Rob’s intro [00:00:00]
Rob Wiblin: Hi listeners, this is the 80,000 Hours Podcast, where we have unusually in-depth conversations about the world’s most pressing problems, what you can do to solve them, and whether there’s a horse version of me in a parallel universe hosting their own podcast. I’m Rob Wiblin, Head of Research at 80,000 Hours.
Today’s episode is an exploration of one of the most fascinating ideas out there — the many-worlds theory of quantum mechanics.
I was lucky enough to speak with David Wallace, who wrote The Emergent Multiverse: Quantum Theory According To The Everett Interpretation, and who just might be the best person in the world to interview on this subject.
I had so many burning questions about what actually happens and doesn’t happen if many-worlds is true, like:
- Is there a world where I become president of the USA?
- Is there a world where I murdered someone this morning?
- And is there a world where I’m a horse or some other animal?
I also wanted to get to the bottom of what listeners were most curious about: does this have ethical implications for how we should live?
So I also asked David questions like:
- Are our actions getting more or less important?
- Whether the branching of the universe creates more goodness by there being more stuff
- And whether there’s any way that we could conceivably influence other branches of the multiverse
After our initial recording with David, some early reviewers — even some with physics backgrounds — found the opening hard to follow. So if you notice a change in the audio, that’s because we went back and recorded a new introduction to Quantum Mechanics to go near the beginning.
I’m not gonna lie — it’s still not the easiest thing in the world to follow. So if you find it challenging, you won’t be alone! But we’ve made an extra effort to add links throughout the transcript on the website to make sure you can get explanations for all the most important concepts.
On the website we’ve also linked to some video explainers of quantum physics which you can check out and which have an easier time of it because they can visualise things for you.
The good news is that even if you don’t get all the technical details a lot of the subsequent discussion should make almost complete sense.
Later in the conversation we go on to talk about:
- Why Quantum Mechanics even needs an interpretation
- Whether we can count how many worlds there would be
- The debate around what QM is, and why a consensus hasn’t emerged
- The progress in physics over the last 50 years, and the practical value of physics today
- The philosophy of time
- And much more besides
Alright, without further ado, I bring you David Wallace.
The interview begins [00:02:15]
Rob Wiblin: Today I’m speaking with David Wallace. David is a professor of history and philosophy of science at the University of Pittsburgh. Before transitioning into philosophy of physics, David completed a PhD in theoretical physics at Oxford University. David is one of the world’s foremost experts on the philosophy of quantum physics. In particular, he is known for developing and defending the Everett interpretation of quantum mechanics, often called the ‘many-worlds interpretation.’ To that end, he published the book The Emergent Multiverse in 2012, which David Deutsch called, “An outstanding achievement, and the current state of the art in the Everett interpretation.” He has many other interests in philosophy of physics as well, including quantum field theory, statistical mechanics, and general relativity. Thanks so much for coming on the podcast, David.
David Wallace: It’s great to be here, thanks for the invitation.
Rob Wiblin: I hope we’ll get to talk about whether quantum physics has any important ethical implications, and why the correct interpretation of it is a controversial question at all. But first, what are you working on at the moment and why do you think it’s important?
David Wallace: Right at the minute I’ve been thinking a lot about black holes, and about questions about whether black holes are properly quantum mechanical systems. It’s kind of related to a lot of my thinking about the interpretation of quantum mechanics before, because a lot of the reason I was interested in those questions was as a means to an end to really be able to think properly about some of the other deep questions that come up in physics. Almost all of which have a quantum mechanical background. Why is it important? Well black holes are the nearest we’ve got to a window into how a quantum theory of gravity would work. And that’s something near to the holy grail of what we’re looking for in contemporary theoretical physics. And they’re the kind of things that throw up a bunch of interesting conceptual questions, but conceptual questions that link into what physicists are working on themselves at the moment. And that’s the kind of place I’d like to hope that philosophy of physics can make a contribution to physics.
Rob Wiblin: So you’re trying to figure out a theory of quantum gravity, and that’s maybe a way of fixing up the remaining problems with the standard model of physics, some of the parts that are still not quite right?
David Wallace: Well kind of. I mean the problem really with the standard model is it’s a bit too perfect. We know it’s not the last word, we know it’s an approximation that holds only at relatively low energies, and we know that ultimately it’s got to be combined with gravity somehow. But getting direct empirical access to that is a pain. And one of the dirty little secrets of the discovery of the Higgs in the Large Hadron Collider is that discovering the Higgs meant nothing itself, except the Higgs is about the most boring possible thing that can come out of the whole process. So experimentally we’re at a bit of a loss there, unfortunately.
Rob Wiblin: Right. So the hope was that it would turn up more surprising results. They would fix one another and it would show us a new path. But not so?
David Wallace: Pretty much, yeah. There was what looked like a really pretty good argument that at energies not much higher than the energy of the Higgs, then something would go wrong with the standard model. That argument must’ve been wrong, because the standard model seems to be working just fine at energies quite a lot higher than the Higgs.
Rob Wiblin: It’s too good.
David Wallace: Yeah, but we don’t know what’s wrong with that argument. But this is another thing I’ve been thinking of, although it’s difficult to know where to progress from it.
Rob Wiblin: Right. This all sounds terribly like actual physics, but I think you’ve actually decided to work on philosophy of physics, at least recently. How did that come about?
David Wallace: Well really I think my work is kind of interdisciplinary. It’s sort of on the borderland between physics and philosophy. And some bits of it, like the bits I’ve just been talking about really, are really close to physics. Some bits of it sit closer to mainstream philosophy. But really where I came in was I was a physicist, but I was interested in a bunch of these conceptual questions, which were kind of half philosophy. And for relatively mundane reasons about how the academic scene in Britain and America works, it became kind of clear that if I wanted to pursue that as a career, then doing the kind of research I wanted to do in a philosophy department was probably a better move than trying to carry on doing it in the physics department. At least at my early career stage. So I retrained, I did my philosophy PhD. And I have an interest that you’d call core philosophy, but my starting point has always really been that I’m a physics-trained person who wants to work on these questions where physics and philosophy are talking to each other.
Rob Wiblin: Right. Is there much antagonism between physics and philosophers of physics? I saw a slightly mischievous quote about how physicists don’t have any use for philosophy of physics; that it’s “as useful as ornithology is to a bird,” I think someone said.
David Wallace: Yeah I think that’s Steven Weinberg, I can’t remember. Yeah, I’ve seen that quote a few times. I mean Tim Maudlin pointed out that ornithology is more useful to birds than you might think. But I mean putting that aside. Whoever the quote was from, what they had in mind was something like, the passive study of the method of science is not very useful to scientists. And maybe that’s true, maybe it’s not. I think the replication crisis, in very different bits of science, is a good wake-up call to say that sometimes you do need to pay attention and be reflective about the methodology of science.
David Wallace: So there are places where I think, even on its own terms, that would be too quick. But the kind of things I’m interested in as a philosopher of physics tend to be less about that kind of “What’s the method of physics,” and more about the kind of conceptual questions that come up in physics. And in that space sometimes you get hostility, but you also get quite a lot of the conversations. I think there’s a view among some physicists that philosophers of physics, or philosophers more generally, sort of pontificate about things which really turn on facts about physics without knowing anything much about the physics.
David Wallace: This is not an entirely incorrect criticism, I should say, but it’s also the case there’s a lot of people who I think, I hope, increasingly do try to understand the physics, and do really try to do work that crosses the boundary. And by and large, I think physicists are willing to have interesting conversations with anyone who’s working on these things, if they kind of know what they’re talking about. I think often I have an initial interaction with physicists where there’s a sort of skepticism that you do know what you’re talking about, and it goes away relatively quickly when you get technical. And I sometimes joke that the most valuable thing to me about having a PhD in physics is that I get to truthfully say, “I have a PhD in physics.”
Rob Wiblin: Yeah I guess that adds a bunch of credibility.
David Wallace: Yeah, to some degree.
Introduction to quantum mechanics [00:08:10]
Rob Wiblin: All right, so before we go on any further, I think it might behoove us to give the audience, and indeed me, a little bit of a reminder of what quantum mechanics is all about, because it is a somewhat challenging and confusing topic that is easy to get a little bit muddled in your mind. I’ve gone back and read some Wikipedia articles for this, but I’m not sure I 100% understand it, or I’m not sure I ever did. Yeah, first off, what is quantum mechanics?
David Wallace: Well, quantum mechanics is our best theory of the very small — of atoms and molecules, of the subatomic particles that make them up. But because big things are made of small things, then quantum theory is really our best theory of pretty much everything. So it underpins most of modern physics, from scales right down to the scale of the Higgs boson that gives master particles that we try to look at in the Large Hadron Collider in CERN, all the way up to the quantum fluctuations in the early universe that give rise to the structure of galaxies on the largest scales. And it’s relied on by physics at every scale in between that. Computers will do as just one example where we need quantum mechanics to understand how their components work.
Rob Wiblin: And what is so distinctive about it? I suppose, you know, in Newtonian mechanics, the things keep going, they bounce off of other stuff. That feels all very intuitive, but quantum mechanics has a different flavor to it.
David Wallace: Radically so, yeah. The simplest way to get at the difference is to say that, in quantum mechanics, if any object can have one or other property, then it can somehow have both properties at the same time. So, particles aren’t just on the left or on the right. They’re on the left and on the right at the same time. They’re not just spinning this way or spinning that way. They’re spinning this way and that way at the same time.
David Wallace: And then if that was just confined to the microscopic, then that might be okay. Maybe we could just say, “Look, we’re evolved plains apes. Natural selection didn’t suit us to intuit the very small. Maybe we just need a new language to talk about it.” But anytime you engage with this indefiniteness, this “two things at the same time” — what physicists call “superpositions” — then the multiplicity kind of infects the system that’s engaging with it. So if I’ve got one particle that’s in two places at the same time, and I scatter another particle off it, now the other particle would bounce differently if the particle is in one place than if it was in another place. So now, suddenly the scattering particle is doing two things at the same time.
Rob Wiblin: Right, so at its core, it’s got this oddity that it seems like, at the very small level, particles, I guess electrons, very small things, can be in these superpositions where in some sense, there’s fundamental uncertainty, or it’s just like there’s no fact of the matter perhaps, about where it is or what charge it has yet. But that doesn’t confine itself to the microscopic level, that infects the larger scale as well, even though we don’t see things in multiple positions, and I’m not even sure exactly what that would look like.
David Wallace: Right, exactly. So there’s a temptation to try to think about this “two things at the same time” as just uncertainty or lack of information, that somehow this is just a fancy way of saying, “Well, it might be one thing. It might be another thing. We don’t know which one.” The short answer is that doesn’t work. To do all the explaining work quantum mechanics does, it needs a phenomenon called “interference,” where the two possibilities can reinforce or cancel out in ways that wouldn’t make sense if these were just probabilities. And you’re absolutely right that what really matters is the extent to which this stuff doesn’t stay microscopic.
David Wallace: So, at least according to the theory, if a scientist makes a measurement to say, “Where’s the particle?” and the particle is in two places at the same time, the theory tells us that the measurement device predicts two results at the same time. If the measurement device was an old-fashioned pointer, and it was supposed to point to the left if the particle was on the left and point to the right if the particle was on the right, then according to the theory, if the particle was on the left and on the right at the same time, then the pointer is pointing to the left and to the right at the same time. And that doesn’t just sound unintuitive — that sounds ridiculous.
Rob Wiblin: Yeah.
David Wallace: Maybe we can’t even understand what it would be, but at any rate, that’s not what we see pointers doing.
Rob Wiblin: Right. It is radically counterintuitive, and yet it is kind of the consensus theory. It is what we work with. So yeah, why are we kind of compelled to believe something like this? What couldn’t we possibly explain otherwise?
David Wallace: Transistors, DNA, most of modern chemistry, all of particle physics, why nuclear weapons work, why the sun shines, why the galaxies are where they are. I mean, name me a phenomenon in physics you’ve heard of, and I’ll tell you that quantum mechanics is probably needed to explain it.
Rob Wiblin: Right. I guess, is there a simple archetypal example? I suppose the archetypal thing that people found hard to explain was sending… was it electrons through a slit, and then it seems like two different things can happen, but these two different possible outcomes kind of interfere with one another in this peculiar way, even though neither one of them has happened, but both of them has happened.
David Wallace: Good. Exactly. I mean, this is kind of easier to do in a non-audio medium, but to try to visualize it, you might imagine I’ve got a situation where my electron can go along one of two paths, and having gone along one of two paths, it can then, again, choose one of two paths to go along. So, you can think that the electron’s going to the left or to the right, and then the electron beams cross, and again, they can choose to go to the left or to the right. What you find is, if you send electrons… so originally they can only go along the left path. Then, at the second choice, half of them go to the left and half of them go to the right. Likewise, if you send the electrons only along the right-hand path, half of them go to the left, and half of them go to the right.
David Wallace: So, you’d naturally think it doesn’t matter which path the electron goes along. It’s 50% likely to go left at the end, 50% likely to go right at the end. That’s not what we find. We find, if you set the experiment up right — and these are not easy experiments to do with electrons, but they can be done and they have been done — then you could set things up so that, eventually, all the electrons end up on the left, or all the electrons end up on the right, or whichever you like, anything in between according to how carefully you tune the experiment. And what we say in the jargon of quantum mechanics is that the electron path to the left interferes with the electron path to the right. That interference cancels out the possibility of the electron coming out one place at the end. It reinforces the electron coming out the other end.
David Wallace: And it doesn’t work to think that what’s going on is just half of the electrons go one way, and half of the electrons go the other way, and the half that go one way interfere with the half that go the other way, because we could do the experiment with one electron at a time. Again, that’s even harder, but it’s doable, and the results don’t change at all. So, what we seem to have to say is somehow the electron’s going along the left-hand path and the right-hand path at the same time, and the version of the electron going along the left-hand path and the version of the electron going along the right-hand path interfere with each other.
Rob Wiblin: Okay. If you were sending lots of electrons down this path and then looking at the results, then you could say, “Well, these different electrons have interfered with one another and created this funny effect.” But the fact that you see this result even when you’re sending down just one, it’s like, “What’s it interfering with it?”
David Wallace: Exactly. That’s right.
Rob Wiblin: Basically, the answer has to be, the other version of itself, another path that the electron itself took.
David Wallace: Exactly. Yes.
Rob Wiblin: There’s no other option.
David Wallace: That’s right. Yeah. Since we’re talking about the many-worlds theory in this podcast, and the many-worlds theory is controversial, I just want to stress that this bit of quantum mechanics is not controversial. What words you might use to describe it can depend on your take on quantum mechanics, and I’m probably talking about it in a many-worldsy kind of way. But the basic underlying idea, the formalism of our theory’s got to represent an electron going along the left-hand path and the very same electron going along the right-hand path at the same time — that’s just commonplace in modern physics. Everyone accepts that.
Rob Wiblin: I guess one way that people put this sometimes is that, I guess, particles can both be… or electrons can both be particles and waves. Is that kind of looking at these two natures? That sometimes it’s clear where something is, what path it’s gone down, and other times it’s ambiguous and has this kind of wave-like property?
David Wallace: Kind of, yeah. I mean, the sort of classic way of talking about particles and waves is right in its way, but it can be quite tricky to map onto the way we talk about quantum theory now. The best way to map it is to say that, when we talk about a particle as a wave, what that wave is constituted of is really the particle being in lots of places at the same time.
David Wallace: So a classic way of talking about interference wouldn’t just be my little stylized model of the electron only having two paths to go down — it would be a case where the electron can go any way it likes. So I’ve got a source of electrons, and then some distance from that source, I’ve got a screen with some sort of holes in it through which the electrons can go. And between the source and the screen, the electron can go anywhere it likes. So in that situation, it’s not just that the electron is in one position and another position at the same time. It’s kind of, at the same time, everywhere in space between the source and the screen.
Rob Wiblin: Right.
David Wallace: And the way we describe that everywhere in space is, mathematically, by means of a wave, and that’s what we mean by saying there’s a wave description.
Rob Wiblin: Yeah. I guess sometimes, for simplicity, we talk about, “Oh, it could go left or right, or it could be up or down, or spin left, spin right.” But you’re saying, I guess… I didn’t study physics, but I did study a bit of chemistry, and we would always talk about electron clouds, and it was like, we would talk about… So you’d have this kind of donut shape, and you’d say, well I guess in chemistry at least, we’ll talk about it as like, “Oh, this is the probability distribution of where the electron is.” But I guess you’re saying… Well, to begin with, that cloud extends indefinitely in every direction. It just gets extremely improbable for an electron to be very far away from the place that it… from the center of where it ought to be.
David Wallace: Yeah.
Rob Wiblin: I guess it doesn’t just have kind of discrete outcomes, like one or two. There’s outcomes everywhere in between. It could be 1.11, 1.12… but yeah, so there’s kind of an unlimited number of outcomes if you look at, funnily enough. How should I make sense of that? It seems very odd.
David Wallace: Yeah. Okay, good. You’re absolutely right that, when we talk in chemistry or chemical physics, we talk about an electron cloud. And sometimes we try to talk about that like probability. We want to say the cloud measures how likely the electron is to be at a certain place. But by itself, that doesn’t work. If you just suppose that we were ignorant of where the electron was, that wouldn’t work for chemistry. The probabilities come in because if we look to see where the electron is, we’ll find it in a definite place. And the quantum theory tells us what those probabilities are, and the cloud can be understood as telling us those numbers. But we can’t get rid of the cloud and just keep the probabilities, because the cloud’s doing a lot of other work. The cloud is doing things like modeling how the energy of the atom is stored.
David Wallace: Again, as you know from your chemistry, if, instead of measuring where the electron is, I measure how much energy the electron is, then I find that’s, in the classic sense, quantized. It comes in discrete chunks, and each of those chunks corresponds not to a particular place the electron could be, but to a particular cloud, a particular superposition of different places the electron could be. Different superpositions of where the electron could be have different energies. And the higher the energy is, the more the superposition is concentrated on electron states that are quite a long way from the nucleus of the atom. But in all those states then, they still have the possibility of the electron being close to the atom or far from the atom.
Why does quantum mechanics need an interpretation? [00:19:42]
Rob Wiblin: Yeah. This all seems pretty weird. I guess, what aspect of the weirdness kind of calls for interpretation, so to speak? Because people don’t talk about an interpretation… Well, maybe they do, but I don’t hear people talk about an interpretation of gravity or an interpretation of inertia in the same way.
David Wallace: Right. Yeah. That’s right. I mean, those raise philosophical problems, but they’re less urgent. And the real thing that calls for interpretation is this question about what happens when the “two things at once”-ness, the superposition, gets up to the macroscopic scale.
Rob Wiblin: Yeah.
David Wallace: Because we don’t seem to see pointers in two places at once. If we borrow Schrödinger’s famous morbid example, then we don’t seem to see cats alive and dead at the same time. So, interpretation kind of understates the problem. What we need is a way of making sense of the theory, and the kind of in-practice way that physicists do it is that, when the superposition is at the scale of the macroscopic, at the scale of the everyday, we stop treating it as a superposition. And then, and only then, we start treating it as probability. So we stop saying the cat is alive and dead at the same time. We say it might be alive, or it might be dead, and these are the probabilities. We stop saying the measurement needle is pointing two ways at once. We say it’s pointing one way or the other, and these are the probabilities. But there’s no good understanding or rule as to when that should happen.
David Wallace: The actual dynamics of quantum theory doesn’t ever transfer us from two things at once to one thing or the other. It’s a deterministic theory. It’s got no probabilities in it, so we kind of end up using a very ad hoc rule to say this is where we decide to treat this as a measurement. Since measurement devices are made of atoms, they obey ordinary physical principles. There’s something pretty problematic, pretty difficult to make sense of in the idea that somehow, just because we choose to call something a measurement device, the very laws of physics change for it.
Rob Wiblin: Yeah, so the fundamental, or least one of the fundamental mysteries is, why does this effect seem to happen? Or why do you get resolution of this superposition when you look at it?
David Wallace: Yeah, exactly.
Rob Wiblin: Given that people looking at something doesn’t seem like it should have any special privileged status within physics, or within philosophy really.
David Wallace: Well, exactly. Yes. Yeah, or looking in. If I look at something, it makes sense that looking at it should affect me.
Rob Wiblin: Yeah.
David Wallace: Otherwise, if my brain doesn’t go into a different state according to what I see, then I haven’t really succeeded in seeing anything at all. But it doesn’t seem that the thing I look at should particularly care that —
Rob Wiblin: That you were looking in its direction.
David Wallace: That’s right. Or, look, it might care because it’s a particular physical interaction. I mean, my looking at something is a pretty nonviolent process, but often in physics, we, scare quotes, “look at something” by hitting it really hard with something else. You would expect that to make a difference, but you wouldn’t expect that to make a difference just because it counts as a measurement.
Rob Wiblin: Right.
David Wallace: If I hit some particle really hard with another particle, what happens shouldn’t depend on whether I was doing it because I wanted to make a measurement, or I was doing it just because I don’t like particles and want to break them.
