Seems like a good time for an update on a graph from my 2015 post on AI.

- Scott Alexander Eli Lifland Thomas Larsen The AGI death race is in nobody's interests. Step one is to make sure that's common knowledge. Step two is to hammer out the details of a treaty, such that it's rational for both sides to accept the treaty and irrational for either side to try to cheat (e.g. because cheating

Thanks Jon Scott and Fox News for yesterday’s AI discussion. It’s a troubling sign when the His Holiness moves faster than the United States Congress to recognize we’re at a pivotal moment for humanity.

Nate Soares and I are publishing a traditional book: _If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All_. Coming in Sep 2025. You should probably read it! Given that, we'd like you to preorder it! Nowish!

Most important book of the decade:

The best move in Russian Roulette is to stop playing.

Ursula von der Leyen: "When the current budget was negotiated, we thought AI would only approach human reasoning around 2050. Now we expect this to happen already next year." ec.europa.eu/commission/pre…

JD Vance did something every single political leader should do.

As a general rule, basically everyone has the correct commonsense view that AI is a dangerous technology. The issue is galaxy brain academics + VCs etc. who overthink it and come to wrong conclusions, and industry people who have an incentive to downplay it.

Should we be racing to build superintelligent AI? Here's my conversation with "Keep The Future Human" author Anthony Aguirre and MIRI CEO Malo Bourgon who both strongly believe we shouldn't... ...a controversial take here on TPOT, but given the stakes in either direction, it's

Vance was asked if the US Government is capable of pausing AI in a loss-of-control scenario. “The honest answer to that is that I don’t know, because part of this arms race component is: if we take a pause, does the People’s Republic of China not take a pause?” Great to see

One very weird thing is that AI companies are trying really really hard to make it clear that they're going for RSI, and functionally no one on Capitol Hill believes them (or even understands this is the literal Business Plan)

Banger stats from AI Public Opinion The AIPI JD Vance Americans want government action on AI!

FUD doesn't make the case for the anti-thesis. It works by raising the costs of making the case for the correct thesis. No one takes voluntary commitments seriously, but they make rallying against corps and passing laws harder.

You think politicians won't listen about AGI risk? Or that you can't talk to them? Think again! This is THE definite piece writing up how ControlAI did it, and you can too!

Thinking that mechanistic interpretability is the key to understanding AI safety is like thinking that neuroscience is the key to understanding political science.

🚨 Mind blowing interview with Turing award-winning Yoshua Bengio for BBC Newsnight one of the three founding fathers of AI, is now warning: “The worst-case scenario is human extinction.” AI isn’t just risky — it could end us. 1/10 🧵 📺 youtu.be/c4Zx849dOiY?t=0

This will be fascinating. tylercowen, Tom Davidson, Connor Leahy and Mike Webb debating AI in London on 1 July. Register below.

Had a great time talking with the ControlAI team about my time at OpenAI and other light & breezy AI topics - full episode link below