Aschenbrenner bet everything on AI — and lost
Leopold Aschenbrenner’s AI hedge fund Situational Awareness just went from $45 billion in assets at the start of July to $10 billion at the start of August after Aschenbrenner had to sell the fund’s public stock holdings to Ken Griffin at Citadel to pay his margin debts. [Bloomberg, archive]…
Leopold Aschenbrenner’s AI hedge fund Situational Awareness just went from $45 billion in assets at the start of July to $10 billion at the start of August after Aschenbrenner had to sell the fund’s public stock holdings to Ken Griffin at Citadel to pay his margin debts. [Bloomberg, archive] Aschenbrenner is just 24. He’s been an AI doomsday crank since he was a teenager. He started at Columbia University at 15 studying economics. He started an effective altruism group there. After college, Aschenbrenner worked at the FTX crypto exchange’s charity arm — more AI doomers. Then he went to OpenAI to research how not to cause the AI apocalypse. I first heard of Aschenbrenner in 2024, from his essay “Situational Awareness,” on how AI was going to hit superintelligence. [blog] The essay is a love song to the ever onward and upward AI future. Even as the chatbots he was working on had topped out their S-curve of progress long before. I particularly liked the bit about speedruns in Minecraft. This is the sort of thing Aschenbrenner puts forward as evidence of the power of superintelligence: As a visualization of how wild this could be, look at some Youtube videos of video game speedruns, such as this one of beating Minecraft in 20 seconds. Now imagine this applied to all domains of science, technology, and the economy. The video Aschenbrenner links to is a computer doing a human-mapped speedrun. He is claiming the existence of video game speedruns like this means AI will take over! Because what if an AI could do this, huh? And also, what if Minecraft was real life? The AI could speed run you! In Minecraft! A lot of people were wowed by his essay. People who didn’t know anything about tech, but they sure did appreciate a curve that pointed upward. He got a retweet from Ivanka Trump. [Twitter, archive] Aschenbrenner founded a hedge fund, also called Situational Awareness, to put his ideas on AI into practice. With other people’s money. [Situational Awareness LP] Aschenbrenner had no experience as a trader. But he sure did have the right friends: [FT, archive] Aschenbrenner captivated his investors with bold ideas about AI drawn from his network comprising effective altruism adherents — the controversial “do good” philosophy whose devotees populated both the collapsed crypto exchange FTX and AI powerhouse Anthropic. “Controversial” means they’re famous for being weirdos with too much money. In this case it means a kid can get serious money from them to do a job he has never done. But he did take the job seriously: Obviously not blowing up is sort of like task number one and two, or whatever. Good to hear. Situational Awareness went great guns — because it was buying into a bubble. In the finance markets, you can talk any gibbering lunacy and nobody cares if you’re making money. Even if they know the reason you’re making the money is stupid. In fact, Aschenbrenner literally had no plan if the number went down: [NYT, archive] What was Mr. Aschenbrenner’s plan if the A.I. revolution didn’t pan out quite as hoped? The hedge-fund founder had no detailed response, the investor recalled. Mr. Aschenbrenner simply truly believed it would all work out. Then Aschbrenner did what a lot of inexperienced traders do — he figured the good times would go on forever. It’s a new paradigm! Things are different now! So he made a lot of leveraged bets — where you borrow money to trade more. These have more chance of upside — and more risk of downside. Over July, some numbers went down. In particular, memory maker stocks had a downturn. Aschenbrenner was caught off guard and he had to frantically sell other stocks: When Intel reported strong earnings on July 23, its share price surprisingly fell and speculation began to mount that someone was aggressively selling into the market. That seller was Aschenbrenner, according to multiple people familiar with the matter. “He was trying to recover his losses,” one of the people said. “He had the illusion of still being in control … he didn’t have control at all.” On 24 July, Aschenbrenner wrote to investors to reassure them everything was fine! Also, volatility happens sometimes. But don’t worry about it. At the same time, he was calling around the investors, offering them deals to put even more money into the fund. By last Wednesday, Aschenbrenner’s margin positions were too much for his lenders. Goldman Sachs called wanting its money back. Aschenbrenner called around the larger Wall Street hedge funds to sell off what he could fast enough to cover his margin debts. Citadel did the deal to buy Situational Awareness’s public stocks on Thursday. [WSJ, archive] Situational Awareness is still alive. But smaller. One unnamed person involved in the whole mess told the Financial Times: The real question is how did anyone invest so much money, and then lend so much money, to a . . . kid with no personal experience and no infrastructure? Because when a bubble’s on the upward slope, everyone looks like a genius. Even as the slightest thing rattling can take it down. The whole AI bubble is built on leveraged investments, dubious valuations of private equity shares, circular financing, lending to your customers to keep your sales numbers up, and off the books ventures to hold the debts. It’s a giant bomb, it’s going to explode, and it’s going to take out a ton of stuff. Situational Awareness was the simple version and it failed because it was one weirdo riding a number going up. But reality does not work like a Minecraft speedrun. Video — PodcastSource: Pivot to AI — Published — Category: Business