Rob Wiblin: So I guess I’m not 100% sure how to phrase this question, but I suppose, as we’ll talk about later, on one interpretation, these measurements produce splitting, or like, multiple worlds that result. On another one, things could have gone either way, and then one of them is chosen, and that’s the thing that really exists, and the other ones don’t. But either way, there’s something that qualifies as a measurement or a thing that causes splitting. And that’s easy to understand if you’re thinking about it as like, “It could go left, it could go right, got to do one of the two discrete things.” But given that it’s just actually a smearing across a continuous thing, what counts as a measurement? Or what counts as a different outcome, given this kind of unlimited number of outcomes, really?
David Wallace: So if I’ve got a genuinely continuous process, like I’m not measuring left or right, I’m just looking at where the particle might be. Or maybe if it’s a radioactive decay process, and I’m timing how long it takes to decay. Then we can certainly talk about the superposition, the multiplicity being magnified up to the scale of the measurement device. And in the many-worlds theory, we can interpret that as a branching into worlds. But what it’s telling us is that that branching doesn’t have any built-in discreteness to it. It’s up to us, just descriptively, how discrete we want to make it.
David Wallace: If you were to imagine… Don’t think about the branching structure as a bunch of distinct bright lines radiating off from a point. Think of it as a sort of blurry fan, and we can cut that fan into chunks to analyze if we feel like it. And if we cut the fan into chunks too finely, we’ll find that the chunks are still interfering with each other. So they can’t be treated as distinct, genuinely distinct possibilities where we could see one but not another, but there’s no exact maximal fineness as to how finely you cut it up. There’s a fuzziness as to when cutting it up too finely stops working. And so that tells us that the phenomenon of macroscopic superposition of multiple things going on, even at the macro level, is robust and real. But the way in which we decide to discretize it and to break it up is very often not robust and real — it’s just a way of describing what’s going on.
Rob Wiblin: Okay. So it seems like things have to deviate a particular amount, or the resolutions have to be sufficiently different before you cease to get interference, and you can say these things have split. What is the thing that determines what is enough? It seems like it’s going to be somewhat arbitrary, or is this like another physical constant where it’s like, “Well, once it’s a meter, then that’s enough”?
David Wallace: Yeah, it is somewhat arbitrary. I mean, as always with these things, what’s important to note is what’s making it arbitrary is that we’re using ordinary human language to describe something whose sharpest description is mathematical. And so the arbitrariness is, to some extent, just coming from that misfit between our language — which kind of insists on things being discrete and sharply defined — and the mathematics, which is much more relaxed about fuzziness and differences of degree.
David Wallace: But in any case, yeah, what it comes down to is there are physical processes that lead to a lot of macroscopic variation — processes that give rise to a superposition of all the different times which the atom might decay, or a superposition of all the different results that the detector might show. And those processes don’t necessarily have to be intrinsically discrete, but our way of describing them tends to sort of force that discreteness on it.
Rob Wiblin: Yeah, so, the classic superposition is an electron or an atom or something. But you’re saying it is possible to get significantly larger things than that to be in superpositions, I guess proteins, or big bits of DNA, or something on that level. Which is much bigger than individual atoms, let alone subatomic particles, but nowhere near as big as people. How do you get larger things like that to remain in a superposition? Like do you have to make them extremely cold so they don’t melt?
David Wallace: Well, if you take the theory seriously, if you trust the mathematics of it, then getting things into superpositions isn’t what’s difficult. What’s difficult is getting them to stop infecting things around them with yet more superpositions. So if I want to prepare a big macroscopic object like me in a superposition, it’s dead easy. I just watch a Geiger counter, say. The atom is in a superposition of decaying, at all the times it might decay, so the Geiger counter will be in a superposition of clicking at all the times it might do. But precisely because that superposition is so uncontrolled, I’ll go into a superposition, everything around me will go into a superposition. Then we can’t do the sorts of experiments that distinguish superpositions just from alternate possibilities. To do those experiments, we need to let the superposition expand out from a microscopic system to a slightly bigger system, but go thus far and no further.
David Wallace: And that does tend to involve, as you say, things like keep it cold, keep it isolated, don’t let other things interact with it. If you let another thing interact with a system in a superposition, then the superposition tends to get out and spread to that other thing, and that can get uncontrolled really, really fast. For instance, we often talk about electrons as the things we want to put in superpositions. Actually, if you do experiments with interference, you very often don’t use electrons because electrons are charged, and because they’re charged, they have an electric field, which messes with stuff around them quite a lot. So, it’s really hard to create large-scale superpositions of electrons without the superposition escaping the lab, causing the electron to get entangled with lots of other particles, which get entangled with lots more particles after that. It’s much easier to do superpositions with photons — particles of light — or with neutrons — which are neutral, have no charge — because neither of those particles generate the sort of long-range electric forces that make it hard to keep superpositions under control.
Rob Wiblin: I see. So if I wanted to get something big, conceivably a person, say, to remain in a superposition, I would have to isolate it hella well from everything else — just like stick it in a vacuum, like a pure vacuum far away in some part of the galaxy where there’s absolutely nothing, and make sure it’s emitting no light, so no one can see it, and I guess make sure it’s extremely cold, so it’s not accidentally emitting some particles out that could interfere with anything else. And then, conceivably, a larger object could remain in a superposition as a whole.
David Wallace: Right, but notice it’s not that it’s hard to get the person into a superposition —
Rob Wiblin: But to keep them there.
David Wallace: — like I said, that’s dead easy. And it’s not hard for the person to stay in a superposition. What’s hard is to keep it the case that only that person is in a superposition.
Rob Wiblin: Ah. Oh, rather than other things around it as well.
David Wallace: Exactly. That’s right. So if the person in a superposition emits some particles, that doesn’t stop them being in a superposition — it just means that the particles are in the superposition too. Then the thing they hit is in the superposition too. And pretty soon, the whole forward light cone of the universe around the person is in the superposition too. So it’s keeping the superpositions locked away that’s hard.
Rob Wiblin: So in a sense, it’s easy to keep an extremely tiny thing in a superposition, and it’s easy to keep the whole universe in a superposition. It’s just the things in between that are a bit of a trickier state in practice.
David Wallace: That’s exactly right. The truth is you’ll never do it with people.
Rob Wiblin: Yeah.
David Wallace: Doesn’t matter how good that box is, gravitational radiation will suffice to break the superposition. People are just too big and complicated. It’s kind of quite closely related to these arguments that people might have heard in the physics of heat, where in principle, if I take a glass of water and I put an ice cube in it, and I let the ice cube melt, in principle, it ought to be possible to run the whole process backwards so that the water warms up and the ice cube reappears. But as a practical matter, there’s no possible chance you’ll ever be able to do that, because it would involve a totally unmanageably fine control over all the various microscopic bits of the water and the ice. This is just the same — it’s actually at a technical level very closely related. The kind of fine control you’d need over every last constituent of the human-sized object in a superposition isn’t viable.
Quantum mechanics in basic language [00:30:37]
Rob Wiblin: Yeah. Okay, so I’m going to try to say back what quantum mechanics is in my head in the most basic language, and see how not wrong I can get it. So all kinds of objects, it seems like they can be in multiple different states simultaneously, and we can study this more easily and understand it more easily at the microscopic level. And there we see that even a single tiny primitive thing, it seems like it has to be able to be in multiple different states simultaneously, because we can observe those different positions it could be in interfering with one another, as if they were both there messing with one another.
David Wallace: Yes.
Rob Wiblin: And basically, that then scales up to causing more macroscopic effects that might be visible on the level of people, so this bizarreness doesn’t remain confined to the microscopic level. And we’re trying to figure out, how can something be in two states at once? How does it interfere with itself across these different things? This isn’t the kind of physics that we’re familiar with, but we have this mathematical model that we call quantum mechanics, which can explain that and then can explain basically every observation that results from it both at the small level and at the big level, basically. It’s just that it’s super weird.
David Wallace: Yeah. That’s pretty much right. The only thing I’d add to that is that, at the macroscopic level, the superpositions behave a lot like just probabilistic alternatives, like one thing or another, because you can’t do the interference experiments for really large systems. They’re too complicated. So what that means is there’s no way to tell directly experimentally whether suddenly one of these large superpositions just magically stops being in a superposition and just jumps randomly into one outcome or another. That’s something that, arguably, is pretty ad hoc. There’s no justification for it, and every experiment we do on superpositions on larger and larger scales tells us it doesn’t happen on larger and larger scales, but you couldn’t rule out it happening somewhere.
David Wallace: The many-worlds way of looking at quantum mechanics basically says, “Look. We’ve got no particular reason to think that’s happening, and it complicates other equations to no good effect, and we don’t know how to build it into our theories in any principled way, so why assume it happens?” But there are alternatives that either say, “Look. We don’t care that it’s ad hoc. We’re just going to do it anyway,” or that say, “Okay, challenge accepted. Let’s try to come up with a non-ad hoc physical theory that we could test that modifies quantum mechanics to get rid of these effects.”
Rob Wiblin: By superposition, you mean a superposition is this state — this funny thing where an object, an electron or whatever, is in this ambiguous two-thing-at-once state. We call that… The technical term for that is a superposition, right?
David Wallace: That’s right. Yes.
Quantum field theory [00:33:13]
Rob Wiblin: Just before we move on, so quantum mechanics and quantum physics kind of seem like they’re very similar, but I’ve heard of this other thing in researching for this episode called quantum field theory. Is that some sort of extension, more sophisticated version of quantum physics?
David Wallace: Kind of. The best way to think about it is that quantum theory, quantum physics, is a framework in which you can do lots of different bits of physics. So we can have a quantum physics of particles, for instance, a quantum physics of electrons and protons. That’s the kind of quantum physics we do in, say, doing chemistry, and the things we’re talking about are electrons and protons. And because it’s quantum, they could be in multiple states at once, they could be in superpositions. But what they can’t do is, for instance, be created or destroyed, or move at speeds close to the speed of light — not because the electrons and protons can’t do those things, but just because the theory we’re using is just not describing that stuff.
David Wallace: If we want to describe particles in situations where they can move close to the speed of light, if we want to describe particles in situations where they can be created or destroyed, then a quantum theory which has particles as its basic constituents doesn’t work — we need a quantum theory that can handle things like radiation, things like light. We need a quantum theory that can handle the idea that particles might be created and destroyed. That quantum theory is the quantum theory that underlies particle physics, and that’s called quantum field theory.
David Wallace: So quantum field theory is a particular quantum theory. It’s in the big umbrella of quantum theories, but it’s the most important theory under that umbrella. In fact, it’s a little umbrella of its own because there are many specific quantum field theories. In relativistic physics then, the relation between field and particle is quite close, and so when we talk about quarks as constituents of protons, we can also talk about a quark field. In some ways, the relation between the field and the particle is kind of like the relation between the ocean and the waves of water or the surface of the ocean.
Rob Wiblin: Okay, so it’s kind of an extension of using similar thinking to explain a wider range of phenomena. Indeed, it sounds like kind of trying to cover everything under this umbrella.
David Wallace: Yeah. Pretty much all of modern deep theoretical physics is done under the umbrella of quantum field theory. But the thing is, quantum field theory doesn’t change the basic paradoxes of quantum mechanics. The entities that can be doing two things at the same time are stranger, and more alien, and more complicated, but they could still be doing two things at the same time.
Rob Wiblin: I see.
David Wallace: And so the fundamental interpretative questions in quantum mechanics don’t really change when you go from the sort of relatively simple quantum mechanics of a small number of particles to the much more technically complicated realm of quantum field theory.
Rob Wiblin: Okay, so it sounds like, for our purposes here, we can kind of set aside quantum field theory. There’s this extra thing, but it’s not terribly going to affect anything that we’re going to talk about.
David Wallace: Yeah, I think pretty much. Most of the philosophical problems that come out of the interpretation of quantum mechanics kind of look the same at whatever scale you’re looking.
Rob Wiblin: So quantum mechanics, I thought it was kind of a settled thing — people weren’t rewriting the textbook on quantum mechanics. But it sounds like quantum field theory is still an active area of research where people are changing and improving their theories in order to explain more. Is that right?
David Wallace: Yeah, that’s exactly right. I mean, again, if you think about quantum physics as this general framework in which lots of theories are done, then the quantum theory of non-relativistic particles is pretty solid and pretty stable. The quantum theory of electrons and light is pretty stable. It’s been stable since the 50s. The quantum theory of the Higgs boson, and the electroweak interaction, and nuclear force of things, that’s been stable since about 1980.
David Wallace: But what happens beyond the Higgs boson, we don’t know. Maybe we’ll get clues from the Large Hadron Collider. What’s going on when you try to understand how gravity comes into this picture at incredibly small scales in the very early universe, that we hardly know at all. But all of our theories trying to do this, or pretty much all of them, are still quantum theories. So the standard model of particle physics is another quantum theory. The particle physics theories people try to work out to go beyond the Higgs boson, that’s a quantum theory. String theory is one more quantum theory.
Rob Wiblin: Okay. I think that’s probably enough explaining quantum mechanics for now. I think some people would say that that’s not a fully comprehensive and entirely satisfactory explanation of quantum mechanics. I suspect you probably couldn’t pass your physics exams based on that.
David Wallace: Oh, God. Right.
Rob Wiblin: But hopefully it’s enough to continue forward.
David Wallace: Well, I mean, there’s no way around this. At some level, the fully comprehensive one requires mathematics.
Rob Wiblin: Right. Yeah.
David Wallace: You’re never going to completely bypass that, and just like in chemistry, we can chat about orbitals, and energy levels, and electron clouds, but as we both know, if you actually want to talk about the energy levels of hydrogen, there’s no shortcut to doing some of the mathematics of it.
Rob Wiblin: Yeah, so we’ve got multiple problems here. One thing is that me and probably a lot of the audience perhaps are not quite up to scratch on their mathematics to understand this at such a formal technical level. The other problem is audio does not super lend itself to doing mathematics.
David Wallace: Yeah.
Rob Wiblin: Probably we’re going to need a visual medium, like a piece of paper or YouTube video.
David Wallace: Yeah. I mean, I’m trying to sit on my hands at the moment because the temptation even talking audio is to try to gesture and do that kind of thing to help visualize it, but —
Rob Wiblin: Yeah. Alas, gesticulation will not help us here. But yeah, for this exact reason, we’re going to stick up a link to a more comprehensive audio podcast version that goes into what we’ve been talking about in a bit more detail and maybe a bit more slowly. We’ll also stick up a link to a YouTube video where people who want to see a visual thing about, what is this electron cloud and what are these slit experiments with the electrons? For those who haven’t seen that one already, I think it could be well worth going and checking that out, although I suspect that you’ll be able to follow what comes next, even if you don’t.
Different theories of quantum mechanics [00:38:49]
Rob Wiblin: Alright, there’s a bunch of different theories and models that people have put forward. And I guess I’m interested to have you go through kind of briefly each of the main ones, at least the top three or four, and explain what their general take is.
David Wallace: Sure. I mean, I think you could really break it up into answers to that question. I mean, one way of answering that question is to really double down on the idea that observers and measurements are a core part of what your theory is, and really rethink what a philosophy of science should be on that basis. And the other branch of approaches, which are closer to the kind of thing I work on, and which most people in philosophy of physics have thought about, tend to say, “Well, you’re exactly right. There shouldn’t be any fundamental role for measurement or observation as a primitive in my theory. And so if there is, then we better reformulate the theory.” So what you get in that first set of approaches are very much an attempt to rethink science, generally, really, but quantum mechanics in particular. Not as some description of a sort of third-party describable world, in which humans are just more physical systems, but to rethink science as something that just fundamentally is a way of describing how the world seems to me, the observer.
David Wallace: And some of these things border on solipsistic, they start being very much a picture of, “Well, okay, I am the observer. I am the person the theory’s about.” You Robert are just more physics. And so physical predictions about you are relevant to what I the observer see. I’m not giving a very kind description of these frameworks. There are more sophisticated things to say in their defense, but ultimately there’s a fairly common criticism and philosophy that says, “Look, there’s something wrong if your theory does put this fundamental role in for observers. It’s getting in the way of the idea that humans are physical systems obeying physical laws, and that measurement processes are just a subclass of physical processes that we decided to stick a measurement sticker on to.” If you look at strategies to try to remove the role of the observer, there’s kind of two directions you can go. One is to say, “Well, the theory as written absolutely has a role for the observer. So much the worse for the theory as written, let’s change it.”
David Wallace: So if the way you’re thinking of this is that quantum mechanics predicts that systems enter these strange superpositions, both things happening at one state, the cat’s alive and dead at the same time, if that’s what the theory predicts, then the theory is wrong. It needs to be changed to a theory that doesn’t predict that. And no amount of double-talk about how the observer changes things is acceptable. And so these are the approaches that say something like, well either the collapse of the wave function, the transition from alive plus dead to just alive, needs to be written in as a bonafide respectable dynamical law, and not relate to observers at all, or, alternatively, something else needs to be added to the theory, some hidden variables whose real job it is to describe whether the cat’s alive or dead. And the quantum description is incomplete in that sense.
Rob Wiblin: I see. Yeah. So maybe we should put some names on these different categories, so people can track them a bit more easily in their head. So you’re describing non-many-worlds ideas here that are trying to make more respectable the idea that well there is only one world, and the question is like, “How do we go from the uncertain situation to having one world that’s been singled out that actually happens?”
David Wallace: The first kind of line I was talking about, where you want to bring the observer up to some preferred status, those kinds of strategies come under the kind of general label of ‘Copenhagen interpretation.’ You can get very fine-grained about the naming, but basically, the Copenhagen interpretation is that kind of approach. Approaches of that kind share at least a certain amount of DNA with the Copenhagen interpretation. The kind of mess-with-the-physics approaches I was talking about generally get labeled as ‘dynamical collapse theories,’ that try to change the equations of quantum mechanics so that these weird superpositions, live plus dead states, don’t survive, don’t remain. Or hidden variable theories, theories that want to add some stuff to quantum mechanics so as to have that stuff represent the definite goings on and say that the quantum state itself is not doing the representing.
David Wallace: I tend to label strategies like Copenhagen as change-the-philosophy strategies, and strategies like dynamical collapse and hidden variables as change-the-physics strategies. And one of the observations you have is that it’s mostly physicists who like the change-the-philosophy strategies, and it’s mostly philosophers who like the change-the-physics strategies. And the last class of strategies, which I think is what the many-worlds theory exemplifies, basically says, maybe you don’t need to do either. Maybe if you think hard enough about that alive plus dead cat, you’ll realize that it’s not as incompatible with what we observe as you think it is.
Many-worlds theory [00:43:14]
Rob Wiblin: Yeah. Do you want to flesh out this option of ‘don’t change the philosophy and don’t change the physics, just accept that the world is much bigger or perhaps much stranger than we thought’? I guess this is the many-worlds theory, or the Everett interpretation, they’re kind of two names for the same thing.
David Wallace: Yeah. So look at it this way: Quantum theory predicts that cats are alive and dead at the same time. And our immediate response is, “That couldn’t possibly be true.” Why not? Well, I’ve seen lots of cats. You’ve seen lots of cats. We’ve never seen them alive and dead at the same time. What would it look like to see a cat that was alive and dead at the same time? I think your general impression is that it would be sort of like being really drunk, seeing double or something.
Rob Wiblin: Right, yeah.
David Wallace: That’s your intuition as to what it would look like if you saw a cat that was alive and dead at the same time. But intuition’s a lousy way to predict what you’ll see in a physical theory. There’s a lovely, almost certainly apocryphal, too-good-to-be-true story about Wittgenstein, the philosopher. So supposedly he’s crossing the court in Cambridge with a colleague, and he sort of stops suddenly, as Wittgenstein is wont to do, and says, “Why was everyone so resistant, so surprised by the idea that the earth went around the sun?” And his colleague said, “Well, because it looks as if the sun goes around the earth.” And supposedly Wittgenstein thinks for a minute and says, “Well, what would it have looked like if the earth went around the sun?”
Rob Wiblin: …the same.
David Wallace: Exactly. Yeah. Because this is how it looks, and the earth does go around the sun. And so what was really going on in that kind of, “Well, it doesn’t look like it” is something like our intuition of what it would look like if the earth went around the sun is not the same as how it actually looks. So our intuition is that the sun would be essentially whizzing past, and we would seem to be flung backwards onto the earth by the force of our acceleration, or something. But if you actually ask the physics what it would look like, you realize those things wouldn’t happen. Those are bad intuitions about what being on a moving planet is like. And so similarly, in the quantum case, ask the theory what it would be like to see a cat that’s alive and dead at the same time. You don’t get the seeing double answer; you don’t get the being drunk answer. You want to think something like this: If I saw a live cat, I’d go into a state that you might describe as seeing a live cat.
David Wallace: And if I saw a dead cat, then I go into a state you might call seeing a dead cat. So if I see a cat that’s alive and dead at the same time, according to the equations of quantum mechanics, then I go into a state which is seeing a live cat and seeing a dead cat at the same time. And if I tell you about it on this podcast, then you go into a state of hearing David say the cat’s alive, and hearing David saying the cat’s dead at the same time. And when people listen to the podcast, everyone who hears it goes into this mixture of hearing you reporting the cat’s alive, and hearing you reporting the cat’s dead at the same time. And in a pretty short order, the whole planet knows the cat is alive, and everyone knows the cat is dead at the same time. And those two bits of the theory aren’t talking to each other anymore, these are sort of separate strands of reality inside the quantum state.
