Turns out you can’t bluff your way past an AI.
In a historic win, a team of scientists from the University of Alberta’s Computer Poker Research Group — which includes researchers from Charles University in Prague and the Czech Technical University — has created an AI that defeated professional poker players last December.
The researchers detailed their creation — a GPU-trained AI system called DeepStack — in a recent issue of Science magazine.
The victory in heads-up no-limit Texas hold’em is the latest in a growing string of victories over human competitors.
This one, however, is particularly significant. Unlike games of “perfect information” — such as checkers, chess and go, where both players are seeing the same pieces — poker players have to use intuition to guess what cards the other players are holding.
“There’s in fact a piece of information about the game — the other players cards — that I don’t know, and that’s what the games about,” said Michael Bowling, a professor in the University of Alberta’s Faculty of Science, in a conversation with tech journalist Michael Copeland in the latest episode of our AI Podcast. “The excitement, and the AI challenge, comes from that.”
That work promises to yield applications in the real world, where we often have to make decisions based on incomplete information. To hear the full story, tune into the latest edition of the AI Podcast.
AI Podcast: Deep Learning Hears Once Extinct Bird
And if you missed our podcast last week, it’s worth a listen: We spoke with Matthew McKown, CEO of Conservation Metrics, about how deep learning techniques helped rediscover a bird that was once thought extinct, and how GPU-powered AI now helps biologists crunch vast quantities of data to spot trends that would have been impossible to detect before.
How to Tune in to the AI Podcast
Our AI Podcast is available through iTunes, Google Play Music, DoggCatcher, Overcast, Podbay, Pocket Casts, PodCruncher, PodKicker, Stitcher and Soundcloud. If your favorite isn’t listed here, drop us a note.
Image credits: Ross Elliot (featured) via Flickr, John Ulan for UAlberta.