AQR’s Asness Says Machine Learning Won’t Replace Stock-PickersBy and
‘They’ll be finding patterns that don’t repeat,’ manager says
Buffett’s classic strategies still work, conference crowd told
Cliff Asness, founder of $172 billion AQR Capital Management, said he’s skeptical that new techniques in machine learning will supplant humans in fundamental investment decision-making.
Speaking at the Sohn Canada conference presented by Capitalize for Kids, Asness said machine learning is likely to dominate among firms that trade at high frequencies, using “a ton of data and a repeated experiment again and again,” where the process can help increase the accuracy of results.
“Fundamental stock-picking is simply not about that,” he said Thursday at the Toronto event.
“I don’t think those patterns -- to the extent they find them, and I don’t think they’ll find anything because I think they’ll be finding patterns that don’t repeat,” Asness said. “I don’t think it will take over the need to have individuals look at companies.”
Traditional money managers from Steve Cohen to Paul Tudor Jones have made investments in data science and machine learning. According to a study of 100 hedge fund managers that was conducted by KPMG, 58 percent say they think artificial intelligence will have a medium-to-high impact on their sector.
Asness, 50, said AQR, which employs systematic trading techniques, isn’t using machine learning but instead is making more simple investment predictions.
“I still believe a lot of these classic strategies -- the kind of models Warren Buffett’s using -- are going to continue to work on average, are not particularly correlated to markets,” he said. “My best idea is to find a lot of them and do a little bit of all of them.”
In a wide-ranging discussion, the AQR founder also reiterated his belief that the traditional hedge-fund fee model -- 2 percent of assets managed and a 20 percent cut of profits -- won’t last. While a rare manager may deserve such a high cut, “for the whole industry I don’t think the math will work out for 2 and 20.”
Asness, who manages hedge funds as well as other vehicles, has been critical of managers who are charging hefty fees even while they underperform broader indexes. In an Oct. 21 post on Greenwich, Connecticut-based AQR’s website, he wrote that “hedge fund managers just have to stop blaming others for their problems.” He said their attacks on quants, indexing, algorithmic and high-frequency trading, exchange-traded funds and risk parity are devoid of logic. He’s previously encouraged managers to lower fees.