National Post (National Edition)
AI hedge fund gives up after just two years
NEW YORK • The fledgling world of AI hedge funds is claiming one of its first casualties.
Sentient Investment Management is notifying investors of plans to liquidate the hedge fund it started in late 2016, according to people with knowledge of the situation.
The fund managed less than US$100 million and hasn’t made money this year after gaining 4 per cent in 2017, one of the people said.
Sentient, based in San Francisco, used artificial intelligence techniques, including machine learning and so-called evolutionary algorithms, to trade stocks globally. It followed a market-neutral strategy in which bets on rising prices were matched by wagers on falling ones.
Jeff Holman, Sentient’s chief investment officer, declined to comment.
Hedge funds have been exploring AI, upgrading trading technologies and employing data scientists as the industry confronts years of mediocre returns. The investments appeared to be paying off — until now. Before this year, the Eurekahedge AI Hedge Fund Index gained an average of 10.5 per cent annually since its 2011 inception. This year, the measure of 15 funds is little changed.
Other funds that say they use AI for trading include Cerebellum Capital, Acatis Investment GmbH and Man Group Plc’s AHL unit.
Sentient Investment Management grew out of Sentient Technologies Inc., an AI startup formed by Babak Hodjat and Antoine Blondeau. Before starting the hedge fund, they spent almost a decade developing an AI system capable of scouring billions of pieces of data, spotting trends, learning and trading stocks.
The firm’s team of technology-industry veterans were betting that software responsible for teaching computers to drive cars, beat the world’s best poker players and translate languages would give their fund an edge on Wall Street pros.
Sentient’s hedge fund deployed thousands of computers, using algorithms to essentially create trillions of virtual traders, which it called “genes.” They were given hypothetical sums to trade in simulated situations created from historical data. The unsuccessful genes were discarded, while those that made money were spliced together with others.
Technology giants including Google kicked off the AI boom several years ago by showing how a relatively new approach called deep learning could dramatically improve some software programs and services.
Since then, hundreds of startups have sprouted.
There’s growing concern, however, among some AI researchers that the technology may not be reliable enough to tackle real-world challenges, like autonomous driving or making investment decisions.
Filip Piekniewski, an expert in the AI field, recently predicted the advent of an “AI Winter,” a period of disillusionment and evaporating funding for such research. Breakthroughs appear to be slowing and those that do occur require everlarger amounts of data and computer power, he said.
“Just because you have special tech and AI doesn’t mean you’re off to the races,” said Vasant Dhar, a professor at New York University. “This tech is interesting but fraught with all kinds of risks.”