What’s the latest edge in quant finance, which relies on math whizzes to drive computer-driven funds? Humans.
In an ironic twist, the people behind quant funds are using some of the same language human-run stockpickers and investors have used for decades to defend their style against computer-run alternatives.
“Human creativity” is what will get quants ahead, said Amadeo Alentorn, head of systematic equities at Jupiter Asset Management, adding that there might be “too much hype” for what generative AI can do for investment management right now.
While advances from companies like OpenAI have allowed quants to speed up their processes and work on more complex projects, the technology is not yet to the point where firms can turn their systematic funds over completely to the machines.
“By far the most important part of generative AI is the end user,” said Timothee Consigny, CTO of H2O Asset Management, a Paris-based macro manager. Consigny, speaking at a London conference for the industry, called Quant Strats, compared the technology to driving a car; everyone now has access to F1-fast cars thanks to generative AI, but not everyone is able to drive them effectively.
In other words, on its own, AI is not going to help win the “alpha war,” said Matthias Uhl, head of analytics and quant solutions at UBS Asset Management.
None other than Citadel founder Ken Griffin agrees with the assessment, saying at Wednesday’s Robin Hood conference in New York that generative AI “falls short” when it comes to finding market-beating ideas, according to Bloomberg.
A timesaver and marketing tool, but not a revolution
Back-office grunts and marketing teams might be the biggest beneficiaries of the technology so far. Alentorn said the explosion of the technology has “helped selling our funds” because investors are more comfortable with things being run by a computer now. Uhl meanwhile said it’s mostly doing “mundane” tasks.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
This aligns with what funds told the Alternative Investment Management Association last year in a survey. While firms were using AI, the primary use cases were for “time savings on administrative tasks” and “content generation” for investor relations teams.
It’s not to say quants dislike the technology. Stephan Kessler, who heads quantitative investment research for Morgan Stanley, said, “it allows us to run more systematic strategies in areas we haven’t done before.” He said the firm has AI combing bond prospectuses for crucial info, a task that used to take a person days is now done in minutes.
“It allows us to code more complex things faster,” Kessler added.
But large language models are, in a way, a blank page and need to learn from funds what is important and what isn’t. The software itself is not that valuable; as David Shelton, global head of FICC electronic trading and FX quantitative strategies group for Bank of America, pointed out, AI companies are “giving away the code.”
Therefore, what you feed the model is more important than what it was trained on, said Haoxue Wang, quant at Izzy Englander’s Millennium.
“A language model can’t read your mind,” Wang said.
In other words, the edge still lies with people — for now.
