Former DeepMinders Develop AI To Pick Stocks

Former DeepMinders Develop AI To Pick Stocks

A trio of Alphabet-owned DeepMind employees left the company in January in order to start their own company and develop AI capable of picking stocks and cryptocurrencies to invest in before their prices rise, reports CNBC. The three employees—Martin Schmid, Rudolf Kadlec, and Matej Moravcik—formerly worked at IBM. In 2017, they created DeepStack, an AI that could win against professional poker players at heads-up no-limit Texas hold ‘em poker. That concept will be applied to the financial markets at their new company, EquiLibre Technologies. “…rather than playing poker, our algorithms will play algorithmic trading,” Schmid told CNBC.

Using reinforcement learning—a technique that rewards the AI every time it achieves a particular goal—the AI system will be trained to buy and sell shares in order to make a profit. Since so much trading in the industry is already algorithmic, Schmid says he isn’t worried about policy makers instituting regulations on the technology. The article points to Candlestick and Yuyostox, two other stock-picking AI products being utilized in the market. In the long term, Equilibre has set its sights on using its AI to establish a new hedge fund or selling it to a major institutional bank.

The three former DeepMind employees aren’t Equilibre’s only connection to DeepMind: both Michael Bowling, head of DeepMind’s Edmonton office and Richard Sutton, co-author of DeepMind’s “Reward Is Enough” paper are on its advisory board. Sutton’s paper put forth the controversial idea that by constantly rewarding an algorithm every time it does what you want it to do, the AI will eventually develop signs of intelligence.

The new start-up, which is based in Prague, Czech Republic, has been backed by several venture capitalists—a testament to the pedigree that working at DeepMind gives Equilibre. “If you worked at Google, DeepMind and other places,” Schmid told CNBC, “you probably do not suck.”