If we can’t control MechaHitler, how will we steer AGI?

It’s Chana and Aric again.

Firstly, we cannot believe the response to our last video. Thank you all so much for creating an initial swell of views and support, for your feedback and comments and subscriptions. We’re at 186,000 subscribers, and 6 million views. Truly, truly, incredible; we’re so delighted to be doing this with you.

Also, um, feel free to do that all again? Because AI in Context has a second video for you.

There are two types of people in the world:

1. People who are understandably very confused about why the word “MechaHitler” appears in the title of this sober, thoughtful, serious blog post.
2. People grimacing because they know why “MechaHitler” appears in the title of this sober, thoughtful, serious blog post.

If you’re in category 1, here’s the summary:

You might have heard of Elon Musk’s AI model, Grok, which can interact with users and post directly to Twitter. Earlier this year, it suddenly turned from being a fairly neutral commentator on events to a sexually-harassing, Nazi-minded troll calling itself ‘MechaHitler.’

Our new video is about that incident, getting into the wild details of the MechaHitler story that even the terminally online may have missed.

But it’s also about putting this news story in context: how Elon Musk and his AI company xAI function, where the broader AI industry is going, and what this incident reveals about our readiness for more powerful AI systems.

Aric and the team did hundreds of hours of research so we could describe exactly what happened.

Why MechaHitler matters

We’re only as safe as our least safe AI company. The more reckless and irresponsible one of them is, the more reckless and irresponsible they’re all incentivised to be.

We were lucky that xAI, the company behind Grok, could resolve the issue within 16 hours (though only by reverting to an earlier version). But, we argue, the incident heralds bigger, much worse things in the future — unless we take this as the warning that it is.

In fact, the MechaHitler incident might be one of the best warnings we get before things go really off the rails in a way we can’t just quickly patch up.

Fortunately, society at large — and the inspired, talented folks reading this newsletter — have a lot of ways to respond. We at 80,000 Hours are excited to help you work out where to go from here.

There’s need for:

• Policy experts crafting and implementing safety standards
• Technical researchers working on AI control and alignment
• Evaluators testing systems before deployment
• Communicators helping society understand these risks

Outside your career, you can also help hold AI CEOs accountable by tracking safety promises kept and broken, and speaking up on social media — these folks are surprisingly online (watch the video to find out more).

Why watch the video

Whether you’re in the market for information, alarm, or the dulcet tones of Aric’s voice, we’ve got you.

The video includes:

• Getting a real sense of the story, wild details included: It involves unintended code changes, intentional jailbreaks, runaway feedback loops, and defense department contracts. We include details we haven’t seen reported on anywhere else.
• Expert interviews: Including with New York Times columnist and author Kevin Roose, who once endured attempted seduction by a wayward chatbot.
• Historical context: How this connects to previous AI failures (like Microsoft’s ‘Tay’ and Microsoft Bing’s ‘Sydney’) and what we’re not learning from them.
• How to think about the players: What kind of a company xAI is, and Elon Musk’s history as (believe it or not) a prominent advocate for AI safety.
• The bigger picture: As AI systems gain real-world capabilities through robotics, defense applications, and autonomous decision-making, we risk dangerous loss of control and extreme concentration of power.
• What you can do to help: We lay out some ways people are eliciting change from AI CEOs, and make the case for speaking out or working to change things yourself.

It can be hard to know what to make of AI news like MechaHitler. Is it silly sensationalism? Does it actually matter? We want to help put the news in context: what the story actually tells us about the safety landscape, about how competent AIs are, and whether it seems like things are on track. For now, they’re not. But it doesn’t have to be that way.

Check out our video here, and give us feedback on it here. We want to know your thoughts, and don’t hold back — did anything seem wrong or confusing? Was there anything we missed?

Other resources

• Check out our other videos examining transformative AI
• Apply for free one-on-one career advising if you’re interested in working on these problems