AI Pioneer Who Built First Hedge Fund Won't Trust ChatGPT With His Money
Vasant Dhar helped bring AI to Wall Street in 1994, but remains cautious about today's generative AI tools for investing.
Few people have more credibility on the intersection of artificial intelligence and financial markets than Vasant Dhar. A professor and quantitative researcher, Dhar was among the earliest practitioners to apply machine learning to trading when he helped build one of Wall Street's first AI-driven hedge funds back in 1994 — long before the term "AI" became a marketing buzzword attached to nearly every financial product on the market.
What makes Dhar's perspective particularly worth examining is precisely the skepticism he now brings to the current generation of generative AI tools. Despite his pioneering role, he stops short of endorsing systems like ChatGPT for managing or guiding personal investment decisions. His hesitation speaks to a broader and underappreciated distinction in the AI landscape: the difference between narrow, statistically grounded models trained specifically on market data and large language models designed primarily to generate fluent, human-sounding text.
Read more Micron Stock Slides Toward Worst Monthly Loss in Over a Decade →
The concern isn't that today's AI is unsophisticated — it's that it's sophisticated in the wrong ways for financial decision-making. Large language models are optimized to produce confident, coherent responses, a trait that can be deeply misleading in a domain where uncertainty quantification and intellectual humility are essential. Markets punish overconfidence, and a system that rarely says "I don't know" is a poor fit for an environment defined by incomplete information and structural change.
For everyday investors, Dhar's caution carries a practical implication: the AI tools most accessible to consumers are not the same tools that have earned credibility in professional trading environments. Institutional-grade quantitative systems are built on rigorous backtesting, disciplined risk controls, and domain-specific training pipelines. Asking a general-purpose chatbot for a stock tip occupies an entirely different — and far less reliable — category.
Dhar's three-decade vantage point offers a useful corrective to the hype cycle now surrounding AI in finance. His message to investors isn't to ignore AI, but to understand what kind of AI they're actually dealing with. Continue reading at MarketWatch.com