Rob Wiblin: So the many-worlds theory says that, in the Schrödinger’s cat case, you do see the cat as alive and the cat as dead. Both things happen, and the world splits into different timelines based on which one you end up in, or which thing happened in your part of the wave function. People sometimes think of the word as splitting each time there’s a quantum event like that. But you think that’s not quite the right way to picture it. What’s the right way of imagining this kind of branching tree structure?
David Wallace: Well I don’t think splitting does a bad job of describing it, but you have to understand that all of that is non-fundamental. It’s not that there’s some new fundamental law of physics that says, “Suddenly the world is split.” It’s rather that if you look at what the actual laws of physics tell you, you started off with the world being structured to represent one set of goings on, and then it changes in a way that now it’s structured to represent two sets of goings on. If you shine a light through a partial mirror and you originally had one part of light, and when it hits the mirror one part of light goes off in one direction one goes off in the other direction, did something split? Yeah. But not as a matter of fundamental law. It’s just that the natural way to describe the underlying goings on is that I have two parts of light rather than one part of light.
David Wallace: And so similarly, in the quantum case, the natural way of describing what’s going on is, before the measurement, things were structured in the way of there being one classical world. And after the measurement, there were two parts of disconnected bits of structure in the world. And one of them describes the live cat world, and one of them describes the dead cat world. And then you can layer various bits of metaphor. And then if you want to say, “Well, actually there was a vast number of worlds and they differentiated one from another,” David Deutsch has that way of talking, for instance, you can do that. If you want to say, “Well, the world split,” you can do that too. But the physics doesn’t care. It’s just a way of talking about a higher level of differentiation appearing in the underlying equations. And none of it is fundamental; there’s no completely sharp notion of how many worlds there are, for instance.
Rob Wiblin: I see. So as I understand it, at the point where you take some quantum measurement and these two branches start to move apart from one another, very early on they can interact somehow, but then they kind of stop interacting gradually as they diverge and become more different in how exactly they look. Can you explain that?
David Wallace: Pretty much, yeah. The way in which [superpositions] (https://en.wikipedia.org/wiki/Quantum_superposition#:~:text=For%20an%20equation%20describing%20a,to%20obey%20the%20superposition%20principle) —states that are doing two things at once — can interact is what we call ‘interference.’ So in a certain sense, the two-slit experiment — you can talk about it in terms of we have a split into two worlds, but then the worlds interfere with each other and sort of reemerge together. That’s a perfectly legit way of talking, even if it’s a little bit metaphorical. But that kind of interference, that way in which the presence of multiple worlds shows itself, gets suppressed very quickly when the scale of splitting gets large. And the reason for that basically is you could only ever see the interaction between worlds if you can step outside. It’s not as if I within my world can do an experiment that I describe as interacting with another world. There are experiments that somebody outside the whole branching process could do, which are well described as causing the worlds to interfere.
David Wallace: But the very fact you can talk that way means the inside description becomes less legit. The inside description, where I can think about myself in a world as isolated from other things, only really kind of works as a description if this interference isn’t present. And yeah if you try to build a quantum computer or something, one way of thinking about it is your job is to keep the splitting inside the box, so that you can mess with it and bring it back together. That’s hard to do. And normal events, like looking at a cat, tend to cause things to split in ways that are pretty radical, pretty irreversible. So only gods or aliens could cause the worlds to interact to that point.
Rob Wiblin: Yeah. Okay. So naively you might think well, just after one of these divergences, if these worlds are still able to interact somewhat and have interference, well you’d be able to detect the other world and show that the many-worlds interpretation is correct, because you’d be able to observe these other worlds affecting you. But is it that precisely because they can interact, they haven’t truly diverged yet, and so there’s kind of a contradiction here that you can never see something that you can’t see?
David Wallace: Sort of. I mean, it’s sort of a semantic contradiction. You could say that by definition what I mean by a genuine branching is that it’s irreversible. I can’t bring it back round again. But if you want to step away from the semantics, the real question is how big a superposition can you get, and still have it visible through interference? And there’s no hard logical upper limit to that. I mean, as a practical matter, if you try putting humans in superpositions, there’s just no chance you’ll keep the system controlled enough. But people have put buckminsterfullerene molecules in superpositions, they’ve put bacteria I think in superposition states. Arguably the night sky is a relic of superposition states that span the cosmos, depending how you think about it. Do you have to describe that stuff in many-worlds terms?
David Wallace: Well, that’s contentious. Could it be that even though we see all these superpositions on all these various scales, nonetheless, when it really gets to the human scale, some new process kicks in and destroys all but one of the superpositions? Well again, you can’t rule that out empirically. But the sort of basic underlying idea that quantum mechanics supports superpositions and interference between superpositions, and entanglement that causes the superpositions to spread out, that’s something that’s just been verified sufficiently thoroughly in experiment by now that it’s not really in serious doubt. You can fight about whether ‘many worlds’ is the right language to describe it. And you can fight about what it means for humans, but you can’t really fight about the fact there’s superpositions going on at that scale. I mean people fight about everything, so I have colleagues who disagree.
Rob Wiblin: But mostly. Okay. And a superposition is when you’ve kind of introduced some uncertainty, some ambiguity about the position of something. And you’re saying, you can do that. You can kind of begin to diverge and then prevent that divergence from happening. And that’s putting something in a superposition and then bringing it back?
David Wallace: Yes, except I wouldn’t say that there’s uncertainty, because uncertainty implies there’s something you don’t know but you could know. It’s not that you don’t know whether the particle is on the left or on the right, it’s that it’s somehow on the left and right at the same time.
What stuff actually happens [00:52:09]
Rob Wiblin: Yeah. So a different line of questioning here. I guess, under the Everett interpretation, what stuff actually does happen? This is something that I’ve kind of tried to get to the bottom of while preparing for this, and it seems very hard. Because from one point of view you might think, well, everything happens, but what is everything in this branching structure? Is every configuration of matter achieved in some way? Or is it something much narrower than that?
David Wallace: Yeah. Okay. I mean the boring quick answer is anything that you thought had a probability greater than zero, according to quantum theory, happens in some branch. But if you then want to interrogate that and ask, what does that mean? Well, are there branches in which you can fly like Superman? No, flying like Superman is physically impossible. Are there branches in which you roll a dice a million times and you get a six every time, yeah, absolutely. That’s highly improbable, but there’s nothing stopping that from happening. Is there a branch in which the charge of the electron is different from what it is? We don’t know. Because our current physics doesn’t tell us whether the charge of the electron is a fundamental thing that’s just written into the laws of physics, or whether it’s actually something a bit more parochial that will come out of some deeper physics. If it’s a bit more parochial, there’ll be some branches where there’s one charge of the electron and some branches will be another charge. If it’s fundamental, then the charge of the electron will be the same everywhere.
Rob Wiblin: Right yeah. Is there a path where I’m U.S. president, constitutional requirements notwithstanding? How do we look at a particular scenario and say whether it’s compatible with the laws of physics or not?
David Wallace: My quick guess is probably no. But it’s slightly delicate, because there are probably configurations of incredibly implausible worlds that have the same shape as the configuration in which you became U.S. president. Because of some ridiculously unlikely but not completely impossible series of little fluctuations or disturbances. But is there a history of things happening in which you became U.S. president? I’d be pretty surprised, because I don’t think the events that caused you not to become us president are well characterized as pieces of random chance. I mean, try this as a slightly more mundane example. I mean, it’s a mundane but kind of morally charged example. Is there a branch in which I decided to shoot someone this morning? I don’t own a gun, let’s pretend I own a gun. I hope the answer’s no, because I’m pretty sure I’m not the kind of person that will randomly shoot someone.
David Wallace: And I’m pretty sure that’s not a matter of random fluctuations in my brain. I don’t think it’s like if I walk past someone there’s then a random quantum chance that I shoot someone. There are people out there who when they walk past somebody, have a random chance of killing them. They’re psychopaths with very severe illnesses of various kinds. Ordinary people basically are not in that kind of category. That’s something we’d normally call physically possible, there’s no law of physics that prevents me from shooting someone, but equally the fact that I didn’t shoot someone is not a matter of random chance. So there isn’t a branch in which I shot that person.
Rob Wiblin: And the reason it’s not a matter of random chance is just that there’s no number of random changes at the quantum level that would be sufficient to drive you to do that, because there’s too much error correction, I guess? Or if anything began to do that, then it just wouldn’t be sustained because there’d be factors pushing back against it?
David Wallace: Yeah. Something like that. I mean, there’s probably a non-zero chance of some amazingly unlikely series of fluctuations in my neurons, such that they all fire in such a way that my arm does move and pull the trigger. But I wouldn’t call that kind of thing me deciding to shoot someone. That’s more like a freak muscular spasm or something. Yeah. I mean, this is in some ways as much a philosophy of mind point as a physics point. I mean, to decide to do something is to have reasons and there to be the kind of high-level processes in your brain that count as forming reasons and intentions and acting on them. You can plausibly believe that some of the process of doing that is chancy. I mean, the fact that I decided to wear a blue shirt rather than a white shirt this morning was whimsical.
David Wallace: And maybe that whim is explainable in some deep deterministic way, or maybe it’s genuine quantum chance. But one’s reasoned decisions aren’t whims. It’s particularly easy to say, “Shall I kill someone?” It doesn’t seem very plausible that those decisions for reasons are things that are chancy. I mean, it’s a little bit of a guess about how the philosophy of mind and how the psychology will turn out here. But on reasonable guesses, I don’t think there are going to be these branches where you do weird, awful things or something.
Rob Wiblin: Yeah. I’m guessing that you might have seen Rick and Morty, and perhaps the episode where they’re kind of jumping through different worlds. And I think they go to one where all of the people are chairs, and the chairs are pizza, and the phones are people. And then they’re just sitting between them.
David Wallace: I haven’t.
Rob Wiblin: Okay. Right. I guess I was going to ask, is there a world where people are horses, or if there’s people, they look like totally different animals, or is that too far?
David Wallace: Well I mean, is there a world in which the planet has been civilized and developed, and a civilization has been built by a bunch of creatures which are biological descendants of horses? I don’t know, I’m not a biologist. Plausibly. I don’t know whether there’s a plausible evolutionary pathway. But very plausibly there’s a branch of the multiverse in which the dinosaurs didn’t die out, and descendants of velociraptors developed intelligence and so on. That’s extremely plausible. Then it almost becomes a definitional question, do you want to say that’s a world where people are horses? I’ve got to say, look, it’s a really distant world. And it’s a world where humans didn’t have to evolve at all, and horses developed intelligence. I mean, there isn’t going to be a world in which you’re a horse.
Rob Wiblin: Because that will be too different.
David Wallace: Yeah. I mean maybe fantastically distantly across the multiverse, there’s a world in which not only were there horses, but one of the horses is called Robert Wiblin and does podcasts for an organization called 80,000 Hours. And even in this world, the randomness of how language develops is that they’re speaking something that has the same structure as English. But I don’t think I want to say that a fantastically distant descendant of horses is you just being a horse. It’s somebody who, by an incredible…
Rob Wiblin: …just somebody that resembles me.
David Wallace: …somewhere in the multiverse, resembles you. Yeah.
Rob Wiblin: So in order to answer these questions, it seems like you have to go through some process of thinking about what things can be changed through quantum fluctuations. And it sounds like we don’t have a totally unambiguous answer to that. That’s a slightly open question.
David Wallace: Yeah. We don’t have a completely unambiguous answer, but we’ve got quite a good answer to it. And the answer basically goes, quantum fluctuations… There’s three big sources of that. One of them is explicit stuff we do in the lab. We actually do a quantum experiment intentionally. That’s a very rare special case. The second is where random quantum fluctuations get magnified up by some natural process. So here’s a mundane example. If you’ve ever seen a flickering fluorescent light tube, that flickering process is a quantum-mechanically random process. So it flickers differently in different worlds. Here’s a slightly more morbid example: whether a given cosmic ray causes a mutation that triggers cancer in you, that is a quantum-mechanically random process. The third category, and I think the most important for working out which of these worlds happen, is that anything that’s classically chaotic becomes quantum mechanically indeterminate relatively quickly.
Rob Wiblin: Yeah, I see. And I guess, because the brain is error correcting and designed not to be chaotic, precisely because it has this very specific function to serve, that’s one reason why it doesn’t just go off in every different direction.
David Wallace: Exactly. Yeah. The brain does not seem to be a randomizing device of that kind. But the weather is, for instance, chaotic. The butterfly flaps its wings or not, then the weather will turn out differently. So if the butterfly’s in a superposition of flapping its wings or not, then the weather will end up in a superposition of different states. So the weather, we can be pretty certain, is different in different branches of the multiverse. And much more dramatic things like the contingencies of chance that lead to one evolutionary process happening and another one not, hence my reason for thinking there are probably sentient horses and sentient velociraptors, it’s again because there’s enough chaotic processes. And that will just get magnified up to quantum chance.
Can we count the worlds? [00:59:55]
Rob Wiblin: Yeah. In the many-worlds hypothesis, are there kind of countably many worlds or uncountably infinite many worlds, or some higher level of number of worlds? Or is talking about there being a number of things misunderstanding it?
David Wallace: Yeah. I mean, there isn’t really a precise notion of counting it. And the reason for that is because whether you call a bit of the structured quantum universe one world or several very similar worlds is really a matter of how coarse-grained or fine-grained you study it. So if you take something that looks like a single world and you study it on a finer grain, you’ll find it’s actually a bunch of very similar distinct worlds. And as you study at a finer grain, you’ll find even more of that. Eventually you’ll reach a point where those finer worlds are actually interfering with each other and the world language becomes less applicable, but there’s no super sharp level where that happens.
Rob Wiblin: So it’s a little bit like asking how many numbers are there between zero and one or something like that.
David Wallace: It’s not exactly like that, because there’s a pretty determined answer to that. There’s continuum infinity. It’s more like asking, I don’t know, how many clouds are there in the sky. Which collections of wisps count as a cloud? There are obviously wrong answers. If it’s cloudy the answer is not zero. And if your answer is 10 to the power of 30, you did something wrong as well. But there’s a whole bunch of reasonable answers. So by the same token, how many worlds are there? At least 10 to the power of 10 to the power of 20. There’s not a sharp, most sensible answer. It becomes a bit definitional.
Rob Wiblin: So if I think about how many branching events are happening at any point in time, do I have to think about how many quantum observation events there are across the entire universe in any given second? Or is it just how many there are in my local environment?
David Wallace: It’s the latter. So the branching process isn’t instantaneous, it’s constrained… Well, it’s theoretically constrained by the speed of the fastest interaction that spreads the branching. In practice, that’s the speed of light in most cases. It’s never faster than the speed of light. It’s usually not slower. So you can think about those random quantum events causing a branch to spread out into the future from the earth. And then other branch events that are happening in the Andromeda galaxy are causing branching to spread out throughout the galaxy. Two million years after those events happen, we will be branched by them. And two million years after they happen for us they will for Andromeda, but they’re not happening everywhere in the universe all at once.
Rob Wiblin: Yeah. It’s a little bit hard to picture that because you’ve got like branches coming out from every different point and then meeting one another. And so you have two branches that intersect because they’ve now met, the differences. What happens then? Now they split into four I guess.
David Wallace: That’s right. Exactly. And this comes up in some of these sort of classic quantum experiments where you and I share a particle and go to distant planets or something. And you measure your particle, and you get one of two answers. And I measure my particle, and I get one of two answers. So your measurement causes you to branch. My measurement causes me to branch. And years later, when the signals get back between us, your measurement then causes me to branch and my measurement then causes you to branch. And eventually you’ve got four branches, each of which encompasses both me and you, but there’s an intermediate period where there are two local branches encompassing me but not you, and two local branches encompassing you, but not me.
Rob Wiblin: So this is slightly a hard question to phrase, but I suppose, one way of imagining this branching process is that the world is getting bigger and bigger in some sense, as you’ve got the branches multiplying. Another one is imagining that every time this happens, you’ve got the same total size, but it’s just like splitting into ever finer threads. Is that a meaningful question? And if so, which one is it?
David Wallace: It’s a little bit metaphorical, but probably the latter is the better way of talking. But I mean, in a sense, the world is becoming more and more variegated. That’s a way of putting it. It’s becoming more and more highly structured. And a world in which there’s only one thing going on is much, much less structured than the world in which there are a gazillion things going on. There are sorts of measures you can give for how thick each world is in a certain sense, but they’re a little bit metaphorical. They don’t translate to anything like how much energy each world has or something like that.
Rob Wiblin: Okay. Right. What relationship does the Everett approach have to modal realism?
David Wallace: Probably not all that much, but I know smart people who would say differently.
Rob Wiblin: Oh, interesting. Okay.
David Wallace: So if you think about the modal realist idea, it says something like all possibilities, not even the physical possibilities…all possibilities are real. And we’ll use the machinery of this modal multiverse to make sense of various concepts like possibility and causation. You could, if you want, try to use the branches of the Everett interpretation to be kind of concrete surrogates for those modal worlds, but you don’t have to. Suppose you thought that modal realism was right, or basically right. Then you might be inclined to think, well, actually Everett’s an improvement on modal realism. And instead of using these completely causally disconnected worlds, I’ll use the Everett branches instead. But if you were skeptical about modal realism, if you thought that maybe we want to use the language of possible worlds to talk about modality or causation or action or something, but we can think of these as mathematical fictions, then I think you should probably carry on thinking they’re mathematical fictions. The Everett interpretation shouldn’t lead you to change your mind about that.
Why anyone believes any of these [01:05:01]
Rob Wiblin: Okay. We’ve done a bunch of elaboration on various different theories, and especially on the many-worlds hypothesis. What’s the case, either intuitively or empirically, in favor of each of these main interpretations? Why does anyone believe any of them?
David Wallace: There’s no intuitive case for many-worlds theory. Let me start there. It kind of goes back to that sort of, shall I change the physics or change the philosophy route. If you think that there are good philosophical reasons that you want to think about a scientific theory as a kind of description of a third-person reality without having an observer play a primitive role, and if you think that our current physics is really good, then you’re going to be inclined to want to find a way of making sense of quantum mechanics that doesn’t need you to change either the physics or the philosophy. And that basically leads you to the Everett interpretation.
David Wallace: And if you think Everett doesn’t work, and lots of people do, but you still hold onto that basic idea that theories should not have a primitive role for observation, and if you’re much, much more optimistic than I am about your prospects for redoing 20th century physics from scratch, then you might be led to want to modify quantum mechanics in one way or another. And if you’re relaxed about changing science, then you might well be led to Copenhagen.
Rob Wiblin: I see. So remind me, what are the names of the theories in the middle one, where you think that you’re going to update the physics?
David Wallace: Dynamical collapse theories and hidden variable theories.
Rob Wiblin: Okay.
David Wallace: I can give some classic examples of those. So the De Broglie–Bohm theory is a classic example of a hidden variable theory, for instance.
Rob Wiblin: So my impression is that, among philosophers of physics, the many-worlds theory is maybe the most popular. At least as a category it’s pretty popular. But then among physicists themselves, it seems like the Copenhagen interpretation is perhaps more popular. It doesn’t seem like any of these has a big majority in its favor. Is that about right?
David Wallace: Not entirely. I mean the fact that you think Everett is the most popular strategy among philosophers of physics says something kind of interesting about the—
Rob Wiblin: …my research process?
David Wallace: Well, no, I was going to say about the kind of overlap between the 80,000 Hours community and that broader intellectual community and the philosophy community. I would say Everett is a respected minority position among philosophers of physics. There’s probably substantially more support for what I was calling change-the-physics strategies, dynamical collapse strategies, hidden variable strategies. Among physicists, there’s almost no support for changing the physics. And Everett is probably roughly tied for first place with something like Copenhagen.
Rob Wiblin: Okay. So the benefit of the Copenhagen interpretation is that you don’t have to think that there’s all of this extra stuff. And I guess you have an explanation for how you’re getting these specific observations. It’s sufficient to explain the observations, but the cost you pay is you’ve created this new fundamental thing called observation and observers. Why are physicists okay with that? You’d think that they would be resistant to introducing this slightly mystical concept of an observer as like a fundamental property of the universe.
David Wallace: Yeah. I mean, physics is big and some people like that idea, but mostly they’d agree with you.
Rob Wiblin: Okay, right.
David Wallace: But you have to bear in mind, I said Everett’s tied for first place with something like Copenhagen. Really first place is occupied by, “Can we not think about this, please? We’d like to get on and do some calculations.” And that position sort of bounces back and forth between these various positions. A lot of physicists will say no, there absolutely shouldn’t be a role for the observer. But they won’t commit exactly to what they think the alternative is. In my optimism, I think of these people as sort of tacit Everettians. They’ve kind of adopted the Everett interpretation, but they don’t really realize it.
Rob Wiblin: I see. Okay. I thought you might say that this is kind of a heated issue or a heated disagreement, but it sounds like maybe it’s a cold disagreement, because people just aren’t interested in talking about it. Or they find it inconvenient.
