Over the past year or so, a number of AI hedge funds have sprung up, promising to beat human traders by analyzing and responding to the market in a faster and deeper way. These AI funds go beyond the algorithms used by traditional data-driven funds by attempting to imitate—and improve on—the way the human brain works.
Some early results have been impressive. In Japan, the Simplex Equity Futures Strategy Fund managed to end at a 3.4 percent gain on June 24, the day of Brexit, when funds across the country plummeted. A survey of 12 AI funds from across the globe showed they gained, on average, almost 7 percent this year. At a recent conference for AI academics, half the companies recruiting employees were financial firms, a marked contrast from previous years.
Shaunak Khire, who launched Emma AI, a Silicon Valley-based AI hedge fund this summer, says AIs like his have major advantages over the traditional hedge fund.
“There’s no way a human analyst can cover that much data,” Khire says. “It’s literally not possible for a human brain to process that much information in that short a period of time.”
An AI fund might take into account financial data from markets across the world, historical data, news articles, international monetary policy, knowledge about human trading behavior, company backgrounds and more.
Emma AI, for example, has been trading on GlaxoSmithKline, covering every data point there is on the pharmaceutical company, even looking at filings going back to the 1970s, Khire says.
Many funds already use quantitative investing strategies, developing computer algorithms to make predictions. The difference between these funds and the AI variety is so-called “deep learning” or artificial neural networks—an AI can learn without human input, whereas algorithms cannot.
Emma AI, like many AI funds, uses Bayesian analysis to replicate the human decision-making process. This means it is able to take advantage of new information to update its perspectives and strategies. This is essentially what humans do, only an AI can do it more quickly and, in theory, more rationally. A human analyst is prone to making mistakes based on fear or overexcitement or greed, problems not faced by computers.
A fund launched earlier this year by San Francisco-based AI firm Sentient Technologies uses a type of AI inspired by plant and animal evolution. This “evolutionary computation” constantly creates new algorithms and incorporates the best ones into its old algorithms, making itself ever better and stronger. Hong Kong-based AI fund Aidyia uses evolutionary computation, among other strategies.
In the future, Khire sees entire funds managed by AI, though perhaps with some sort of human control or override for regulatory purposes. He also, however, sees AI as playing a role in regulation. Wall Street reforms passed by Congress in recent years have caused many banks to increase the number of employees dedicated to compliance.
“You could automate all that,” Khire says.
Some experts are skeptical about whether AI will really revolutionize the financial industry the way AI firm founders claim. Some say the AI technologies used are not all that different from more traditional algorithms used in data-driven firms. Others say financial markets are too capricious to be predicted by the kind of AI that exists today.
Khire says he’s not trying to reinvent the wheel, only to improve it incrementally.
“The AI’s only goal is to find opportunities that have a lower risk threshold than the S&P  and have a higher rate of return than the S&P,” he says. “It’s a low bar.”
Still, simply beating the S&P 500, the American stock market index often used as a monitor of Wall Street's overall performance, over the long term is not easy. But Khire and others are convinced they’ve got the goods. The question is: would you bet your money on it?