- The Pentagon has hired an AI bot designed to win poker games.
- Libratus, which became well-known after winning $677,000 against a group of poker experts, has been hired for two years at $10 million.
- The bot managed to defeat four experts in "no limit Texas Hold'em".
Around two years ago, an AI bot became a sensation after managing to defeat several expert poker players in no-limit Texas Hold'em, managing to bag $677,000 in play money from the poker champions by calculating how they might respond to its decisions.
Two researchers from Carnegie Mellon University built Libratus using a technology called computational game theory.
For an AI, poker is more difficult to learn than chess, for example. This is due to it not being a "perfect information game," in the sense that while players can see their own cards, they can't see those of their rivals.
With poker being a game of skill, factors like body language and deception come into play, which required the software to come up with betting strategies and even to demonstrate the ability to bluff.
The US army has now hired the robot
One of the developers, Tuomas Sandholm, founded a startup called Strategy Robot to adapt the technology to government needs.
Remarkably, the robot has now been hired by the Pentagon for $10 million.
According to a Wired report, Libratus will now spend the next two years working "in support of" a Pentagon agency called the Defense Innovation Unit.
Some are concerned about the military's interest in AI
With China recruiting children to build bots for the military, the US army is far from the only military group interested in AI; Russia is also exploring the potential military applications of AI, with President Putin saying whoever leads in AI "will become the ruler of the world."
The military's growing curiosity as regards AI has previously been troubling for many of those working on developing the technology, with a number of Google's AI researchers having joined thousands of employees in protest against the company's work on Project Maven, a program set up to exercise commercially available AI techniques on US missions.
Sandholm, however, believes that the concerns about military interest in AI are misplaced, saying: "I think AI's going to make the world a much safer place."