Can A Card-Playing Artificial Intelligence Beat A Bunch Of Poker Pros?

Faster hardware and better software means artificial intelligence is only going to get more, er, intelligent. Games are one way to put electronics smarts through their paces and when researchers got bored of Chess they, naturally, moved onto Texas Hold'em. So, how well does a dedicated, poker-playing cyber-brain go against the professionals? Not well.

"Brains vs Artificial Intelligence" is a little competition being held by Carnegie Mellon University's School of Computer Science and it pits AI "Claudico" against four pro poker players — Dong Kim, Jason Les, Bjorn Li and Doug Polk. Seeing as a computer probably doesn't have much of a "tell", other than its requisite processing delay, you'd think the AI would be favoured.

At the time of writing, Claudico, after over 42,000 hands, is behind the humans by a collective 600,000 chips, according to Steve Crowe over at Robotics Trends. While the competition hasn't completely dominated the organics, they're statistically a class above.

Not that high-level poker is easy for a machine to process. Like Chess, there are many different "moves" the AI can take. The difference is that you don't have all the information in poker — the hands of other players and the cards in the deck are all unknowns, not to mention any attempts at analysing bluffs.

A human brain is more capable in such an environment, but the team behind Claudio have built adaptation into its design... and it's performance has improved as the competition has gone on. From the Robotics Trends piece:

This is quite possibly the highest-quality poker ever seen," Sandholm continued. "Furthermore, both Claudico and the humans are improving their game throughout the event. Both sides are already playing stronger than they were at the start. I have been extremely impressed by the pros' ability to improve and adjust their game no matter what curve balls Claudico has thrown at them."

If you're keen, you can watch the matches live via the University website.

[Carnegie University, via Robotics Trends]

Image: Ross Elliott / Flickr, licensed under Creative Commons 2.0

Trending Stories Right Now