Joshua House
Staff Writer
In a recent 20-day no-limit Texas Hold'em competition, “Brains vs. Artificial Intelligence: Upping the Ante” an AI by the name Libratus, took down poker pros by a margin of over 1.75 million chips. This took place from Jan. 11 and finished up on Jan. 30 at Rivers Casino in Pittsburgh, PA. Libratus was developed by a team at Carnegie Mellon University including Prof. Tuomas Sandholm and Ph. D. student Noam Brown. Last year, a similar system named Claudico that was developed by the same team lost by over 750,000 chips to four other poker pros.
This might not seem like a big deal, but it is shocking for both the poker community and the programming one. First of all, people have been trying for years to develop computers that could beat the best of the best in given games. A turning point came in May 1997 when the IBM Deep Blue Computer beat Chess Grandmaster Gary Kasparov. The difference between a computer beating someone in chess and beating a player in poker is in the game. Chess is a game that is all strategy and a computer can calculate out all possible moves and outcomes. Poker is a whole different game due to how it involves both skill, luck, and a lot of bluffing. In poker there are situations that players and the computer can’t really prepare for just because of how infrequently they happen. Thus the AI has to develop its own unique way of thinking, which has been a game changer for human players.
Libratus over-bets frequently, wagering far more to win a hand than is currently up for grabs in the pot.
“If you have $200 in the middle and $20,000 in your stack, you can bet that,” says Doug Polk, a poker pro who bested a previous AI built by CMU in 2015. “But humans don’t really like that. It feels like you’re risking a lot of money to win so little. The computer doesn’t have that psychology. It just looks at the best play.”
The way that the computer was able to pull this off came from its ability to learn from its mistakes and its complete disregard for the value of money. Unlike a human, Libratus doesn’t have any plans to use money to buy items it wants; it only wants to win. When Libratus messed up early on, the poker pros found that it rarely ever made the same mistake twice. It kept changing its game and remained unpredictable for even the best to figure out. Also Libratus would do insane overbet bluffs for a small pot of chips that poker pros would be forced to fold. When the poker pros were sleeping in between eight-hour sessions, the computer would still be busy at work learning from itself. What is kind of scary in itself is that the actual creators of this system do not even know how the computer itself plays.
But what are the implications of Libratus to the programming community?
Frank Pfenning Head of the Carnegie Mellon school of Computer Science stated when asked about the big win: “Developing an AI that can do that successfully is a tremendous step forward scientifically and has numerous applications. Imagine that your smartphone will someday be able to negotiate the best price on a new car for you. That's just the beginning.”
In the future these forms of computers could have use in military strategy, business-to-business negotiations, finance, and even in the medical field due to its ability to make split second decisions.