David Wallace: Yeah it’s hot in philosophy, it’s definitely cold in physics. And almost at the professional level… I mean, sociologically it’s still not a great idea, I think, to spend a lot of your research time thinking and talking about interpretations of quantum mechanics, if you’re a pre-tenured physicist. It’s better than it was, but it’s still generally not a recommended career move.
Rob Wiblin: Yeah. I guess famously tenured professors of physics can say whatever they want about whatever topics.
David Wallace: More so anyway. Yeah.
Rob Wiblin: In brief, why do you think many worlds is the most plausible understanding? I suppose the reason is that it saves you from having to change the physics and saves you from having to change the philosophy, but why then are you so resistant to doing either of those things?
David Wallace: So I’m resistant to changing the physics because there is just a hell of a lot of very strong evidence for the theoretical edifice of modern quantum theory. Most advocates of strategies that want to change the physics…basically I think those people grossly underestimate the complexity of the task. You’ll find very, very few experts in modern quantum field theory who think that changing the physics to solve the measurement problem is a live option, for instance. There’s a sort of secondary reason there, which is that even if changing the physics is the way to go, then qua philosopher of physics, that’s not a good contribution for me to make. That’s a big first-order research program, and the kind of work I do tends to be about teasing out the conceptual and structural implications of theories we have. The task of developing a completely new theory is not really inside the space that my community, my skill set, is suited to. So I think philosophers shouldn’t really try to change the physics, even if that’s a good project. And vice versa: physicists shouldn’t try to change the philosophy. It doesn’t play to their strengths.
Changing the physics [01:10:41]
Rob Wiblin: So the middle path was to change the physics. So find some new theory that allows you to explain why it is that you get wave function collapse, or perhaps it will be interpreted differently than wave function collapse once you have that theory. I guess you think that’s not your department, but maybe you’re also kind of pessimistic that it’s actually going to work. Why is that?
David Wallace: I suppose because it’s really hard to come up with good physical theories at all. I mean, put it this way: If we knew for a fact that it was impossible to make sense of it with the physics we had, then I suppose you’d have a logical argument that there would have to be an alternative way to make sense of it. But if you don’t think that, then it’s almost a little quixotic to think that there could be a different theory which has a different quality that you like instead of the current one. I mean, look, it’s not a million miles away from the following.
David Wallace: Let’s suppose there’s something you really don’t like about Darwin’s theory of evolution. And if you were talking to a evolutionary biologist and you asked them to explain how to make sense of this feature that you don’t like, and they answered you, and then you were to ask them, well, what makes you think that’s the right way to go, rather than trying to chase the biology and come up with a completely new explanation for evolutionary complexity that doesn’t appeal to natural selection? That would be a very strange thing to ask. I don’t think it’s quite as clear cut in this case, but it’s a little bit like that. You need really good evidence to think that highly successful scientific theories should be just sort of ripped off and restarted from scratch.
Rob Wiblin: Yeah, I see. So in the biology analogy, there’s so much to be explained about the natural world, about living things. And evolution explains a huge fraction of them with an incredible amount of coherence. It fits so many pieces of evidence and it all kind of fits together internally. And you’re saying, if you found one thing that you didn’t like about it, would you say “I’m going to find a completely different theory than this which fits all of this data in the same way. Not only is there one sufficiently good theory, there’s actually two that would fit the data very nicely, but one of them will be nicer to my aesthetic preferences.” And that’s kind of a stereotype of what’s being done.
David Wallace: Yeah. I mean, I’m being a little unfair, but that’s the basic shape of it. Quantum mechanics explains everything from the scales of what’s going on in particle accelerators to the distribution and structure of the night sky. And it fills in all the gaps in between as well. And it underpins the whole of modern electronics, it underpins nuclear power, nuclear weapons, it underpins the reason that the colored objects I’m seeing are colored the way they are. It tells you why the sun shines. It does a lot of work.
Rob Wiblin: Okay. So an alternative theory has a lot of work cut out for it.
David Wallace: Exactly. And it’s fair to say that people have done kind of essays in the craft here. They’ve taken a little fragment of quantum mechanics and constructed a proof of concept of how you might modify that little fragment of quantum mechanics to fix it. So we have dynamical collapse theories of a little fragment of quantum mechanics, and we have hidden variable theories of a little fragment of quantum mechanics, but it really is a little fragment.
Rob Wiblin: I see. I’m imagining that the folks working on this project, if they were defending it, would say, “We’re not trying to come up with something that’s completely different. It’s going to preserve lots of aspects of existing quantum physics. We’re just going to modify it in some ways that then resolve these frustrating elements of the present theory.” Is that right?
David Wallace: Yes, that’s right. And they’d probably also say, and it’s not a completely unreasonable thing to say, “Look, there’s hardly any of us. And that’s why it’s taking us such a long time. We’re outnumbered 1,000 to 1 by the mainstream physics community.” So at the sub-level, it becomes what do you bet on, what do you think is likely to be a fruitful research program. It would probably get a bit too much into the technical weeds to say why I don’t think it’s likely to make that work, but I do think that there are quite severe technical problems that are underestimated by some of the attempts to develop these things. Not all, but some.
Changing the philosophy [01:14:21]
Rob Wiblin: Okay. So we’ve talked about the option of changing the physics there. Maybe let’s turn to the philosophy now. One problem here, or one puzzle here, is that it seems like the world to us is going to look exactly the same under all of these interpretations. And if so, how can one know whether one is right? And is that even a meaningful question, if the observations are all the same?
David Wallace: Yeah. I think it’s overstated how much that’s the case. I think there are many applications of quantum mechanics that we don’t really know how to systematically make sense of in the kind of language of Copenhagen, of the language of external observation. But that kind of way of doing quantum mechanics is kind of suited to an idea where you’re in a lab, you do a discreet preparation of your system. You let it evolve for a little while under the equations of quantum mechanics, and then you do a discrete measurement of it. But let’s suppose what you’re interested in is understanding the nuclear fusion processes of the sun. You’re dealing with a much richer, more messy, complicated process. Our data on how stellar evolution has gone includes things like what we can infer from the fossil record about variations in the emission patterns of the sun over time.
David Wallace: And it’s kind of a bit procrustean to try to say that the various bits of the fossil record counters a quantum observation of the state of the sun 100 million years ago. You’ve really got something that doesn’t have this nice discreet, we’ll make a quantum measurement and then appeal to the collapse of the wave function. You’ve got something that really tries to describe an ongoing dynamical process. You could say the same about this formation of structure in the early universe, where quantum fluctuations get magnified up to be the seeds of galaxies, but there’s no really sharp point to say it’s the observation that does that, or shifts from quantum to classical.
David Wallace: So I think actually if you want to do quantum mechanics in situations that leave the lab context, then doing something that’s roughly Everettian style is difficult to avoid. And you’ll find that if you look at particle physicists and string theorists, their style of quantum mechanics is pretty Everettian. And it usually, not always, but often, gets described in relatively explicitly Everettian language in that community. If you look at people who do sort of quantum information experiments and theory, they’re much more likely to use the Copenhagen language, as their situation’s better suited to… This goes back to the things I was saying right at the beginning about how there’s more of a feedback loop between your interpretation and your physics practice than you might think.
Rob Wiblin: I see. Is it possible to give a simplified example of an observation or practice that would be different in the many-worlds… Or is it something more subtle than you would actually see something differently in an experiment?
David Wallace: It’s more that doing the modeling at all is difficult to make sense of, outside of a roughly many-worlds way of thinking. If you wanted to model the sun, let’s say, you’re doing a process that really relies on you describing a series of quantum mechanical processes, principally even involving decoherence and things involving being entangled with the environment. The math is the math, but if you take the supposed way of making sense of the math that the Copenhagen interpretation advocates, it just doesn’t obviously fit that situation.
David Wallace: So it’s possible to interpret that math in an Everettian way. It’s not obviously possible to interpret it in a Copenhagen way. So this is another way of saying that it’s much more contested than maybe meets the eye, whether it’s the case that we have just a shared set of phenomena, all of which are describable in many different interpretations. I mean, there’s the math of quantum mechanics used in a kind of Everettian style without collapse of the wave function. And then there’s a question like, can various interpretations justify all of the applications of that maths? And I’d want to say that Everett or Everett-type interpretations can just make sense of the maths in ways that other approaches don’t. So that you couldn’t really say there’s other approaches that are making predictions at all.
Rob Wiblin: If you were trying to model the sun, model how the sun works, using a Copenhagen-ish approach, or some wave function collapse approach, it sounds like you’re saying at some point you would reach an impasse, or you would start running into problems or contradictions. What would that impasse be?
David Wallace: So if you’re trying a collapse approach, it’s not so much contradictions, it’s that your theory can’t handle nuclear physics, and so you can’t get going in the first place. If you try doing the Copenhagen style… Well, I suppose what you get is something like this. The nearest you could get is probably you’d say, well, my observer is the astrophysicist in the here and now, and I’ll do my 100 million years of the quantum physics of the sun. And at the end of the day, I’ll say that my observer collapses the wave function.
Rob Wiblin: I see.
David Wallace: But the problem is that what you’ve done in that process is mostly developed the kind of Everett-style quantum mechanics of that situation over 100 million years. All you’ve really done is right at the end of that you’ve pasted on a statement that says, “And now I collapsed the wave function.”
Rob Wiblin: Okay. Yeah.
David Wallace: So it’s kind of not doing any work.
Rob Wiblin: So you’re saying before there were observers like human beings, the sun was doing its thing, and presumably entering this ever more complicated superposition that couldn’t be resolved because there were no observers. And then at some point observers like humans, or I guess other animals come along, and then they’re meant to be collapsing all of the behavior of the sun, like going back millions and millions, I guess, possibly billions of billions of years. And that’s just bizarre, I guess.
David Wallace: Yeah. I mean, at some level you’re never going to test the Everett interpretation against just some opportunistic collapse of that kind. If you just say, would you ever be able to know that there isn’t just something that reaches in at some late time and discards all but a few of the Everett branches, you’d never rule that out. But at some level, this is beginning to start having the same character as the arguments that you’ll—
Rob Wiblin: …same problems. Yeah.
David Wallace: Yeah. Well, it almost starts to sound like the idea that you can’t rule out the fact that by experiment fossils were just put in the places they were put to look like dinosaurs, but there weren’t any dinosaurs. And there’s only so far you can ever go against that kind of flat skepticism. There are better and stronger ways that the Copenhagen interpretation is used that aren’t, I think, conspiratorial in the way I’ve described, but they’re limited. They have a scope of application, but struggle to handle these sort of big open-world situations.
Rob Wiblin: Yeah. Maybe I found that a little bit hard to understand at first because I didn’t imagine that people who believe that observation causes wave function collapse would think that it has to be people or animals or conscious beings exactly. It must be something about measurement that’s far more general than like, a person sees it, because that seems really wacky.
David Wallace: Yeah. But the problem has always been, there’s never been a good way of describing it. The only good way we have to describe observation as a mechanical process has always been just to go outside and do more quantum mechanics. In other words, the Everett move. If you want observation to do something that isn’t just more physics in that way, it has to sort of happen outside your application of physics. You have to kind of finish doing all the physics and then say, “And now my interpretation is through my observer.” If you ever want a process where there’s kind of a back and forth between evolution and observation, which is what you’d have to have if, say, simple systems count as observers, that either commits you to a kind of big modification of physics that we have no clue how to do, or it commits you to something like the Everett interpretation or something that I think would be Everett plus a kind of skeptical gloss, or something.
Instrumentalism vs. realism [01:21:42]
Rob Wiblin: There’s a more general question in philosophy of science that has influenced the interpretation of quantum physics that people have wanted to take. What is the disagreement about what physics is trying to accomplish that has had the most to do with this question of quantum physics interpretation?
David Wallace: I think it really goes back to those sort of issues that came up where I talked about Copenhagen. Is a scientific theory a predictive device, a gadget that helps us manipulate the world and predict experiments? Or is a scientific theory an attempt to explain what the world is doing, where experiments are just part of how we try to learn what the world’s like?
David Wallace: So the kind of approach that people call instrumentalism or operationalism or positivism was very much a way of thinking about scientific theories in the first way. It doesn’t really make sense to ask, is a scientific theory true? It doesn’t make sense to ask whether two scientific theories that make the same predictions are the same theory or not.
David Wallace: A scientific theory is really just a summary of all its predictions. It’s a tool to make those predictions. And the alternative sort of scientific realist approach wants to say more that a scientific theory is trying to describe the world. And of course, the way we’ll test that theory and try to understand if it’s true is by intervening in the world in various ways and observing the world in various ways. But the observations aren’t a core part of how you understand the theory, they’re just means by which you test the theory.
Rob Wiblin: So they have a different focus. The first group, the instrumentalist one, is focused on the observer, or the scientist, saying the point of these theories is to help me, the observer, accurately forecast the observations that I’m going to make. The second one is more externally focused, saying there is a real world out there that follows some regularities and it’s my job to figure out what those regularities are of this real thing. And I think a popular position in that school of thought is structural realism, right?
David Wallace: Well structural realism is a particular way of doing scientific realism. And in some ways it’s a partial concession to some of the worries that people have had about realism, some of the criticisms that the instrumentalists have raised. Because if you’re not careful, scientific realism can lead you to start committing yourself to an awful lot of totally unobservable, uncheckable facts about the world.
Rob Wiblin: Why is that?
David Wallace: Well, suppose you want to say, “I believe in electric fields.” Well, what’s an electric field? I can tell you mathematically, structurally, how we represent electric fields in my physical theories, but if I ask what really is an electric field… Is an electric field some object that permeates space? Or is an electric field not really a thing at all, but just a way of talking about the properties that points of space have? If you’re not careful, as a scientific realist, you end up committed to the idea that those are live distinct possibilities. But the math doesn’t care.
David Wallace: And because we only use the theory through the math, then the experiments don’t care. So if instrumentalism commits you too strongly to the idea that there’s nothing more to a theory than its experiment’s predictions, realism can commit you a bit too far to the fact that there’s a huge amount that we’ll never know through experiment, but nonetheless, there are truths about it.
David Wallace: Structural realism wants to say, “What scientific theories are in the business of doing is telling us how the world is structured. And there’s no structural difference between the two claims about the electric field.” One version of that would just say, “Well, this is a limitation of what we can know. There are non-structural facts about the world that we can’t know, all science can do is tell us the structural facts.”
David Wallace: The stronger version, that is what’s called ontic structural realism says, “Well, really all there is to the world is structure. The most that can be said about the world is how the world is structured. There’s no really meaningful non-structural fact about the world.”
Rob Wiblin: So a lot of smart and reasonable people have leaned towards some kind of instrumentalist take on what science is doing over the years. What can be said in favor of it?
David Wallace: You avoid sticking your neck out, basically. The instinct I think behind that sort of instrumentalism is a laudable instinct that you shouldn’t commit to more than you have evidence for. In the early 20th century, what was driving a lot of this sort of move in philosophy was a feeling that philosophy was trafficking in a whole lot of meaningless questions that were not a sensible thing to spend time on, but more than that, not even really well-defined. You were arguing about things where really you were just spitting out words that didn’t mean anything.
David Wallace: So a lot of the motivation is to say, “Okay, given that all we have is the empirical data, how can we understand our science in a way that commits us to know more than we actually get from the empirical data?” And I want to stress, I think that was a non-stupid idea. I think in these much more realist days, sometimes that whole period gets dismissed as a foolish wrong turn, but those guys were not stupid. They did have good reasons for it.
Rob Wiblin: So it was a reaction to people not being disciplined enough about deciding what questions are real questions of fact about the world?
David Wallace: I’m not a great historian of that period, so I couldn’t swear to getting the details right, but that was certainly the way a lot of the people writing in this tradition would write. So I mean take a silly example: Let’s suppose you want to argue about whether the world is upside down or right side up. That’s not a good example of a fact that we’ll never know because the experimental evidence doesn’t speak to it, it’s an example of a nonsense pseudo-question.
Rob Wiblin: Yeah.
David Wallace: And a lot of what people advocating these instrumentalist ways of thinking were doing was they were trying to have a reliable way to distinguish the nonsense pseudo-questions from reasonable questions. And a lot of their basis was, look, if evidence doesn’t bear on it, we can’t say anything sensible about it, because we can never even know what the words meant in that setup. The way we learn how language works is through the correlation between language and experience. So if this is something that just doesn’t connect to experience at all, then you don’t even know what the words mean. That’s the kind of idea that was driving it.
Rob Wiblin: And what was its main problem?
David Wallace: So in its own terms, the main problem is that it relies on a way of reconstructing the content of a scientific theory or a philosophical theory that’s way too simplistic. So to the sort of classic version of how theories of this kind went is it’s something like the theory is a summary of experiments. So if I make a theoretical claim like, “Matter is made up of electrons and protons,” what that translates to is something like, “If I do this experiment, I’ll see this thing. And if I do that experiment, I’ll see that thing.”
David Wallace: Or if I say the world used to be inhabited by giant reptiles, that becomes a way of saying, “If I look in such-and-such fossil bed, I’ll see this thing. And then if I look in such-and-such other fossil bed, I’ll see this other thing.” But the reason I have to give you these sketch descriptions rather than a proper real description is because it’s not possible to give a real description there.
David Wallace: In the language that philosophers use, observations are theory laden. You can’t really describe observations without using theoretical language. And you can’t really identify in a kind of articulated way which bit of your theory is proved or disproved by which bit of your evidence. Your theory as a whole confronts the experimental evidence as a whole. And so the way of trying to break down the content of the theory and re-express the content of the theory just in terms of direct observational claims just doesn’t really get off the ground. You can’t really make it work.
David Wallace: I mean, here’s a silly science example from these days. Think about lasers. If I say I’ve got an experiment involving a laser, well, what makes a given gadget a laser is really theoretical. Laser is not a direct term in our observation language that we knew and we could have understood when we were stomping around the savannah.
David Wallace: Laser is a theoretical technical term. Most of us probably don’t even know what laser literally means, and we use lasers in all sorts of contexts. You can’t really make the sharp distinction between observational claims and theoretical claims that you need to get off the ground in an approach which says the theoretical claims are just a way of talking about the observational claims.
Rob Wiblin: Instrumentalism also has this problem of not seemingly corresponding with what we actually feel like we’re trying to accomplish when we, for example, use fossils to say that we think that there were dinosaurs. Can you elaborate on that one?
David Wallace: Absolutely, yes. I mean, it’s a slight caricature. The difference between the instrumentalist and the realist would be something like the instrumentalist thinks that we talk about dinosaurs because we want to study fossils. And the realist thinks that we study fossils because we want to talk about dinosaurs.
David Wallace: I was saying earlier what the problem for instrumentalism is in its own terms, but absolutely one of the problems of instrumentalism from the point of view of the practitioner of science is it’s just pretty obvious that relatively few people in science are interested in photo plates or fossils or particle accelerators for their own sake; they’re interested in them because what they really want to know about are these deeper things about the world.
Rob Wiblin: A positive argument in favor of a realist take on things is that it just makes so much more sense that science works and can explain these regularities at all, at least as simply as it does. That’s so much more compatible with there being an actual world out there that does obey these regularities. It would kind of be a miracle if science was able to do this if there weren’t such a real world. And so on that basis, we should suppose that there is such a thing.
David Wallace: Absolutely. But it’s worth saying that argument only really even starts going if we’ve got a way of thinking about our theories that makes it even a live option that there isn’t an outside world. So instrumentalism tries to develop a way of thinking about our theories where we can even understand the theories as not talking about an external world, and I’ve said why I think that project ultimately fails.
David Wallace: There’s a different sort of thing we could do that’s not instrumentalism, but kind of is motivated by instrumentalism, which is to say something like, “Well, the realist is right about how to understand our theories, but we shouldn’t believe them. So yes, our theory is about dinosaurs, but we shouldn’t believe in dinosaurs. We should use the theory as a great tool to predict fossils, but we still shouldn’t believe it.”
David Wallace: And the kind of observation you’re just making there, which gets called the ‘no miracles argument,’ is exactly an attack on that second sort of anti-realist position, that sort of skeptical position that says we should understand our theories that make unobservable claims, but we shouldn’t believe those claims. And the response, at least in part, is something like, “Well, it’s an inexplicable miracle for the theory to be so successful if these claims it makes about the unobservable world are not true claims.”
Rob Wiblin: So we’ve got these two broad schools of thought in philosophy of science, instrumentalism and realism. Those who were approaching quantum physics with an instrumentalist take, how did that color their impression?
David Wallace: Well, I don’t think it’s a coincidence that the heyday of instrumentalism in philosophy was also the point of the foundation of quantum mechanics. And certainly the physicists of the time talked to philosophers to a much greater degree than is true these days. And so I think the Copenhagen style approach to thinking of quantum mechanics is very much an approach in the spirit of the instrumentalist way of approaching scientific theories. And conversely, I think the instrumentalists were partly cheered on in their belief by the fact that it seemed like this fantastic new theory that was being developed was a very strong exemplar of precisely what they thought scientific theories should be doing.
Rob Wiblin: What was the connection? Why did it make them more enthusiastic about the Copenhagen interpretation?
