When Genius Fails: The Intellectual Arrogance of the AI Labs

Leopold Aschenbrenner, known for being part of OpenAI's Superalignment team (from which he was fired over alleged leaks, which he disputes) and for a 2024 essay on the imminence of AGI, was running a $20 billion hedge fund called Situational Awareness LP. The fund has now blown up, providing what the author calls a "$20 billion demonstration" that being an expert in one field doesn't make you an expert in all. The author, an ex-hedge fund professional with an AI background, uses this as a springboard to critique the lack of intellectual humility in frontier AI lab culture, which he sees as extending into many verticals beyond money management.

The author compares the blowup to the canonical example of Long-Term Capital Management (LTCM), which combined Nobel laureates and Wall Street's best bond traders. LTCM quadrupled investors' money in four years before spectacularly collapsing in 1998, requiring a Federal Reserve-orchestrated bailout by Wall Street banks. The author notes that many more hedge funds have blown up than have been consistently excellent, and that Situational Awareness speedran the entire rise-and-fall process.

He clarifies that the issue isn't peak returns, since a lucky roulette player can hit 3,100% returns, nor is it simply "beating the market" in a bull market or bubble. Instead, what matters for a hedge fund is consistency across bull and bear markets, because that's what justifies the expensive fees. The author also mentions the existence of markets other than stocks, such as bonds, implying a broader perspective that AI experts may lack.