Ken Griffin calls generative AI useless

Ken Griffin’s point was narrower than “AI is useless.” In a Stanford Graduate School of Business conversation published on YouTube on May 4, 2026, the Citadel founder said generative AI is not proving useful for the core hedge-fund task of finding durable market edge. He was speaking with Stanford finance professor Amit Seru at the 2026 Stanford Leadership Forum. (youtube.com) The specific claim making the rounds is that generative AI may help “for the next five minutes,” but not much beyond that in markets. I could not independently verify that exact five-minute wording from a primary transcript in the sources available here, but Griffin has made the same underlying argument on the record before. In an October 15, 2025 Bloomberg interview at the JPMorgan Robin Hood Investors Conference in New York, Griffin said: “With GenAI there are clearly ways it enhances productivity, but for uncovering alpha it just falls short.” (bloomberg.com) That distinction matters. Griffin was not arguing that AI has no business value. He was separating two uses: productivity inside a firm versus alpha generation inside a hedge fund. In his telling, generative models can help people work faster, but they have not yet shown they can reliably produce market-beating returns. (bloomberg.com) That view also fits how Citadel describes itself. Stanford’s event page says Griffin built Citadel around “advanced quantitative analytics and powerful software,” and says the firm’s culture is defined by “continuous learning, problem-solving, and meritocracy.” Citadel’s own careers pages emphasize quantitative research, engineering, and fast deployment of research into investment decisions. (gsb.stanford.edu) So the subtext of the clip is less “old-school finance rejects AI” than “one of the biggest hedge-fund managers is drawing a line between chat-style AI and the statistical methods that already run much of modern trading.” Citadel says its quantitative researchers build models that power systematic strategies and inform investment decisions, while its research engineers and traders work on automated systems at scale. (citadel.com) The recruiting angle follows directly from that. Citadel’s public hiring materials say it recruits across quantitative research, trading, engineering and operations, and repeatedly stress analytical ability, technical depth and problem-solving. That makes Griffin’s comments read as a hiring signal as much as a technology view: the firm still wants researchers and engineers who can build and test trading systems, not just prompt large language models. (citadel.com) There is also a timing point. The Stanford appearance was published on May 4, 2026, but the same basic skepticism had already appeared publicly in October 2025. The newer clip appears to have revived and sharpened a position Griffin had already stated: generative AI may improve workflow, but he does not see evidence that it is delivering persistent hedge-fund alpha. (youtube.com) If you want the cleanest verified takeaway, it is this: Ken Griffin has publicly said generative AI helps productivity but “falls short” at uncovering alpha, and Citadel continues to present itself as a firm built around quantitative research, software and recruiting for technical talent. (bloomberg.com)