David Wallace: Well, Copenhagen’s all about claims about observation and measurement, observation and measurement are sort of in the primitives of how we think about the theory. Copenhagen doesn’t let you say things about the system, except in the context of what you can observe about the system. These are all the virtues of an instrumentalist’s take on a scientific theory that the other instrumentalist movement liked.
Rob Wiblin: And I suppose the instrumentalist approach allowed you to say, inasmuch as someone starts talking about many worlds, or wants to say, but what’s really going on? They can respond that that’s not a meaningful scientific question, that in fact they don’t have to answer that because they can explain their observations and that’s enough?
David Wallace: Well, if they can [laughing], but yes in principle.
Rob Wiblin: Right.
David Wallace: But look, I mean, I’m not just joking there, that kind of goes back to what I was saying about the problems of instrumentalism. Exactly the sorts of things I was saying earlier in the podcast about how if you really try to use the Copenhagen interpretation to describe and make sense of big, complex scientific applications, you find you can’t do it. That’s really just a special case of the kind of problem for instrumentalism of actually being able to separate theory from observation that I was talking about in the latter part. So you can really see that the things we’ve learned from philosophy of science, they’re not telling us why we shouldn’t entertain the Copenhagen interpretation, they’re telling us why we can entertain it and see what’s wrong with it in its own terms.
Rob Wiblin: It sounded like you were saying that over the last 70 years there’s been a resurgence of a realist take on what science is up to. How has that influenced people’s interpretations of quantum physics?
David Wallace: So I think it’s a little difficult to tell, but I think it’s sort of indirect. So if you think about the history of the subject, in the pre-history of modern philosophy of physics, people like Reichenbach and Karl Popper in the 1920s, 1930s, and 1940s were saying various broadly instrumentalist things. Well, not in Popper’s case, but various things about quantum mechanics.
David Wallace: But then there’s really quite a long fallow period where talk about the interpretation of quantum mechanics isn’t getting much air time in philosophy. You only really see interpretation of quantum mechanics becoming a big topic in the 1980s and later. And by that point, the kind of realist position in philosophy of science is quite close to being a consensus among philosophers of science.
David Wallace: I think maybe the indirect connection is maybe those weren’t good questions to be asking, or they were seen not to be good questions to be asking, except from a realist starting point. But I’m not sure here. I mean, look, it’s just a historical fact that in physics as well, the period from about WWII to the 1980s was a period where people didn’t really talk about the interpretation of quantum mechanics. They waved their hand in the direction of the Copenhagen interpretation and called it a day. And then there was a resurgence in physics and in philosophy starting around the 1980s. I think it’s a really complex, multi-strand question why that was.
Objections to many-worlds [01:35:26]
Rob Wiblin: We might return to that question of why people weren’t interested in this for that period, but let’s talk now about other objections to the Everett interpretation, reasons that everyone hasn’t been on board with it from the start. What argument or observation gives you the most doubts about the Everett interpretation?
David Wallace: It’s pretty different from what the main objections are, actually. I worry about it because it’s so cosmological. It makes sense as a theory of the universe, and it struggles to make sense as a kind of theory of substance of particular small systems. And I worry that scope for scientific theories can go beyond what we’re normally comfortable with.
David Wallace: I think trying to do the physics of the whole universe is a demanding and worthwhile goal, but it’s not something we’ve mostly been doing so far. We’ve been doing the physics of particular chunks of the universe. So I personally get nervous about the Everett interpretation being so cosmological in that sense.
Rob Wiblin: I guess a natural objection that I suppose a lot of people would have, I’m not sure whether scientists have this objection or just lay people like me, but it’s saying the many worlds theory is claiming that the universe is just full of so much stuff. You were calling it highly structured earlier. I guess I would say there’s just so much going on, so much detail here, it’s assuming lots of stuff being there and lots of details that otherwise maybe you wouldn’t have to imagine being there. Should we prefer a theory that requires us to think that there’s less stuff or information in the world?
David Wallace: Short answer: There’s no good reason to. I mean, look, here’s an analogy. If you think, how good would a theory of the world be that basically included the solar system and nothing else beyond it? And there’s lots you couldn’t explain, but in terms of what it could predict and manage, it would get all the predictions right, except a few predictions about some very faint lights in the sky that you often can’t even see. And some even more tenuous predictions about some even fainter lights you see in the sky, if you use really good telescopes.
David Wallace: There’s only really a very small amount wrong, descriptively, with a theory that says all there is is the solar system. And yet what we actually do as scientists is believe in the existence of 100 billion galaxies, each containing 100 billion stars, the vast majority of which are not visible at all individually. We extend our cosmos a ridiculous degree more than the solar system.
David Wallace: We’re not even slightly motivated to think that that’s not a sensible thing to do, just because it’s so big. Ultimately there’s no good reason I can see why having more stuff rather than having less stuff is bad. There’s lots of reasons for saying, “Don’t believe in anything without evidence.” That’s always good advice. But if the evidence points you towards a big thing, that’s not a reason to distrust the evidence.
Rob Wiblin: I think a lot of people will think, “Well, you should try to find a way of explaining your observations while thinking that there’s less stuff,” because we’ve been told that it’s important to have a simpler explanation, a more parsimonious explanation for things whenever it’s possible.
Rob Wiblin: And that then naturally carries over to thinking, well, if you think that there’s all of these extra galaxies out there, then that’s adding complexity to what you would have to say if you wanted to describe the entire world. And so maybe that should be penalized for its added complexity. But I suppose that’s kind of a misunderstanding of what our preference for parsimony should actually be.
David Wallace: I think that’s right. Normally speaking, if you look in science, the kind of reliable applications of this principle of parsimony, it’s normally just to sheer away things that aren’t doing any work. We’re not normally in situations where we’ve got two rival scientific theories and there’s no sense in which one of them is a sub-part of the other, but one of them is somewhat more complex than the other, and so we’ll go for the simpler one. Normally we just say, “Okay, we don’t know, let’s collect some more data.”
David Wallace: The theory that the dinosaurs were wiped out by an asteroid is simpler than the theory that there was an asteroid that hit the earth, but the dinosaurs were wiped out by a mixture of the asteroid and some volcanoes. Maybe your suspicion that parsimony is good might lead you to preferentially fund research on the asteroid project.
David Wallace: But fundamentally it would just be incredibly premature to say, “Because the asteroid theory is simpler, that’s what killed the dinosaurs.” The right thing to do would just be to collect more data. The reliable applications of the simplicity rule are normally where you’ve got two theories, the first theory is moderately complex, but then the second theory is the first theory plus some more stuff, plus some more mathematics stuff, and you’ve just got no independent evidence for it, there’s nothing about that that’s doing any work that wasn’t being judged as well.
David Wallace: But in any case, this is all about simplicity rather than amount of stuff. And while it’s true that a world of more galaxies is more complicated, most of that complexity is just sort of random contingency.
David Wallace: So if you think all of that rich detailed structure comes from a pretty simple initial quantum mechanical state plus quantum mechanical fluctuations that magnify up, that quantum state doesn’t get any more complex if you decide to make it just a bit physically bigger. It has more space to play out quantum mechanical randomness, so the actual world, or in many-worlds terms, our branch of the world, is way more complex. That’s not the kind of complexity that we’re sort of counting here.
David Wallace: It’s kind of like how complex is a fractal? I mean a fractal, you just sort of give a bitmap of a fractal, it’s ridiculously complex. And the bigger you make it, the more complex it is. But if you ask what’s the elegant mathematical description of a fractal, it’s just a nice short equation. And that’s the sense in which a fractal… From a theorizing point of view, fractals are simple.
Rob Wiblin: I see. So I suppose the Everett interpretation implies that there’s more detail, more stuff in the world, but the underlying mathematics isn’t more complicated. And that’s the area where we would actually be worried if it was, if it had lots of extra pieces that weren’t adding any explanatory power?
David Wallace: Yes, absolutely. Although the only other thing I’d add to that is that, in a sense, I think having this conversation about complexity concedes too much to the competition. Because it’s not as if we have this hypothetically more complex, but ontologically more parsimonious alternative that explains all the data. The main reason I’m driven to Everett is the lack of what seemed to be viable alternatives, rather than a feeling that I weighed Everett up against other viable alternatives and I prefer it more.
Rob Wiblin: What’s a key objection to the Everett interpretation that people raise with you?
David Wallace: So the objection you raise about parsimony I actually don’t hear so much from philosophers or physicists. Most commonly from philosophers you hear a worry about probability. How do we understand probability in this theory?
Rob Wiblin: Yeah, I’ve heard that this is an objection, but it’s always struck me as an odd one because my sentiment would be, if the Everett interpretation messes with the philosophy of probability, so much the worse for probability. Why would I influence my views on physics based on probability? Which I don’t really feel like I have that good a grasp on it, and I’m not sure whether it’s fundamentally real.
David Wallace: I think that’s a good instinct. It’s largely mine, I guess. Let’s suppose you thought that we completely understood probability prior to quantum mechanics. Suppose we had a totally, thoroughly satisfactory understanding of it; there were no remaining foundational problems at all. And then somehow quantum mechanics came along and there were then some interpretations that try to do probability in different things.
David Wallace: And then we could say, “Well, some interpretations of quantum mechanics conform to our completely satisfactory understanding of probability and some don’t.” Well, then you might worry. You might say, “Well, there’s other ones that better make sense of probability themselves, otherwise it’s a black mark against them.” As I think you recognize, we’re just not remotely in that situation in the philosophy of probability. Probability is a big mystery.
Rob Wiblin: We’ve slightly jumped the gun — why do people think there’s a tension between the Everett interpretation and probability?
David Wallace: So quantum physics seems to be tested by probability statements. Quantum mechanics is not in the business of saying this will happen or that will happen, it’s in the business of saying this will happen with probability 0.627 or something. So unless the theory delivers probabilities, it can’t make contact with the experimental data.
David Wallace: And people have at least thought that there’s only really two ways in which probability can enter physics. It can either enter because of our ignorance of the initial conditions, or it can enter because the equations of the theory aren’t deterministic. On different runs of the experiment, different outcomes can occur.
David Wallace: So traditionally in sort of our understanding of the way gases work or something, we want to say that it’s our ignorance of the initial conditions that explain why sometimes a little particle fluctuating in the breeze goes one way and sometimes goes the other way. In our kind of traditional pre-thinking about quantum mechanics understanding of what you might mean by radioactive decay, we normally think, “Well, it’s just indeterministic, there’s genuine randomness in the equations.”
David Wallace: But the Everett interpretation doesn’t have either of those things. There’s no relevant ignorance of initial conditions that’s going on here, and the Schrödinger equation is deterministic. It doesn’t predict the system might do this or it might do that, it predicts the system will definitely do both this and that at the same time. So there’s a concern that there’s just no space in the Everett interpretation in which you could insert the kind of objective probability we seem to need to explain how the theory is tested.
Rob Wiblin: So because in the Everett interpretation all of the different outcomes do happen, how would it be meaningful to say that one thing is more likely than another, even though we do want to say that some things happen with greater probability? That feels very central. But you have a response that’s really quite devastating here, which is saying not only is this not a problem, in fact, the Everett interpretation allows us to make sense of probability in a way that we never could using a previous interpretation.
David Wallace: Yeah, so the way that attempts to go is something like this. If you look at attempts to make a theory of probability in philosophy of probability, you broadly see two ideas people have. Maybe probabilities are frequencies, or maybe they’re about symmetries. And the frequency approach probably doesn’t work for reasons that maybe I won’t go into just now.
David Wallace: And the symmetry approach, it looks great if you think about something like I roll a die, why do I think the probability of getting a six is one-sixth? Well, initially you might think, well, there’s six possible outcomes, but you realize that’s not persuasive, there’s infinitely many possible outcomes. More persuasive is, the die is symmetric. There’s a symmetry that relates to different sides. If the die were loaded, that would break the symmetry, but a genuinely fair die will be symmetric, and that’s why the probabilities are one-sixth in each case.
David Wallace: But you realize that something is going to have to break the symmetry, because one outcome happens rather than the other. Ultimately, on this particular roll of the die, I got a six. So the initial conditions can’t genuinely have been symmetric.
David Wallace: They can’t have been, because ultimately one thing has to happen rather than another. Well, the Everett interpretation says, “No, it’s not true that one thing has to happen rather than the other.” In the Everett interpretation, if you roll your perfectly symmetric die with a perfectly symmetric starting condition, you get a perfectly symmetric outcome, which is an equal superposition of all of the six ways the die can land, if it’s a quantum mechanical die or something.
David Wallace: So the Everett interpretation gives you a route whereby from symmetry considerations, we can get probability considerations. We can understand how certain experiments are such that we ought to treat each outcome as equally likely. And then once you have a notion of equal likelihood, you can build up unequal likelihoods by imagining that you divide the unequal likelihoods into smaller cases so that you’d have to say it so that ultimately all the likeliness is equal. So there’s just a strategy and a method available in the Everett interpretation for parlaying symmetry into probability that was never available in a single universe theory.
Rob Wiblin: I think you’ve kind of answered it there, but how do you respond to the objection that, but everything happened, so therefore everything must have been equally likely? Although there are, at some level, each time there was a branching event, yes, both of them happened, but in many more of the branches, some actual outcome that you care about resulted. Whereas in other ones, it’s only in some tiny fraction of the branches that some outcome was achieved.
David Wallace: Something like that. I mean, the caveat is that trying to put any count on the number of branches doesn’t really work. So it’s more that when we say in most of the branches these things happened, that’s a sort of metaphor that picks out the collection of branches where collectively the probability is high. So you can’t really appeal to the count of branches to explain the probabilities.
David Wallace: But to respond to the objection you raise, I mean, everything happens, therefore everything is equally likely. That’s not even in its own terms mathematically coherent. If there’s no very reliable count of how many things happen, it can’t get off the ground. But even if it was a reliable count, I mean, let’s suppose that two things happen, so they’re both equally likely. Let’s suppose the two things are heads and tails.
David Wallace: And then in the heads world, let’s suppose we toss another coin. So now there were three things that could have happened. Heads, tails, heads, heads, and tails. So now if we think that if everything happens they’re all equally likely, we have to think it’s one third likely we get heads, tails. One third likely that we got heads, heads, and one third likely we got tails. But that’s inconsistent with the claim that actually it was half likely that we get heads. The probabilities just don’t add up.
David Wallace: So that rule for assigning probabilities is not a consistent rule. And it just generally is actually very hard to write down consistent rules for assigning probabilities in a universe that keeps branching. One way of simplifying the claim I and others want to make here is that actually if you push it, the only consistent rule you can write down to assign probabilities turns out to be the ordinary quantum mechanical rule.
Rob Wiblin: We’ve dealt with probability then, and I’ll stick up a link to a really excellent lecture that I listened to that you made about this question. What’s another objection to the Everett interpretation that’s in mainstream discourse?
David Wallace: The other main objection that has come up has been what sometimes gets called the ‘preferred basis problem,’ which is, how do you actually justify this talk of branching? I’ve got my quantum state that’s something like a living and dead cat at the same time. And in the language of quantum mechanics, you’d write that as something like live cat + dead cat. But what really licenses you to say there are two cats, rather than there’s just one indefinite cat? What justifies this move to this many-worlds language?
Rob Wiblin: And what does?
David Wallace: I’d say a proper understanding of how emergence happens in physical theories. That’s why my book’s called The Emergent Multiverse. I think if you look generally at what our stories are about when we’re allowed to use high-level language like cats and dogs and chairs and tables, rather than just talking to fundamental physics, the answer is when do we get robust autonomous structures that do robust autonomous stuff? I mean, the example I’ve used a bunch of times is, if what you want to do is study the hunting patterns of tigers, you could try doing that by studying the microphysics of the molecules in the tigers, but it’s basically not practical, and also even if you did it, you’d miss important high-level structures. What you actually discover is there’s a robustness whereby the atoms and molecules collect together into cells and organelles, and then there’s a robustness further where all of that stuff collects together into animals.
David Wallace: And there’s a descriptive level, which is the level of ecology and hunting pattern optimization and predator/prey relations, that’s mathematically coming out of the lower level stuff, but was invisible unless you looked for it. And it’s the fact that you’ve got that higher level emergent structure that justifies you saying there’s a tiger there. So if you apply that general philosophy of emergence, you apply that to Everett, what you find is what justified you saying there was a cat at all is that the quantum mechanical goings on are structured in a cat-ish kind of a way. So now if you’ve got a superposition, then you’ve got some goings on that are structured in cat-ish kind of way, and you’ve also got some goings on that are structured in a dead cat-ish kind of way, and those goings on don’t talk to each other. There’s no interaction between them. So that’s just applying the ordinary rules of our philosophy of emergence that tells us that there’s a live cat there and a dead cat there.
Rob Wiblin: Yeah, that makes sense to me. And I’ll stick up a link to another related lecture on that objection, which I always thought had a really good explanation.
Why a consensus hasn’t emerged [01:50:59]
Rob Wiblin: So I’m interested to probe why a consensus hasn’t been reached on these questions a little bit more. Obviously I haven’t studied physics since I was, I think, 15 or 16, so I’ve never really felt qualified to weigh in on these issues very much, but I know a lot of people who have a more mathematical and physics-y background, and I guess in my social circles there’s a lot of people who think the many-worlds hypothesis is just so obviously true that they are in disbelief that anyone has an alternative interpretation.
Rob Wiblin: And they’re somewhat drawn to looking for sociological explanations for why it is that everyone doesn’t agree with them, because they think that their position is so obvious. Which is a little bit of a dangerous game to play, because you can almost always come up with excuses for why other people believe something that seems so counterintuitive and strange to you. But I’m curious to know whether you think there have been any social influences that have perhaps led this research project of figuring out how to interpret quantum physics astray?
David Wallace: Oh, goodness. I mean, yes. If I would try to work out what they are, it’s a more subtle matter. I share the reason to be nervous of the sociological argument. It is hard to deny that the culture of physics for about the 50 years after the development of quantum mechanics was very hostile to asking these questions. Why was that? I think there’s a bunch of relatively contingent historical reasons. If you look at the history of this, then physics in this period was an intensely practical and solution-driven problem. The subject was evolving out of that.
David Wallace: I mean, the rise of quantum mechanics got eclipsed by WWII, and the leading lights in physics at the time were pulled into the Manhattan Project and then into the hydrogen bomb project, and were coming out of that around the 1950s with that kind of culture very much present, and also with an incredibly vast range of cool applications of quantum mechanics to explore. And I think there was a lot of “Don’t ask what the theory can do for you, ask what you can do for the theory” about it.
Rob Wiblin: So was the issue that physics got intertwined with practical applications, or that these people came of age at a time when they were trying to solve really concrete problems like defeating the Nazis, and then also figuring out things like, “We’ve just discovered this amazing physics, what can we build out of it?” And that was the mentality, rather than, “Let’s go and sit in an armchair and think about philosophy.”
David Wallace: I think there’s a lot of that. And it’s perhaps not a complete coincidence that a rise in a willingness to ask these questions came around the 1980s, which is kind of the last time theoretical particle physics had very direct experimental work to go on to drive it forward. That’s just armchair guesswork. It wouldn’t entirely surprise me. And I think the apparent absurdity of Everett and the slight feeling it was disreputable to talk about these unobservable parallel universes probably isn’t irrelevant to the account here. I also think it’s worth being a bit more optimistic. I mean, a very large amount of what people do in physics I think these days is Everett style, and can be understood in an Everett way, and it’s helpful to understand it in an Everett way, and it’s basically not helpful to understand it in other ways, but you can muddle along without having to work too hard in understanding it.
David Wallace: I don’t feel like a voice crying in the wilderness here. I feel an awful lot of what physicists do can be described in Everett-y language. And relatively minor moves, like saying, “Well, in Everett terms, one would say this,” rather than just saying “There’s a world in which this happens,” actually diffuse a lot of the instinctive psychological resistance. So I think some of that generally in physics is a bit of the legacy of instrumentalism and a bit of the feeling that it’s disreputable to be saying things that aren’t too closely tied to something experimentally accessible. But again, that’s largely diffusible, I think, in a lot of cases. I mean, this is in physics. Philosophy is another matter.
Rob Wiblin: Right, right. I’ve heard that there were a handful of physicists who took a really strong stance against the Everett interpretation, and perhaps then more junior scientists were less willing to go along with that because they thought it would damage their reputation. Is that an accurate story?
David Wallace: Yes, up to a point. So it’s fairly well documented that Hugh Everett, who came up with these ideas originally… It’s fairly well established that his PhD supervisor strongly discouraged, bordering on suppressed, some of the content there. That I think is correct. And it’s correct that, as late as the turn of the 1980s, I’d seen some fairly reliably sourced accounts of how some people’s graduate students who were working on some of the early decoherence ideas that were important in Everett had career derailings as a consequence. These days, I’d say it’s less that there’s community pressure against Everett as there’s a bit of community pressure against foundations and pure conceptual questions in physics. And some of that, I think, is understandable. I mean, what you’ve got to realize is that if you’re doing an experiment, if you’re making experimental predictions, then the success of your experiment is a very reliable auditing of whether you’re doing good science.
