Forget the ‘tell’: poker is just another game for AI.
A collaborative effort between Facebook and Carnegie Mellon University has done for poker what AlphaGo did for the game of Go. They developed an algorithm that beat a raft of top poker players at six-player no-limit Texas hold ‘em poker. There are eerie parallels with the AlphaGo success story. This variant of poker has a number of apparently tricky features for an algorithm: multiple players, massive complexity, the presence of bluffing. Indeed, the algorithm was not even equipped to read any ‘tells’ let slip by the human players. Yet it trounced them.
The maths might have been different but the principles of the algorithm’s success were similar to those of AlphaGo. It did not use brute computing force on its own to secure the answer. Instead, it only looked forward two or three moves. It was trained by playing against itself. Moreover, it saw no difference between a bluff and a call supported by the cards in its hand. Finally, it simply figured out the probabilities with utter consistency and detachment.
Link to article: https://www.nature.com/articles/d41586-019-02156-9
You might also like to browse other AI posts: https://www.thesentientrobot.com/category/ai/
