In the battle of man versus machine – or quant versus computer – the machines are winning, according to speakers at the Risk Hedge conference on Tuesday (September 13).
The proliferation of data, open-source computer software, machine learning and cheaper computers are threatening to put quants and other types of analysts out of their jobs, said hedge fund managers at the London event.
"Not that long ago if you could do Black-Scholes you could make a lot of money on Wall Street, but now there's an app [to do that] on your phone," Guy Wolf, chief investment officer at Nanolytics Capital Advisors, an investment firm launched by the commodities broker Marex Spectron, told delegates.
Conference delegates heard how Amazon cloud service costs have fallen 84% since 2010. Processing costs have plummeted from $100 for 1 billion calculations in 1980 to less than a millionth of a dollar, while computer error rates in visual recognition tests have decreased to below 5%, meaning computers are better than humans at recognising certain visual images.
Much of what quantitative analysts do already is simply reorganising data so it can be analysed by computer, Wolf said. "I think quant jobs are going to get harder to come by. Open-source development will reduce the importance of technical skill, and increase the importance of creativity and imagination."
"A lot of analysts aren't much more than data manipulators. They don't derive insights, they just perform mechanical tasks. And I think that's true not just in investment management but in risk management as well," he said.
Also speaking at the event, Jeffrey Tarrant, CEO of fund of funds Protégé Partners based in New York, said he had resumed investing in small quant hedge funds after a 15-year lull. Cheaper computers and machine learning mean just one or two quants can now compete with the hundreds employed by firms such as Renaissance Technologies, he said. That was impossible two or three years ago, he told delegates.
Wolf said: "The premium will be on people who can design technology processes rather than the mechanics who can implement them.
"One of the problems is a lot of quantitative people have a very technical background. They don't really understand what the data is. They only understand how to manipulate it."
Meanwhile, industries other than finance lead the understanding of machine learning that promises to revolutionise the hedge fund industry, said Tarrant. Google currently has more than 2,000 machine-learning projects. Its DeepMind machine-learning algorithms have beaten the reigning master at the complex board game Go and have been used to reduce the energy usage of its servers by 40%.
For quants with the imagination and creativity to design processes rather than execute them, Wolf predicted a bright future. Markets are still driven by the same underlying psychological forces, he said. "Now people's actions and emotions are captured within the data points. They're footprints in the snow."
Likewise, Tarrant said humans working with machine-learning computers had proven more capable in tests than machine-learning computers alone.
Quoting chess grandmaster Gary Kasparov, who once said a computer could never beat him but lost to IBM's Deep Blue in 1997, Tarrant concluded that weak humans plus machines plus better processes were superior to strong computers alone. Almost 80% of the top chess performances in 2016 were by human and machine pairs, he said.