David Wallace: If you’re doing theoretical physics, but you’re doing something calculational, then the success of your calculation is likewise a very good audit. If you’re not doing either of those things, if you’re saying conceptual stuff about the structure of theories, it’s much harder to judge, to distinguish good work from bad work. And frankly, there’s an awful lot of very bad work done in that space. The journal Foundations of Physics has traditionally had quite a poor reputation among physicists. Part of that is prejudice against foundational work, but a lot of it, frankly, is a justified view that a lot of the stuff being published in it was really poor work by people who weren’t able to do the kind of thing that was meeting the normal criteria. And one way to see why philosophy has been a better home for some of these is that philosophy is better equipped with the tools to judge good work from bad work in a more conceptual space.
Rob Wiblin: So I suppose it’s natural that the physics community develops an aesthetic that really likes things that can be falsified, where you can figure out if someone’s wrong. And typically that’s good, but I suppose occasionally that could lead you in the wrong direction.
David Wallace: Yeah. And it’s a tricky problem. I think it’s a shame in a way that physics isn’t more tolerant of foundational work, even at a junior level. But I think it’s also true that the reasons it’s not come from some fairly good features about how physics is set up as a discipline. So I don’t have any brilliant suggestions to how that kind of thing would be reformed.
Practical implications of the many-worlds theory [01:57:11]
Rob Wiblin: Alright. Let’s push on to talking about what the audience was definitely most interested in when we asked for audience-submitted questions. We just got this one again and again and again, such that we almost had to ask people to stop submitting it. Should this actually influence anything that our listeners decide to do with their lives, or any decisions that they should make? If the many-worlds theory is right, does that change the impact of any of our actions?
David Wallace: I think it mostly doesn’t. That’s a little bit subtle, though. I mean, I’m guessing a lot of the people asking that question are sympathetic to, or at least understand something like a utilitarian picture of ethics, and nothing in the decision theoretic calculus of doing ethics particularly forces you towards a particular utility or disutility. Rational behavior in this framework can be maximized, including rational ethical behavior, can be maximized in expected utility, but the mere principle that that’s what’s rational doesn’t tell you what the utility function is.
David Wallace: So you could say, for instance, maybe in the many-worlds setup, I now realize that my lucky actions where I did something that could have been bad but in fact it wasn’t, I drove drunk or something, but not nothing bad came of it, I could be more aware that, of course, there’ll be branches in which something terrible came of it, and those branches are no less real than my branch, and the suffering in that branch is no less real than the suffering in my branch, therefore maybe I should be much more risk averse in an Everettian framework. Maybe I should put a much higher disutility on bad things, such that even quite low probabilities of bad things shouldn’t deter me from avoiding them. Maybe that’s true. If you thought that, then maybe learning that Everett was true would cause you to adjust your utility function quite sharply. Psychologically, I can’t report that that’s happened to me. My inclination is to think that even without Everett, if you’d thought clearly enough about low-probability/high-consequence events, you should already have been very worried about them. But that needn’t hold in general.
Rob Wiblin: I’ve heard from other people who’ve thought about this a little bit that they also lean towards thinking that it shouldn’t really impact how we evaluate the goodness and badness of different decisions. And it seems like that mostly stems from the fact that before Everett, we thought, say, that there was a 50% chance of outcome A and a 50% chance of outcome B, and then we do some expected value calculation where we weight them by the probability and their goodness. After many worlds, we say half of the worlds are A and half of the worlds are B, and then we weight them by the fraction of the worlds that are in each one, and then you do an expected value calculation across that, and it just looks the same. The math looks the same as long as you decide to use the fraction of the worlds and the probability the same way within your moral framework.
David Wallace: That’s basically right. The thing I’d add to that is that if Everett’s true, it’s been true all along. So when you originally thought that you were deciding what to do based on the probability, what probability really meant all along was the fraction of branches, you just didn’t know it. So the thing you were doing all along was already the Everettian thing. At some level this comes down to how you think about your metaethics. I mean, there’s a certain very pure style in some corners of philosophy, and probably in some of your readers, that says something like, “The way I should think about my ethics is I should just reason from the beginning as to what the virtuous person would do with no external world input, and then I should do it.” And if that’s your basis, then of course, if you were really badly wrong about the metaphysics of the world, like you didn’t know it was branching, maybe learning that fact would cause you to completely change your ethical assessment.
David Wallace: But if you’ve got a bit more naturalistic take on ethics, ethics are what they are because of how they’ve developed, and you’re not going to be able to find a ‘view from nowhere’ that justifies them, but nonetheless, we’re in the situation we’re in, well then again, the situation we’re in has always been a quantum mechanical situation. To go back to the Wittgenstein example from earlier, should the discovery of the fact that the lights in the sky were stars have changed our ethics? I think the answer is no, in the short run at least. It’s not that cheating on your partner or refusing to give money to save the starving child somehow changes its character because the lights in the sky are other suns.
David Wallace: Of course at some level that transformed our worldview, and in the long run that had big impacts on our ethics and our whole way of thinking about life, and maybe Everett will do that too. But the immediate questions about should I do this or that thing weren’t much changed because we understood how we were situated in the world. The basic mundane things around us were still the same mundane things around us.
Rob Wiblin: So if you were doing linear expected value stuff over consequences, then it seems to make surprisingly little difference. I suppose there’s other theories that might…a theory that says you should never kill anyone now has a slightly trickier time thinking, “Well, you will kill people in some fraction of them, and if it was just a strict prohibition that said it was infinitely bad to ever kill someone, then how do you make sense of that in a thing where so much more happens?” But perhaps that’s a question better asked of a moral philosopher, and it’s possible that just the answer is these theories are going to find it very hard to reconcile themselves, and maybe this would be a reason to go for something else instead.
David Wallace: I guess that could be right. But I mean, over and above that, those theories are already going to struggle with situations where we have a really low probability of killing someone, and they’re still going to… I mean, if you try putting an infinite disutility on killing… I mean you could argue that these are not utilitarian models at all, but if you put infinite disutility on killing people, it’s already the case that you can’t act on that basis because—
Rob Wiblin: …because you could do it by accident.
David Wallace: Yeah. At incredibly low probabilities.
Rob Wiblin: So then you have to say something about, you need to minimize the probability within reason, and then it just becomes you need to minimize the fraction of worlds—
David Wallace: Exactly. You’ve got those kinds of moves being available to you. But as I say, if what you really want to realize is that you should really, really try not to kill people, then you might say that the way you’re going to reflect that in the Everett interpretation is that once you know about Everett, maybe the disutility you place on killing people goes way up. Maybe you decide it’s actually much worse to kill people in a small branch than it is to have a small probability of killing people. Within reason, you can represent that just by changing your utility function.
Rob Wiblin: Yeah, so it might lead to a change if you thought that 1% of worlds was different than 1% probability, but I guess it’s not obvious why you would exactly think that. You could change the math such that this wouldn’t necessarily lead to such dramatic changes. It’s like maybe the switch between classical utilitarianism and prioritarianism or something that has a slightly different function over—
David Wallace: Right. And look, at some level I’m not professionally an ethicist, and I have no particularly developed metaethics. I think it really is a question of what’s the form of your metaethics. Some forms of your metaethics would say logically you shouldn’t change anything because of Everett, because it was always true all along. Other forms of your metaethics, other more relatively purer, relatively moral realistic metaethics might say that you should be open to changing your ethics really quite substantially when you learn this new piece of information. But an awful lot of those changes are absorbable into how I should rearrange my utility function.
Are our actions getting more or less important? [02:04:21]
Rob Wiblin: I want to return to an issue that I raised earlier, but dwell on it a little bit more. A member of the audience wrote in, “Should we think our actions are becoming exponentially less important as time passes, as the measure of the branch we identify ourselves on becomes smaller? Or should we think of our actions in terms of the effect of a policy across many separated branches?” If we’re on an ever-tinier fraction of all of the different branches, then it seems like what we’re doing is becoming radically less important over time. On the other hand, you could think that the universe is massively inflating just constantly all the time, as all of these branches are happening. And it’s not obvious that we can make any observation that would clarify, morally, which one of these things is actually happening. Do you have any comment on that general confusion?
David Wallace: I think the analog of inflation is correctly put. I mean, here’s an even more mundane example. There are more people in the world than there used to be. Does that make my actions less important than they would have been previously? I mean, in one sense, I suppose there’s a way in which you could say, yes, you’re less significant to the cosmos if there are more people like you. If I was the only person in the cosmos, I would be more significant than if I’m one of 7 billion people on Earth, and I’m even less significant to the cosmos if Everett’s true.
David Wallace: But I think it’s a little bit hubristic to be deeply concerned with your significance to the cosmos. I mean, practical questions about what I should do morally are not really going to be changed by these situations, they’re just going to be renormalized. I’m still going to hurt the same number of kittens if I have a kicking kitten policy. The fact that there are more kittens and more other people, if you have that policy, is neither here nor there as to whether I should kick kittens or not.
Rob Wiblin: But I think the idea is if you decided to kick the kitten earlier in time, then that would have affected a larger fraction of all of the branches, right? Because you’ve done it earlier before a branching event. So are events that happen earlier in time of graver moral significance because their consequences play out over a larger number of branches?
David Wallace: I don’t really think so, but I have to say I haven’t thought about it in those terms. Suppose I’m thinking right now, “In fact, I didn’t kick the kitten earlier. Shall I kick the kitten now or not?” Well, the past is the past. I can’t causally influence any of the rest of it, so it’s not that important, the question for your answer. If I’m thinking right now, “Well, shall I kick the kitten now or shall I wait 10 minutes before kicking the kitten?” Now I’m choosing between a strategy where the one David kicks the kitten, or a strategy where a million Davids kick a million kittens. And now, again, it doesn’t really matter. I mean, at the end of either of those processes, there are a million branches in each one of which I kicked a kitten. The ordering of the branch and the kitten kicking doesn’t matter very much. I mean, maybe that’s what your listener is getting at when they talk about strategies across a wide number of worlds.
Rob Wiblin: Yeah. So I was almost wondering, can you directly influence the other branches, and the answer is basically no, very quickly they become separated. But it seems like we might be able to affect what happens in other branches in this evidential or correlational sense, which is that… So you, 10 minutes earlier, you decided I’m going to kick a kitten in 10 minutes. That then causes, or the decision of you then 10 minutes later, when you find out that you did kick the kitten, causes you to learn that all of these other beings in the other branches that are extremely similar to you, virtually identical to you, they almost all decided to kick the kitten as well, because they basically are the same being. So this just becomes very difficult to think about, and a lot of people then want to say, “Well, every action that I take changes the expected goodness of the world in a much bigger way than I might have thought, because it gives me evidence about what I and beings like me have done in all of these other branches of the many worlds.”
David Wallace: Yeah, I mean, I think you’re caching out the issues exactly right. It’s a distinction between evidential decision theory and causal decision theory. So yeah, my doing a certain thing like kicking the kitten is extremely good evidence that I kick the kitten in other branches, even though it’s not, in the normal sense, causally relevant to doing it. There’s a well-established literature of debating how to think about these issues. The easy analogy that people probably mostly know is the voting case, whereby deciding to vote for a third-party candidate is extremely good evidence that other people are going to do it, but it doesn’t cause them to. Or by deciding to vote at all, even. That said, these problems are rather more innocuous in the Everett context than they are in lots of these other situations, because often why we care about the evidential versus causal consequence is that what actually happens is going to be some complicated nonlinear function of what everyone decides.
David Wallace: So shall I bother voting? Well if my decision to vote is extremely good evidence that others will vote but won’t cause them to vote, but if all of us vote, then the good outcome happens, but if only a few of us vote, the good outcome doesn’t happen… My own voting is causally irrelevant to anything useful. I only do a good thing as part of all of the people who vote. But that’s not true in the Everett situation. All of the branches have their good or bad things happening independently. So yes, my kicking the kitten is excellent evidence that other kittens are being kicked, but there’s a direct consequence of me kicking the kitchen, which is I hurt a kitten. I don’t need to think that it’s only good or bad because of the collective effect on the collectivity of kittens. So I think in practice while these things can matter to our self-conception, they’re rarely going to matter to my actual decision as to what to do.
Rob Wiblin: So I wrote a causal analysis of the value of voting, and I did get this response from some people, saying, “No, you should be analyzing it through the lens of your choice to vote. It gives you evidence about other people, and especially people like you choosing to vote more.” Which then leads to the question of how one would do that analysis and do the maths behind that. And as far as I could tell, someone hasn’t actually tried to do an expected value calculation on that basis, maybe because it’s not clear what numbers you’re going to put in. But I didn’t quite understand what you were saying there, like why it cancels out or why this evidential correlational issue shouldn’t influence your decision of whether to vote or not.
David Wallace: You mean in the Everett case or just normally?
Rob Wiblin: In the Everett case, yeah.
David Wallace: Okay, so let’s have a fairly simplistic example here. I’m one of 1,000 participants in some experiment, all of whom are extremely similar to me, and we can each pay $1 to press a button or something. And if at least 800 of us pay the dollar, then all of us get $100, just to make the example as stylized as possible. So in that situation, the causal and evidential decision theories come apart. Because the causal theory says, “Well, the other 999 people will do what they do independent of what I do. The likelihood therefore of my pressing the button shifting us from us all getting the money to us all not getting the money is incredibly small, so I should just keep my dollar.”
David Wallace: And the evidential argument goes something like, “Well, if I press the button, I should expect that most people press the button. If I don’t press the button, I should expect most people don’t press the button, because people are similar to me.” So the expected utility to me pressing the button is quite high. Close to $100. Well, close to $99. And then there’s a longstanding debate about what the right form of decision theory is to use in that context.
David Wallace: But the thing is you can’t do that in the Everett interpretation, because if you try replacing those 1,000 people who are like me with 1,000 branches, then the only way all of us could get a reward which depends collective on what all of us do is if you have some interference between the branches, which you can’t do. The linearity of the Schrödinger equation and the irreversibility of the branching thing means that you’ll never have physics of that kind. You just couldn’t physically realize that in the Everett situation. So you can’t… At least as far as I can see, I haven’t thought about it much before this, you can’t construct those kinds of causal, evidential, disconnects in Everett in just collective action problems in ordinary classical physics.
Does utility increase? [02:12:02]
Rob Wiblin: That makes sense. Let’s return to another bit that we’ve kind of talked about, but I don’t feel like we’ve quite wrapped up. Earlier I said, over time, there’s many more branches. Is the total amount of utility — say that’s being generated across all of these branches — growing over time? Because they’re separating and becoming more numerous, in some sense, although I know counting is problematic here. And it sounded like you were saying no, the total amount of utility isn’t changed by the fact that you get this branching, assuming that the same thing is happening weighted by the fraction of things that have different things going on in them.
Rob Wiblin: I have, as a person, traveled through time, and presumably become presently a smaller fraction of the entire wave function in this model. And yet things feel the same to me. I don’t feel like I’m any less morally significant than I was before. But I suppose it would feel that way regardless, because I can’t sense the wave function splitting, or me being in a thinner part of it. Is that right, and if so, how would we get evidence either way?
David Wallace: I think it’s right as far as it goes. At some level, I’m not really sure what is being said when we want to say that the total utility of the universe in this sense is going up or going down. I don’t really understand what utility means here, outside of at least some idealized decision-threaded context. I guess if we’re asking the question of what should God do, how bad would it be for God to destroy the universe, and would it get worse and worse for God to destroy the universe the more branching happens? I mean, I don’t know. I just don’t know how to think about questions like that. I think they’re starting to get into the space that the instrumentalists…
David Wallace: No, if we’re just asking about anything that will apply to situated moral agents within the universe, then questions of that kind are never going to matter. Because all it would really do if I decided to magnify the utility function as the universe got bigger rather than to keep it normalized at 1 is just to renormalize all the utilities. So that’s never going to affect the choice that you make.
Rob Wiblin: Intuitively it seems like it should imply that, say if you have some trade-off between doing something good and something selfish, that it would be better to do the thing that has long-term positive consequences earlier, because then it seems like it affects more branches. But as you pointed out, because if you do it later you’ll be doing it in more branches on the other end, that cancels out as well.
Rob Wiblin: I can see both how it seems like a meaningful question to say, “Is the total amount of moral value being generated by the situation increasing as there’s more branching?” And I can also see how that just seems like, how would one even begin to answer that, because we don’t have a way of measuring utility anyway.
Rob Wiblin: It does seem like it could be incredibly consequential in some sense, because it’s changing our estimate of how much wellbeing there is in the world by a factor of just unbelievably large numbers, because there’s so many of these branching events. Maybe I am just at the point of asking this question where I’m too confused about what I’m asking for there to be a coherent answer.
David Wallace: I think my real concern would be, there might be ways in which you could raise an ethical question that comes from this, but it needs to be quite situated. You’d need to be asking, “Okay, what ethical choice could you make such that this is going to bear on that choice?” So at some level, if what utility means in our theory is simply something that we want to try to use to decide what the best course of action is, then the absolute scale of utility is meaningless.
David Wallace: If I said, “Well, we used to measure utility in utils, but we’ve now decided to measure them in milliutils, and now there’s 1,000 times more happiness in the world,” that’s obviously nonsense. The only context I can think of where maybe it matters is something like, suppose I think it is better to have more happy people, so should I try to induce a lot of Everett branching so as to make that come about? Or conversely, maybe I think life is a veil of tears and so it’s not a good idea to create more miserable people. So should I restrict metabranching?
Rob Wiblin: Should you increase branching by turning on a neon light?
David Wallace: Yeah. I mean, it’s difficult, although I think probably not impossible, to construct self-consistent strategies that do that. But as a practical matter, the amount of branching that’s happening is so overwhelmingly dominated by random events all over the place in the universe that it’s hard to really see it as something you can actually influence. And I think, as a practical matter, even attempts to count how much branching is going on run into the sort of observations I was making earlier, the actual number of branches is not really a well-defined thing.
Rob Wiblin: I think we’re struggling with an issue here that a lot of people care about, but people haven’t really figured out how to cache this out. At least I tried to find a paper or like anything solid written on this and failed. I don’t quite know why. Maybe it’s because it’s just so confusing.
David Wallace: There isn’t a lot. I think it’s probably an interesting space. But philosophers of physics mostly aren’t ethicists, and ethicists mostly aren’t philosophers of physics. So I think that’s probably what drives it.
Could we influence other branches? [02:17:01]
Rob Wiblin: Is there any way that we could conceivably influence other branches of the multiverse, even with hypothetical, extremely advanced technology, or is that prohibited basically?
David Wallace: It’s basically prohibited. It doesn’t really make sense within one branch to reach out and influence another branch. The nearest you can get to that is to say that somebody could step outside the whole branching process and cause the branch to interfere. So could ridiculously advanced aliens put a bubble around the Milky Way so that the branching doesn’t get beyond the bubble, and then cause the various branches within the bubble to re-interfere with each other? I mean realistically not, but nothing inside the structure of quantum mechanics would rule out that scenario.
David Wallace: I mean, other things about physics would rule that scenario out, it’s not possible, but in terms of quantum mechanics it could happen. But that’s pretty different from the idea that we within our existing branch could reach out to different branches. As long as the principles of quantum mechanics are exact, then that’s not possible.
Rob Wiblin: So a general class of problems for utilitarian or consequentialist theories is infinitarian issues, where what if things are so large that there’s an infinite amount of good and an infinite amount of bad? It seems like they may bite, because the scale of the universe may be that large, in which case we’ll have to modify the theories in ways that, as far as I know, are not yet fully figured out. Does the many-worlds theory potentially make that problem worse if it weren’t present already?
David Wallace: Well, this is mostly above my pay grade, I’m not super familiar with these theories, but my suspicion is it probably doesn’t, because the many-worlds theory has a formal tool in terms of the branch weight or branch fraction that isn’t available in these classical infinity cases. So in the classic way these objections tend to go, I’ve got infinitely many copies of Earth, and nothing to distinguish them. So I could weight different copies of Earth differently. I could weight the first copy half, the second copy a quarter, the third copy an eighth and so on, and then I’d have a perfectly well-defined probability theory. But that conflicts with the pretty obvious argument that the utility of the situation shouldn’t be changed by transposing two Earths, given that there’s nothing that distinguishes them. And so you run into just formal mathematical problems with, in formal terms, defining a probability function, defining a utility function.
David Wallace: In the many-worlds situation, you can’t have infinitely many branches, each of which is indistinguishable from each other, just because you can’t fit it into the math. They have to have a branch weight attached to them. And if each one has a finite branch weight, then their sum would be infinite. And that violates the principle that all the branch rates have to add up to one.
David Wallace: So in the many-worlds context, if you’ve got this vast number, maybe even you’ve got an infinite number of possibilities, you’re always going to have a branch weight that distinguishes them and at least formally gives you a justification for saying, this is why these branches count as more important than these branches in your calculus. Now whether that formal justification is a conceptual justification just brings us back to the probability problem.
Should you do unpleasant things first? [02:19:52]
Rob Wiblin: Okay, so one thing that I was thinking in prepping for the interview — where I thought, “Maybe this could have some practical implication for me?” — whether each time the world branches into two, separate of the many worlds, then it’s like now there’s twice as much value or something like that. If that were the case, and say I was planning out the next two weeks, and I’ve got a particular amount of work that I have to do — work, hypothetically, that I don’t enjoy as much as making this show. So I’ve got a week’s worth of work to do, and then after that, I can party, I can have a good time.
