(Bloomberg) -- AQR Capital Management’s Cliff Asness says the quant firm is actively deploying artificial intelligence — including in the core parts of the investing process that used to be the co-founder’s domain.
The technology is being used to help combine and weight the various investment factors that guide stock picking, thereby steering the nuts and bolts of an allocation method where Asness previously used to issue the final call, he said in an interview on Bloomberg Television on Tuesday.
“AI’s coming for me now,” he said. “It turns out it’s annoyingly better than me.”
The Greenwich, Connecticut-based firm, which runs about $116 billion, is also using AI to generate trading signals from text and boost productivity by speeding up coding, for instance, he said in the interview. It’s a big statement of confidence from a quant who once doubted if buzzy new trends like big data would ever affect the basics of the business.
The likes of AQR typically buy and sell stocks based on factors, or characteristics of shares that are deemed by statistical models to predict relative performance. Combining these factors and determining how important each of them should be is another science altogether, as Asness well knows. In late 2019, he said it was time to raise the weights of the value factor that favors cheap shares — only for this kind of strategy to slump anew versus growth stocks during the pandemic tumult.
In recent years, AQR has authored academic papers on applying machine learning — the AI sub-brand focused on parsing patterns in data — to finance. While old-school quants tend to be warier of overfitting, or mistaking noise for lasting patterns, Bryan Kelly, the firm’s head of machine learning, has preached the virtues of more complex models that can now be built with newer techniques.
It all underscores how the AI-powered era is delivering tangible wins for systematic players — and the shift away from traditional factor investing.
“AI, to be honest, pushes us a little on the spectrum away from some of the traditional things we’ve talked about, and that was uncomfortable for me,” said Asness, who has a finance doctorate from the University of Chicago. “If there’s limited data, you still need some economic priors. If there’s a lot of data, you sometimes don’t.”
--With assistance from Matthew Miller, Sonali Basak and Katie Greifeld.
©2024 Bloomberg L.P.