Citadel founder Ken Griffin claims "generative AI is useless": cannot discover Alpha, Wall Street still needs to manually beat the market.

Ken Griffin at the JPMorgan Robin Hood conference bluntly stated that GenAI has not yet helped hedge funds uncover Alpha, prompting the market to reassess the practical utility, risks, and human-machine collaboration of generative AI (background: ChatGPT analyzes the pullback moment in the crypto market: Is it now the “halftime” of the bull run?). The JPMorgan Robin Hood Investor Conference held in New York yesterday (15th) attracted heavyweight figures from Wall Street and Silicon Valley. Citadel founder Ken Griffin poured cold water on generative artificial intelligence (GenAI) during his speech, pointing out that while GenAI can enhance productivity, “it is still lacking in uncovering Alpha (excess returns).” “GenAI can indeed improve productivity, but when it comes to uncovering Alpha, it still falls short.” In the midst of the hot AI concept stocks, Griffin's attitude has led the outside world to start reflecting on the true value of this technological wave. The cautious stance behind Citadel's practice of Griffin's viewpoint does not indicate that Citadel is resistant to new tools; the company has already integrated GPT-like models into its research process to assist with document summarization, real-time insights, and chatbot-assisted analysis, but key decisions are still led by seasoned investors. In the second quarter, Citadel reduced its long positions in Broadcom by about 82% and in Palantir by approximately 48%, yet quadrupled its stake in Nvidia to over 8 million shares. These moves reflect that Citadel only makes significant bets when it is certain of competitive barriers and hardware leadership, while maintaining distance from overvalued software stocks. Citadel focuses on balancing market momentum and risk control. Market optimism for GenAI: Increased profits? In contrast to Griffin's reservations, some research institutions continue to release positive assessments, stating that hedge funds that adopt GenAI early have seen annual returns increase by 3–5%, particularly benefiting quantitative and stock strategies. GenAI provides faster market signal extraction and portfolio stress testing through the integration of synthetic data, large language models, and modular workflows. According to Moody's CreditView Blog, Agentic AI can monitor trades 24/7, flagging anomalies and enhancing risk management. AIMA's research also shows that 95% of hedge fund managers have used GenAI tools, and 90% of investors expect to see positive contributions within three years. Challenges in Wall Street adoption However, beneath the optimistic data, there are challenges that cannot be ignored. The CFA Institute analysis points out that GenAI models underperform in high-complexity scenarios such as geopolitical impacts, making them prone to inaccuracies during significant fluctuations. In terms of regulation, issues of model opacity and bias remain to be resolved, as once the judgment basis cannot be traced back, it will put pressure on financial institutions' compliance. To unleash GenAI's full potential by 2030, the hedge fund domain still needs 140,000 skilled specialists, posing challenges for infrastructure and talent development. Widespread adoption may also reinforce market homogeneity, increasing systemic risks and concerns about the weakening of human analytical abilities. Hybrid human-machine collaboration: A more realistic blueprint In the tug-of-war among efficiency, risk, and regulation, many practitioners focus on the “human-machine collaboration” model, wherein AI performs initial filtering and pattern searching on vast amounts of data, and investors add their experience and situational judgment to ultimately confirm strategies. This model effectively leverages AI's computational advantages while retaining human sensitivity to extreme events and unstructured information. For Griffin, caution does not equate to rejecting technology but ensures that each investment can yield controllable returns. Once the market understands this point, it may be able to more calmly assess GenAI's position: it is merely an assistant, not a savior. In summary, GenAI is undoubtedly reshaping data processing and operational workflows; short-term profit increases may only reflect efficiency improvements, but stability in generating Alpha and discovering ways to outperform the market at the hedge fund level remain variables. As Wall Street and Silicon Valley discuss the next wave of innovation, Griffin's words remind investors that true competitive advantage comes from the precise use of tools, risk management, and clear judgment not clouded by popular narratives. Related reports Why do algorithmic stablecoins repeat the “death spiral” from Luna to USDe? Taiwanese stablecoin infrastructure company Odin is going public in the US! Directly listed on NASDAQ on 10/16, stock code OWLS Latest interview with Arthur Hayes and Tom Lee: The future of DAT, stablecoins, and prediction markets. This article was first published in BlockTempo, the most influential blockchain news media.