Rob Wiblin: It seems like, if the branching of the universe creates more goodness by there being more stuff — like more complication, more computation going on in aggregate — then I should want to do the unpleasant thing first, when the world is smaller — there’s less detail and less structure. Then I should have the fun time later on when the world is kind of bigger, and there’s more detail, and more structure, because then I’ll be getting more pleasure across all of these many worlds. Does that make sense?
David Wallace: Yeah. I don’t think I buy it, but it makes sense.
Rob Wiblin: Okay.
David Wallace: I mean, here’s the one thing maybe that doesn’t make sense there. You couldn’t say that it’s you getting more pleasure, of course. I mean, each individual version of you is getting the same amount of pleasure. It might be more that you’ve got some preference that, in your future, there are more versions of you, each of whom is getting more pleasure.
Rob Wiblin: Yeah.
David Wallace: I think what you think about that is going to depend on your metaethics. If the way you think about ethics is that ultimately it’s fairly naturalized — you can kind of read off behavior what people’s ethical positions are — then it’s just relatively clear, empirically, that people don’t in fact prefer that. The way in which people consider risk and delay gratification clearly is not sensitive to issues about how much variety turns up in the process of getting it.
David Wallace: So if you had that kind of relatively deflationary metaethics, you’re going to say, “No, it’s just empirically true that people don’t care about that.” If you’ve got a more sort of a priori metaethics — “I sort of think I ought to be able to work out my ethics from first principles” — then you might think, “Well, I’ve learned something big and important about the universe, and so I should change my behavior to allow for it.” Then that’s going to depend on how the rules on that are done. But even then, from a physics point of view, it’s messier than it looks because — and this goes back to something we talked about much earlier in the podcast — the number of branches is not really a well-defined notion in this theory. There’s no very sharp way of deciding how I’m going to carve out reality into these various branches.
David Wallace: I mean, at one grain of looking, you might think that there’s 10 of me after some experiment. At some different grain of looking, there might be 100. At another grain, there’s going to be 1,000. In fact, again, the physics doesn’t really care whether you want to describe the whole process as saying, “The number of copies of me is constantly increasing” or if you want to describe it as saying, “There’s the same number of copies of me, and they’re just getting more and more divergent from each other.” That’s a matter of how we use human language to describe the sort of branching mathematical structure, but there’s no — at least at the level of the physics — there’s nothing about the description that forces you to one or other way of talking there.
Rob Wiblin: Okay. So I find this hard to talk about, but I guess the fact that there’s this ambiguity about how many copies of you there are makes it seem even weirder and somewhat… It’s not clear what claim we’re even making about there being more versions of you, because it’s more just like there’s a smearing of more things going on that kind of correspond with you in some sense. But I guess it does still seem like, over time, the number of different divergent things that we call “you” is increasing. It doesn’t decrease, not at the universe level anyway, so it does seem like, on some metaethical views or… Yeah, I don’t know that anyone’s really thought about “How does this interact with metaethics?” Well, we should probably get an expert on metaethics to think about this, but —
David Wallace: Yeah, it’s not been much described or discussed. To some extent, it runs into just mathematical definitional problems. I mean, there’s a very clear, definable decision theory where you just treat branches like probabilities. Of course, a corollary of that is you wouldn’t care about mere branching. The ambiguity and vagueness of branch definition is going to make it very hard, I think, to have a stable and consistent decision theory that doesn’t hold to this consideration that mere branching doesn’t matter. But of course, that can be challenged. And to some extent, some of my own work on understanding probability and how to think about the many worlds has been arguing that, actually, the only really stable decision theory you’re going to get is one that isn’t sensitive to the mere effect of branching.
David Wallace: Of course, decision theory doesn’t inherently have to be selfish. I can use decision theory to consider ethical questions as well. But again, this is a place where more or less everyone’s out of their element in one direction or other. I’m definitely out of my element in metaethics, and I think there are big open questions worth thinking about here.
Rob Wiblin: Yeah. Yeah.
David Wallace: Here’s a metaphor that might make it at least somewhat easier to understand this kind of indefiniteness of branching. You might imagine that we lived in a two-dimensional universe. Maybe I’m a two-dimensional fish swimming around in a two-dimensional ocean, and then you might imagine that there could be lots of two-dimensional universes, and they’re stacked on top of each other, and they can interact a little bit. So, I interact a little bit with fish a little bit below me or above me but not at all with fish a long way from me.
David Wallace: If that was true, actually, then probably it wouldn’t make sense to say that I was a fish just in one layer of this big stack of two-dimensional universes, because the processes that made me up might kind of do a certain amount of cohering from one layer to another. So what you’re going to get there is a world of rather thin beings and a world in which entities don’t really interact very far through the stack of two-dimensional universes, but where the sort of autonomous chunks of this are not going to be single slices — they’re going to be slightly indefinitely defined chunks of slices.
David Wallace: Once you’ve got that reality, you might imagine I can really take away the definite slices at all, and I can just say my universe is three-dimensional, but the interactions are very strongly confined to the plane, and they only go a little bit up and down. That’s a situation in which you’ve clearly, in some sense, got a multiverse — the things going on very much deeper into the stack or higher in the stack are not interacting with things at this level, but the world doesn’t really have a sharply, a distinctly discrete breaking down into slices.
Rob Wiblin: Okay. Yeah. I’ll try to say that back. You’re imagining an ocean that kind of has a fixed top and a fixed bottom, and it’s made up of all of these two-dimensional planes that are kind of stacked on top of one another.
David Wallace: Right.
Rob Wiblin: Basically, the physics in this world allows for a slice across this ocean plane to have interactions with the slices immediately above and below.
David Wallace: Yeah.
Rob Wiblin: But that means that, because there’s interactions between each layer and the area above and below it, and the closer it is, the more it interacts, the further away it is, the less it interacts —
David Wallace: Yeah.
Rob Wiblin: — it’s not really a specific layer. There’s just, each layer is actually like a smear. It’s actually like a probability distribution across a bunch of different layers, depending on how strong the interactions are between them.
David Wallace: Yes, something like that.
Rob Wiblin: Then, even as this world progresses over time, it becomes apparent that there’s not really an increasing amount. It’s just that the nature of the interactions between the different subplanes shifts over time.
David Wallace: Yeah.
Rob Wiblin: Okay. Yeah, amazing. I was about to ask a question just before you said that, where I was thinking… It’s as if we’ve got the space of numbers, the space of all possible numbers between zero and one. And over time, say, I’m looking closer and closer at all of these numbers being like, “Oh, look. I can keep adding more and more decimal places,” and then you’re like, “As we do this, there’s more numbers. There’s more stuff going on.” But you’re saying, “No, there’s the same amount of numbers.” It’s just that you’re looking more closely as you are adding more and more digits to this incredibly long decimal number.
David Wallace: Yeah. Something like that.
Rob Wiblin: Okay. Yeah.
David Wallace: Because I mean, on any sensible way of thinking about the branches, there are going to be vast numbers of me who are psychologically indistinguishable from one another. Let’s say somebody in a lab in China is currently looking at a Geiger counter. Well, that Geiger counter is constantly causing the world to branch, but not in any way that’s remotely salient to me.
David Wallace: So uncontentiously, if something like the many-worlds theory is true, there are just lots and lots and lots of branches in which I’m having the same experiences, so it’s not as if… Maybe there’s even wilder metaethics where what I care about is the number of versions of me that are sufficiently psychologically different that they’re having different experiences. But if we’re just talking about a mere count, then again, it’s not obvious that the way you’d want to define that means that the count of versions of me is actually going up. Maybe it’s just that the level of variety across versions of me is going up. Again, the mathematics doesn’t care about this.
Rob Wiblin: Yeah.
David Wallace: If there are factual answers to these questions, they’re not answers that you’ll see through the physics.
Rob Wiblin: Right.
David Wallace: There’ll have to be some extra metaphysical gloss over the top.
Rob Wiblin: Or a matter of personal preference and values. It seems like maybe I should get away from thinking in terms of the number of worlds in the many worlds and think about it just in terms of fractions of the total. And then, maybe once you start thinking about it in terms of like, “Well, in this fraction of all the worlds that came out of this, then we have this kind of outcome, and in this fraction, we have that outcome.” Maybe, like, is that more conducive to thinking about it as a fixed ocean with particular fractions that has gotten more and more specific, rather than thinking about it as one that’s getting vastly huger all the time?
David Wallace: Yeah. Certainly, I think, if that’s your metaphor, if you think about the number of worlds being just sort of indefinite, but you can talk in a stable way about the fraction of worlds with a particular property, that, I think… well, at the very least, that kind of keys for a different set of intuitions about the ethical and personal implications.
Rob Wiblin: Yeah. Okay. We’ve reached the end of my list of questions about the Everett interpretation, but we’ve been talking about it for almost two hours or something now so I feel like I can’t just move on without having some sum-up question. Maybe what’s the research frontier here? Or is there anything you’d like to say to wrap up?
David Wallace: I don’t want to underplay how important and striking this is. I’ve been giving quite conservative answers to a lot of these questions, and I’d stick by these answers. But I do think if we’re right about the Everett interpretation being the right way to read quantum mechanics, then during the 20th century, we learned something about the universe and our place in it that’s at least as striking as—
Rob Wiblin: …anything else.
David Wallace: Well, arguably yes. But certainly at least as striking as our discovery that the stars were other suns, and that there were other planets and other galaxies. Our place in the universe has been changed at least as radically by that discovery as by anything else.
Rob Wiblin: We’ll collect a bunch of links of resources and audio and video stuff for people who are interested in learning more about this question, because there’s plenty more that one could say.
Progress in physics over the last 50 years [02:30:55]
Rob Wiblin: I want to push on and talk about whether there are still valuable discoveries to be made in physics. This is a question that’s relevant to an 80,000 Hours audience because people are planning their careers and thinking about how they can do more good. And one option, for very smart quantitative people, might be to go into physics. Before we get to that though, how much progress have we made in fundamental physics in the last 50 years versus the 50 years before that?
David Wallace: Hmm. A lot less. That would be the short answer. The period from say 1930 to 1980, we started with only the rudiments of quantum mechanics coming together, and at the end of that period, at least if we’re talking about small-scale quantum phenomena, then the standard model of particle physics was the thing we’d achieved. And that was an incredible achievement.
David Wallace: If you ask where we are in the development of fundamental physics 40 years later, it’s still the case that the standard model of particle physics is really the summary of our understanding. So while we’ve learned a great deal in that period, we haven’t made anything like the strides in novel theorizing confirmed by observation that we made in the early period.
David Wallace: It’s more exciting if you look at cosmology. Cosmology has advanced a lot, and our observational situation in cosmology has advanced a lot, in the last 40 years. But again, you couldn’t really compare it to the golden period of the mid 20th century. That was just an incredible period for physics.
Rob Wiblin: And is it basically right that our current theories in physics explain all the phenomena that we experience on Earth pretty accurately?
David Wallace: I’d be nervous about ‘explain.’ I think explanations often is a higher or emergent level. What you want to say about that is something like, “Do we think there’s an experiment you could do on the surface of the earth such that it wouldn’t be accurately predicted by modern physics if only we had the computational power?” I think there are excellent reasons to think that we would struggle to do any such experiment.
David Wallace: I mean, in a certain sense, of course we can. Particle accelerators push towards that space. But even the Large Hadron Collider isn’t really succeeding at the moment, and you’d need to spend extremely large amounts of money to build very delicate devices in order to find potential direct violations of our current best physics.
Rob Wiblin: So the problems with our current theories arise at black holes and neutron stars and extremely high-energy environments that we could never get close to anyway, because they would probably kill us. It suggests that coming up with a theory of everything, while an amazing accomplishment, might not really have any practical implications for anything that we can do or any predictions that we need to make about our experiences, is that broadly right?
David Wallace: That might be right. I think there are serious grounds for thinking it’s right. I mean, we don’t know. It might be that there are things we could do with our continued developed physics that don’t correspond to things that have shown up in nature, but could still be done. Maybe there are very sophisticated quantum chroma dynamic states of neutrons and protons and quark matter that could be used to build matter in certain ways that just were never having to be realized in nature. It’s possible. Even that would be something inside standard physics of course, that wouldn’t be triggered by genuinely new fundamental physics.
David Wallace: So I wouldn’t want to say that we can be sure that there won’t be relatively technological applications of physics beyond the standard model, but I certainly think there’s no compelling argument that there is, of anything like the compelling arguments for previous periods in physics.
Rob Wiblin: So at a fundamental level, we just can’t know this, but I suppose if we were betting, just based on the fact that it seems like our theories have converged on getting closer and closer to being able to predict everything and giving us the knowledge that we need in principle to see that something is technologically possible, even if we can’t do it yet and that may be drying up over time, we might expect that the future is just like a continued asymptoting to not really having many practical applications on Earth, at least now?
David Wallace: Yes, I think that’s right. I mean, I think if I had to bet, then my bet would be that post-standard-model physics won’t have direct technological applications until the fairly distant future. A civilization that can travel interstellar distances and build black holes could do things with physics we don’t have at the moment, but we’re quite a long way from those kinds of horizons.
Practical value of physics today [02:35:24]
Rob Wiblin: I find this uncomfortable to say, because it would just be such an amazing accomplishment to come up with a theory of everything, and the idea of discouraging people from doing that in my lifetime seems profane somehow to me. But I guess, if you look at score results and SAT scores, physics as a field is just attracting the smartest of the smartest people in society. And it seems like it’s doing that while the project doesn’t seem to have many practical applications that seem likely to be useful in the near future, or to solve problems that are at least urgent now. Maybe it will be able to do things that are useful once we can reach the black hole in the center of the galaxy. But this suggests that maybe we should kick the can down the road a bit on some of this physics stuff, and solve practical problems. And then leave this as something that future generations can solve with their hopefully vastly superior analytical capabilities.
David Wallace: I don’t think that’s an indefensible position to adopt. Let me give the counter case without necessarily saying the counter case is compelling. The main counter case I’d make is that physics absolutely has the potential to be making a whole bunch of transformative contributions to the world. And the divide between fundamental physics and non-fundamental physics is very blurry in terms of methods.
David Wallace: I’ll give you a concrete example. So at a formal mathematical level, the way we understand the process by which the Higgs boson gives mass to particles is pretty much exactly the same as the way in which we understand how superconductivity is possible. And superconductivity really matters technologically. A room-temperature superconductor would be epically transformative in vast amounts of our infrastructure.
David Wallace: So a whole bunch of things in physics of that kind have a lot of potential to be really important to how we develop as a society in the relatively near term. And you really can’t hive off the community of people doing fundamental physics from the community doing those kinds of applicable physics.
David Wallace: If you try the strategy you were discussing, which would have been defensible on the same grounds 50 years ago, you’d materially have harmed the development of our solid-state understanding of superconductivity, because you’d have closed off the important back and forth that was happening between the solid state physicists and the particle physicists.
David Wallace: So I think there’s at least a live argument that that kind of back and forth of techniques and ideas and applications and concepts really means that doing deep theoretical physics is important and contributory, and you can’t really do it in a way that artificially says, “Only do this part of it.”
Rob Wiblin: I see. So the argument there is that even physics that on its face seems very theoretical, or very abstract, does in fact lead to applications at a reasonable rate. And it has this interface with applied physics where it’s generating valuable inventions or applications in a way that may not be immediately obvious?
David Wallace: Yes, exactly. I mean, there are lots and lots of things that could come out of physics that are really important for the way our world would be in the short, medium, and longer terms. I mean, the ones you can immediately think of tend to be things that have probably got a bit too applied to be directly connected to theoretical physics. So something like battery technology… Even probably these days, how you want to make the latest superconductor, then it’s probably true that that development can be seen as a genuine piece of applied physics, but now we’re not talking about distant future million-year, 1,000-year time horizons, now we’re talking about a matter of a few decades.
David Wallace: Techniques like renormalization group theory, which I won’t go into in detail, but it’s a really important analytical tool in huge amounts of physics and even in bits of science beyond physics, is again, something that was developed out of very theoretical considerations in physics. Part of it is about developing mathematical tools and technology. Part of it is about drawing certain analogies. Part of it’s just to do with the general principle that smart people in a particular discipline are not generally helped in their development in that discipline by being artificially corralled. I mean, I’m not sure I could compellingly demonstrate that to you, but I think it’s fairly plausible.
Rob Wiblin: Yeah. Another objection that someone might raise is saying well, you think it’s unlikely that coming up with a theory of everything that’s a significant step forward would lead to applications, but you don’t know that. The previous discoveries have had massive revolutionary implications and we know that our theory isn’t quite right. So maybe the next thing really could be a big step forward that would unlock massive stuff. And we’re playing here with the nature of the world in which we live at the most fundamental level, so potentially those breakthroughs could be of extreme importance.
David Wallace: I think that’s right. That would be the second half of the case I’d make. I think it’s less compelling than the case via communication between bits of the subject. But I think it’s there. I mean, the pushback against the case says that in those previous examples, there were rich bits of the phenomenology of the world that we didn’t understand. And this goes back to your question about whether there are experiments now on the surface of the earth that we can’t predict with extant physics. This is simplifying a bit, but most of our technology in most of our world is about various manipulations of the ways in which electromagnetic fields interact with matter. And we mostly didn’t understand the phenomena we saw before we understood all of those ideas. But we don’t appear to have the same kind of direct presence in our everyday world of a whole bunch of things we can’t explain that point towards bits of physics that we don’t have.
David Wallace: But I don’t regard that argument as compelling. And I think it is true that if there were transformative things that came out of fundamental physics, they’d be truly transformative. They’d be things that are transformative in the relatively literal sense, so it’s difficult to speculate on what they’d be. I mean, what’s the likelihood that you can build wormhole generators that allow manageable FTL?
Rob Wiblin: Faster than light travel?
David Wallace: Faster than light travel. What’s the likelihood you have to do that? My professional assessment is that it’s very low, but I’m sympathetic to the argument that says I might be wrong. I’d love to be wrong. And the benefits to us of having scalable FTL wormhole technology are unimaginable, really.
Rob Wiblin: Right.
David Wallace: So the argument that says we shouldn’t discount the value of that, I find that argument pretty good. And as a society, shall we run a fundamental physics research program so as to hinge on the possibility of that happening, I think that’s a pretty defensible thing to do. That’s the kind of society-level question rather than the individual-level question of what I should personally do, of course.
Rob Wiblin: Yeah. This isn’t something that I’ve thought all that much about. Sometimes people have the impression that at 80,000 Hours we’re doing mathematical expected value calculations to figure out how much utility will be generated by each career. And I think this case demonstrates, well, we wouldn’t do it anyway, but this example demonstrates just how difficult/impossible that could ever possibly be.
Rob Wiblin: I had this non-consequentialist attachment, I think, to humanity continuing to try to figure out the nature of the universe in which we exist. And maybe it will be justified on that kind of basis anyway, especially given global GDP is $100 trillion a year, there’s a bit to go around. Maybe there does feel like something slightly odd about a field that doesn’t have such a clear case for impact attracting such a large fraction, or a pretty material fraction, of the very smartest people. It’s like maybe more of them should be going into the stuff that seems extremely urgent to solve now rather than a bit later. But it’s a bit… I can’t make a totally comprehensive case for that.
David Wallace: Yeah, I can see the case. I mean, I personally very strongly feel the case that says, look, it’s just compellingly central to the human condition to ask and try to answer these very, very deep questions. “What’s the origin of the universe?” I think a society that is purely driven to think in pragmatic terms about the value of asking that question is in some danger of losing its soul.
David Wallace: I forget who it is, but one of the physicists who was interviewed by Congress about building the superconductor back in the 1980s was asked by some senator, “What will this do to defend this country?” And his answer was, “It will make this country worth defending.”
David Wallace: I feel some of the pull of that, but look, if somebody says, “Well, yes, fair enough. But in the here and now, people are starving and we have climate change and we have existential risk problems, so it’s a luxury to worry about that right now,” I feel the force of that argument as well. And as I say, I think there are more practical arguments you can give to the continued value of studying fundamental physics. But I wouldn’t want to imply that’s why I do it.
Physics careers [02:43:56]
Rob Wiblin: Yeah. Maybe a philosopher of physics is the wrong person to ask for the cutting-edge, most useful things in applied physics, but would you like to suggest anything that if someone was going to pursue a career in physics, where you think there maybe is a more obvious case for impact?
David Wallace: Well as you say, this is not my territory. So this is my non-random sampling. I mean, quantum information obviously has the potential to change a great many things. Although equally, of course, that’s a recognized big field. And the possibility of solving and resolving the climate crisis turns a lot on various sorts of electrification issues. It turns on power storage and generation. It turns on more chemical physics questions of how you handle carbon reabsorption. But it’s slightly more blue skies levels… Again, things like superconductivity, things like materials science…
David Wallace: The truth is I’m not in a space to give direct advice here. And I think part of the point is that a lot of this is not predictable, which would go back to my case as to why I don’t think that someone should feel bad about the sort of contribution that comes from trying to do genuinely cutting-edge, more theoretical work. It’s just very hard to predict where that will go. I mean, the strategy as a society of just trying to understand things deeply and seeing where it’s gone has a lot of payoffs to it that are very, very hard to predict.
