- Martin Schmid, Rudolf Kadlec and Matej Moravcik left DeepMind in January to set up EquiLibre Technologies, relocating to Prague in the Czech Republic.
- The trio used to work at IBM and developed an AI called DeepStack that became the first AI capable of beating professional poker players at heads-up no-limit Texas hold'em poker.
- "Our idea is that rather than playing poker, our algorithms will play algorithmic trading," Schmid told CNBC.
Three former DeepMind employees are trying to train a machine to spot and invest in company stocks and cryptocurrencies before they rise.
Martin Schmid, Rudolf Kadlec and Matej Moravcik left Alphabet-owned DeepMind in January to set up EquiLibre Technologies, relocating from Edmonton in Canada to Prague in the Czech Republic in the process.
The trio all used to work at IBM and in 2017 they developed an AI called DeepStack. It became the first AI capable of beating professional poker players at heads-up no-limit Texas hold'em poker.
Now they're looking to apply some of these concepts to financial markets.
"Our idea is that rather than playing poker, our algorithms will play algorithmic trading," Schmid told CNBC. "We are also looking into crypto."
They intend to use a technique known as reinforcement learning to train an AI system to buy and sell shares and make a profit. Reinforcement learning involves training an AI to achieve a particular goal (be that winning a game of chess or spotting a tumor on a mammogram) by giving it a reward each time.
Schmid said he's not concerned about regulators clamping down on the technology as other companies are already doing similar things. Indeed, EquiLibre Technologies will be competing with the likes of AI algorithmic stock picking products Candlestick and Yuyostox.
"Most of the trading out there is already algorithmic," Schmid said. "We just want to do better algorithms than the ones that are already out there."
In the long run, EquiLibre Technologies hopes to either use the AI it develops to underpin a new hedge fund or sell it to a large institutional bank or another investor.
EquiLibre Technologies' advisory board includes two senior DeepMind staff that are well known in the field of AI.
One is the head of DeepMind's Edmonton office, Michael Bowling, and the other is Richard Sutton, who co-authored DeepMind's controversial "Reward is enough" paper last year. In the paper, the researchers claim that if you keep "rewarding" an algorithm each time it does something you want it to then it will eventually start to show signs of general intelligence.
A number of venture capitalists have already backed EquiLibre Technologies. Schmid said it has raised the largest-ever seed round in the Czech Republic, but refused to disclose the exact figure.
"My understanding is there's always more money than start-ups," he said. "VCs are having a hard time finding the good startups."
Schmid and his co-founders are among a growing number of ex-DeepMind entrepreneurs who are raising money from venture capitalists.
"If you worked at Google, DeepMind and other places, you probably do not suck," Schmid said.
He added that DeepMind staff are also likely to have a good network of tech contacts they could potentially recruit.