Rob Wiblin: It has been working pretty well.
David Wallace: Yeah. I mean, it’s a dangerous move to decide that we want to back away from that strategy entirely. And I also think, in terms of what people do, different styles fit different things. And the differential advantage you get from doing something you’re really good at is important. Even at the fine grain of physics. Even if, maybe let’s say, it’s more important to be doing near-term work on developing better batteries than it is to do longer term work on the foundations of solid state physics, those generally require different skill sets, different sort of tolerances for abstraction and different styles. And you can’t really… Nobody is going to do good physics, really important physics, without being really good at it. This is an elitist view, but I think it’s true.
Rob Wiblin: The low-hanging fruit has been taken.
David Wallace: Yeah. People could be very differentially good at different areas. And it’s hard to work that out until you’ve actually done fairly advanced training in some of those areas. So I think somebody coming into thinking about doing graduate work in physics and thinking upfront, “Here is the thing I want to do on utility grounds,” without really putting a lot of weight into which particular things they think they’ll enjoy and be good at, is probably a mistake even in its own terms. Because you’ll do less good and less significant work.
Rob Wiblin: What about philosophy of physics? Are there any important questions in philosophy of physics that seem unduly neglected to you?
David Wallace: Well, most of the ones that I thought have been unduly neglected, I’ve tried to pay attention to, so…
Rob Wiblin: Right, yeah. We could just look at the list on your website of what you work on.
David Wallace: Yeah. I mean, that’s the thing about being mid-career. I think slightly weirdly, one slightly neglected area in philosophy of physics is direct engagement with mainstream physics. I think sometimes philosophy of physics has a bit of a habit of prioritizing very non-mainstream ideas in physics. Some things that only very small sub-communities of physics take seriously. And I think philosophers of physics often put a very high premium on a certain sort of mathematical clarity and conceptual clarity in the theories they study. Which I think has been… In some ways it’s a shame, because if philosophers can bring anything to the table, I think it’s hopefully the ability to form clarity in unclear spaces.
David Wallace: So I think, yeah, it’s been a principle in my career that what I can best do as a philosopher of physics is engage with the sort of empirically very successful, and mathematically successful, but often conceptually tangled state of mainstream physics, and help untangle it a bit.
David Wallace: Where that’s led… I mean, it’s not so much I’d say that that leads you to big topics. It leads you to different angles on topics. So for instance, I think that statistical mechanics and the philosophy of statistical mechanics, questions about emergence, about the distinction between past and future, are really important topics. And they’re not really neglected, but I think certain styles of doing them have been a bit neglected. Similarly, philosophy of quantum field theory, which is our sort of deeper way of doing quantum mechanics, hasn’t been neglected. But it’s only in relatively recent years that people have been doing it in conversation with the way these things are done in mainstream physics.
Subjective probabilities [02:48:39]
Rob Wiblin: Alright. Let’s push on and do a little bit of a smorgasbord of other issues in philosophy of physics, especially ones that have had audience requests. An audience member wrote in and said, “David criticizes the use of subjective probability in physics, but does he think that subjective probability is ever appropriate?”
David Wallace: Yeah, I’m fine with a lot of uses of subjective probability. So if you ask me what’s the probability that Donald Trump will seek the Republican nomination for president, I guess my answer is about 25%. All that really is is a subjective probability.
Rob Wiblin: A statement about you.
David Wallace: Exactly. Yeah. That’s about the odds I’d take on it if you ask me to bet on it. It doesn’t remotely reduce to something like in 25% of Everettian branches Donald Trump seeks the Republican nomination. I suspect he seeks them in either virtually all the branches or virtually none of them. It’s not that kind of thing. I think there’s plenty of roles for that kind of personal probability that’s not going to have any kind of truly objective grounding.
David Wallace: The place where I don’t think that notion of probability is playing a role is in the kind of probabilities that turn up in physical theories. So things like what’s the half-life of the neutron I don’t think can be characterized as that kind of probability. I’m fairly pluralist about the way I think about these things. I think there are objective chances that play a role in physics. They’re mostly Everett branches, but that’s a separate argument. And then there are personal probabilities that are part of our cognitive process to guide ourselves around an uncertain universe. And these concepts talk to each other in important ways, but they’re distinct and neither of them is dispensable.
The philosophy of time [02:50:14]
Rob Wiblin: Let’s spend a moment on the philosophy of time. Can you explain the problem of there being an arrow of time?
David Wallace: So microscopic physics doesn’t seem to care about the distinction between the past and future. More generally, physics of things with not many moving parts doesn’t depend on the difference between the past and future. If I showed you a picture of the solar system, a sped-up video of the solar system, and I asked you whether I was playing the video forward or backwards, you might be able to tell if you knew quite a lot about astronomy, or you paid careful attention to which side the dawn came, but you wouldn’t be able to tell by immediate observation which it is. The physics of the solar system is not very interested in whether you run it forward or backwards.
David Wallace: But the physics of most systems in the world, in particular the physics of systems of lots and lots of moving parts, cares intimately about the difference between past and future. If I were to video a baby playing, or some ice placed in a cup of water, or some plants growing, or a fire, and I were to ask you if I was playing the video backwards or forward, you could work it out immediately. And so you have to ask where did that directedness come from? How do we go from a situation where at the microscopic level or the small number of degrees of freedom level, the physics doesn’t care about the distinction, to a situation where it does care?
Rob Wiblin: And what’s the solution that you think is most likely to be correct?
David Wallace: So at some level it has to come down to something that breaks the symmetry. Just as a matter of logic, if you put symmetry in and get asymmetry out, you had to smuggle some asymmetry in as well. The fundamentals here are not original to me. At some level I think that solution probably has to be cosmological. At some level we’re talking about certain features of the initial state of the universe, which in some respects encode the fact that the direction of time is one way rather than another way. How to say what those features are, I’m not sure. I think the right way to say it is probably that the initial state of the universe has to be relatively simple.
David Wallace: That is, it doesn’t contain the sort of extremely delicate correlations that would have to be present for a system that was sort of evolving backwards in the opposite direction to the normal thermal directions we see. There’s a lot of different routes people can go. Some people would say this shows a fundamental distinction between past and future. Some say it’s a difference of our inference or a difference of cause. I’m skeptical about most of those moves. I think we want a dynamical understanding in physics terms, and I think if we want an understanding in physics terms, it probably has to be cosmological.
Rob Wiblin: Okay. So just to explain that for the audience. We have this puzzle that the laws of physics don’t seem to feature time. The particles bounce off of things and that looks, at the atomic level, exactly the same whether the tape is being played forward or backwards. So for example, if you had a bunch of gas equilibrized, sitting in a bottle, and that’s what you’re talking about, then it looks the same played forwards and backwards, because the gas molecules are just bouncing back and forth and there’s no difference. And yet, most of the phenomena that we deal with in daily life can only go one way. Like in a fire, a log burns, and yet it never reconstitutes itself, not ever. It’s not just improbable, a log just never comes together with the molecules of air suddenly deciding to become wood.
Rob Wiblin: And so what’s up with that? I guess you could try to figure out some way of incorporating time as a fundamental thing, or an arrow of time as a fundamental thing within the laws of physics. But if you’re not taking that approach, another approach would be to say that there was something about the way the universe was set up, say at the Big Bang or very early on, that kind of had it sort of wound up like a clock somehow, such that it was massively out of equilibrium. And it’s that which is causing the processes to look different in one direction than the other because the initial state was so unnatural, so unstable, now that it’s trying to get to a more probable state, I guess one might say. In order to get that certain things have to happen that look quite different in one direction than the other. Is that kind of right?
David Wallace: It’s kind of right as a statement of the problem, and it’s kind of right as an exegesis of what’s a commonly preferred solution. I don’t think it’s right in terms of what the solution actually is.
Rob Wiblin: Okay, right.
David Wallace: I don’t think that way of doing the cosmology to get the direction right really solves the problem. And a way of seeing that is, let’s suppose that the very early universe was very far out of equilibrium. And there’s questions about what that means exactly. But let’s stipulate we understand it. Well, that doesn’t give you any particular reason to think that the slightly later universe isn’t going to be even further out of equilibrium. The mere statement that it was a very unusual out-of-equilibrium state doesn’t itself tell you anything about—
Rob Wiblin: …the direction it will go.
David Wallace: —where it’s going to go next, yeah. What you actually need to do is to make some kind of assumption that the initial state of the universe wasn’t built in some very carefully correlated way that was going to make it end up in an even further from equilibrium stage. And I think once you’ve put in that assumption about the fact that there isn’t any sort of very fine correlative structure built in in the early universe, then actually, that’s all you need. You don’t then need additional claims about what the macroscopic state of the universe was like. If you can just put in a relative simplicity of the early state, then it should take care of itself from there on.
Rob Wiblin: So if I’m understanding this right, once you have those initial conditions and you specify that it hasn’t been kind of gerrymandered in some way such that it’s going to become even more unlikely over time, then I guess you get this statistical play out. Where it’s more likely for molecules to move in such a direction so that they diffuse evenly over space, and it’s more probable for heat to spread out than it is for it to concentrate. And so, because there’s more ways that that can happen than the reverse, that is why we see an arrow of time in almost all situations that we look at.
David Wallace: Yeah, that’s right. I mean, that’s a fairly standard physics argument, and it’s clearly got the ring of truth to a large extent. I mean one way to see the puzzle we started with is that, if you’re not careful, arguments of that kind prove too much. Once you’ve established that it’s really unlikely that my system will end up moving into a smaller state, and much, much more likely to move towards equilibrium, then you want to say, “Well, why doesn’t that argument work backwards?” Why shouldn’t I say it’s really unlikely that the current state came from one of these really low-probability out-of-equilibrium states? Isn’t it much more likely that the current state came from a higher equilibrium state? And that’s proving too much. We don’t want that to be the case. I mean, Roger Penrose has a lovely quote that says, “The difficulty in statistical mechanics is not to establish why entropy increases into the future, it’s to understand why entropy doesn’t also increase into the past.”
Rob Wiblin: Now I get it. So you’re saying, so the particles are just moving about, and given that the laws of physics are symmetric in time, it’s very easy to explain why maybe entropy is going up forward in time, but then why doesn’t the exact same logic apply going backwards in time?
David Wallace: Exactly right. And one way to think about the reason one ends up appealing to cosmology is, if you say the current state of the universe is not a really specially organized gerrymandered improbable state, now you can understand why entropy will go up into the future, but you don’t understand the past. So you can say, well, instead of saying that’s a statement about the current state, I’ll say that’s a claim about the state 1 million years ago. And now you can explain why entropy has been going up for the last 1 million years. But 1 million years ago, you now can’t explain why it wasn’t going up further. The only way of getting around that regress is to move your kind of special moment—
Rob Wiblin: …to the start.
David Wallace: —all the way to the start, exactly.
Rob Wiblin: So maybe rather than think about things going forward, we want to say, “Why is the universe set up now such that going backwards, however far we see, we see entropy going down?” And that’s the real mystery. And I guess that then just raises the question of why were the initial conditions of the entire universe the way that they were. So I suppose we’ve turned the problem of the arrow of time into just a different problem about why the universe is what it is.
David Wallace: Yes. Although I think the way of making that a little less mystical is to say, nevermind why was the initial state of the universe what it was, what do we need to assume about the initial state of the universe? What features does it have to have in order for thermodynamics, the entropy-increasing things to behave properly?
David Wallace: Now, the further question about why it has those features is then a question for, in principle, future physics. Maybe ultimately something deeper than that. But just pinning down what has to be assumed about that state in order for statistical mechanics to work, that’s something that we can try to answer. That’s a relatively controlled problem. So people have said that what we need to assume is that the early universe was very unlikely. That’s a common line that people take. I think it’s wrong, but I think it’s understandable that it’s argued for.
David Wallace: I think actually the right thing to say is something like, we need to assume that the early universe state wasn’t really unlikely, in the sense that it didn’t contain the sort of very delicate gerrymandered correlations that you need to have present if systems are going to behave weirdly into the future.
Rob Wiblin: Are there any other big questions in the philosophy of thermodynamics that you’re excited by?
David Wallace: The other big question I’m excited about here is probability, again. So statistical mechanics, the kind of probability basis of thermodynamics, is the other big place in physics that we see probabilities. And you see them in quantum mechanics, we’ve talked a lot about that, but we also see them in statistical mechanics. We saw them there first, actually. Physicists started talking about probability in the late 19th century because of statistical mechanics before the quantum revolution. And there’s a common line that says these are just conceptually different notions of probability. And I think that’s probably incorrect. I think in a quantum world, the probabilities of statistical mechanics probably, so to speak, become quantum probabilities. I think these notions merge in quite an interesting way.
David’s experience at Oxford [02:59:51]
Rob Wiblin: Alright, we’re going to have to let you go before too long so that you can play with your kids. I’d like to talk about a few personal things, maybe just to wrap up. One of my colleagues had you as a tutor back at Oxford when they were studying as an undergrad. And they remembered that you were responsible for the physics and philosophy admissions process for a while, and they recalled being impressed with some changes that you made, or tried to make, to the process in order to make it more evidence-based and equitable and fair. Do you mind elaborating on what those changes were and why you wanted to make them?
David Wallace: Sure. I mean, it gets a little bit into the weeds of Oxford’s processes. But I was mostly concerned with various things that were supposed to make it genuinely true that it didn’t matter which college you applied for. One of the issues that Oxford has is that it runs a very decentralized admission system where each college has an admissions team that runs its own process. What’s supposed to happen is that the different colleges coordinate in various ways and run a shared framework, and interview each other’s students to calibrate the process. And in practice that works better in some subjects than others. When I started being involved then it worked extremely well in physics, in my judgment. And a lot of what I was involved in doing was trying to see how we could make this work in physics and philosophy in ways that were fair, that made sure that no one was being disadvantaged. And also that people who wanted to go to colleges with concentration in a subject could do it.
David Wallace: So physics and philosophy is a small subject. In Oxford it’s better to be taught in a college where you’ve got a reasonable critical mass in your subject. So part of what I was doing was trying to organize a system to direct candidates preferentially to colleges that would sort of strongly support the subject. And I got rid of the admissions essay.
Rob Wiblin: Why’s that?
David Wallace: I thought it was probably unfairly advantageous to a certain sort of applicant. Specifically applicants who’d done a humanities A level subject. I’ve put this in equity terms, but to be honest I was more concerned just with wanting stronger students. But there’s a route to being very good at physics and philosophy that involves coming into it with a relatively standard British math, further math, physics and chemistry that you’re often pushed towards if you’re a sort of a science-y oriented candidate. But students of that kind were being put off applying because they were worried they had to have an essay.
David Wallace: And the essay wasn’t actually, to be honest, being given very high weight in the process. If it was given high weight and if it was genuinely an important differentiator, that would have been a case for keeping it and for doing a lot of outreach to persuade people that they shouldn’t worry about that if they came from a non-humanities background. But it wasn’t really doing a great deal of work in the process. Well, again, I don’t want to get too far into the weeds of Oxford admissions, but for various reasons the differential assessments were more being made on the basis of physics and math strength. So it was just a piece of not very functional material that was being kept because people hadn’t thought about it carefully enough, which was putting off students we didn’t want to put off.
Rob Wiblin: Yeah. My colleague mentioned that he thought that you might’ve decided to give more weight to a written test, because it turned out that on some analysis that was more predictive of actual performance than an interview where people are able to just show what social class they’re from by speaking particularly well.
David Wallace: Ben’s giving me too much credit…
Rob Wiblin: Okay.
David Wallace: That was a general feature of how the physics system was working. I was supportive of helping develop some of those as it applied to physics and philosophy, but it was more something about the way Oxford was running generally. That said, I think Oxford’s interviews were, and are, I think, much better in this sense than you might think from the general sort of interview literature. So there’s loads of evidence that unstructured interviews become chats and advantage people who just kind of just relax into it and then draw on lots of irrelevant stuff.
David Wallace: But at least in physics and in philosophy an admissions interview is a really very structured thing. It’s in many ways more like an oral exam. And nobody in the process has any interest at all, for instance, in why you want to come to Oxford. In physics interviews I don’t think I ever asked anyone why they wanted to come to Oxford. I think my colleagues in mathematics used to ask people, but they didn’t write down the answer. They only asked because they knew the students would have prepared an answer.
Rob Wiblin: I see.
David Wallace: I’m not doing it anymore so I can give away my questions. I used to ask people, if they roll two dice, what was the probability of getting two sixes? And then I’d ask them what that statement meant. And we’d explore what they thought probability statements were, which is almost never something they particularly run into in a school context. And actually, you could relatively quickly see the people who had kind of been prepped by a lot of things they’d read that they wanted to show knowledge of, and it was just really annoying. You had to get them to shut up so they could carry on actually thinking about the question you asked.
Rob Wiblin: I see, yeah.
David Wallace: I kind of came away with a view that actual preparation for interviews in Oxford is mostly useless. I mean, there’s one very good bit of preparation for Oxford interviews that private schools provide in Britain, which is that they educate people really well. There’s definitely advantages in an Oxford interview. But the actual prep is largely useless, and largely done by people who have a very outdated picture of how Oxford admissions works.
Rob Wiblin: So Ben recalled that you were perhaps his most thoughtful and dedicated tutor. And I guess Oxford famously has a lot of old practices, and things haven’t always been updated for the 21st century. And he recalled that most tutors would require him to submit handwritten essays, and then would scrawl notes on them that would be incredibly hard to read. But you had managed to bring things into the 20th century, and so allowed people to submit things electronically, and then you would mark them up on a tablet. Which is a lot easier to read and a lot easier to record. What prompted you to do that? And I guess also what do you think of the traditionalism of Oxford in your experience?
David Wallace: Sure. I mean, what prompted me was partly my own workflow. I think what literally prompted me was just moving to tablets that were more manageable and more functional. This is like the dawn of the iPad, I suppose. But from my point of view, if the essay is submitted in paper form and I need to… Even if it was an email, I had to print out the essay, mark it up by hand, and remember to pass it back to the student. And my own workflow gets more fiddly and I don’t have an immediate record of what the student was doing. So my reasons were largely self interested. If it was helpful to the students, then that’s so much the better.
David Wallace: Part of the problem with Oxford is that it’s torn between people who don’t want anything to change at all, and people who want everything to change. And I think protecting a lot of the really core unusual features of Oxford requires fiddling around the edges. So for instance, the core structure of the tutorial system, the basic idea that you’re being taught in very small groups by people who are involved with your education and followed you through it, also who are academically very strong in their field in various ways, that’s I think almost the definitional central feature of Oxford.
David Wallace: Distinguish that really sharply from… I can’t think of an example now because I’ve been gone too long, but there are so many very small things that really just don’t matter at all. You know exactly who is responsible for exactly which bit of decisions. How do we distribute exam essays for marking, all sorts of stuff like this, that actually that’s the place which you need to move in order to make Oxford be in accord with the times. And I haven’t been directly involved with it for five years now, but my understanding is that the pandemic, for all its horrors, has been a bit of a breath of fresh air there, because suddenly you’re forced to change a lot of things. And it suddenly realized it wasn’t so hard. So I think Oxford, having been resisting for decades the idea that they could electronically distribute exam essays for marking, brought it in in the space of a few weeks when they realized that they had to, to handle the pandemic.
Rob Wiblin: Yeah. I heard from a bunch of people that it has been really… The things that people had pushed for for so long and that were said were impossible were suddenly implemented very quickly, as soon as it actually became necessary. I suppose that the benefit of being forced to improvise and forced to change things is especially large when you’re at an organization that has a very long standing culture, and has been around a long time, and so might have built up a lot of practices that they’re resistant to change. So that is very cool.
Rob Wiblin: This has been, I’m not going to lie, one of the more challenging interviews that I’ve done. I think I might have to have an early night tonight. I’m going to be quite tired. But it has been so fantastic to actually be able to get an expert in their field to answer a whole lot of questions that I just could not find the answers to despite searching during the process of preparing for this interview. And I think the audience is going to appreciate some of those answers as well. My guest today has been David Wallace. Thanks so much for coming on the show, David.
David Wallace: Thank you.
Rob’s outro [03:08:43]
Rob Wiblin:If you’ve made it through over 3 hours of chat about quantum mechanics and the social impact of physics research, you’re probably the sort of person who should think about applying to speak with our team one-on-one for free.
We’ve made some hires and so are able to speak personally with more readers and listeners than we’ve ever been able to do before.
Our advisors can talk over your philosophical views, which problem might be a good fit, look over your career plan, introduce you to mentors, and suggest specific organizations that might be able to use your skills. Just go to 80000hours.org/speak to learn more and apply.
Alright, the 80,000 Hours Podcast is produced by Keiran Harris.
Audio mastering by Ben Cordell.
Full transcripts and an extensive collection of links to learn more are available on our site and put together today by both Sofia Davis-Fogel and Katy Moore.
Thanks for joining, talk to you again soon.