BarnBuster
Virtually Unknown Member
uh-oh.
AI Program Beats Top Poker Pros in Ground-Breaking Research
"Artificial intelligence has been a scientific area broadly explored over the past decades. And its application in the game of poker has become the focus of research of two scholars from Carnegie Mellon University. Computer science professor Tuomas Sandholdm and PhD student Noam Brown may have developed an AI program that is able to beat some of poker’s brightest minds in No-Limit Hold’em heads-up matches.
Libratus, the name the program is known by, triumphed over four seasoned poker professionals – Jimmy Chou, Dong Kim, Jason Les, and Daniel McAulay – in a 20-day No-Limit Hold’em heads-up playing session. It had accumulated $1.7 million worth of chips by the end of the tournament.
The program had the difficult task to sort through imperfect information, while trying to detect when and whether its opponent was bluffing. Libratus was designed to deploy a rather different approach from the ones applied by other AI products in the past when learning poker.
Each night, the program analyzed its own strategy from throughout the day. Using a complex meta-algorithm, it explored the holes its opponents had spotted in its behavior at the poker table, prioritized them, and patched the three most obvious ones.
Libratus was thus able to self-improve with each day passing. Its creators explained that previous algorithms had been developed to detect holes in opponents’ playing style. In other words, Libratus’ ability to identify the presence and essence of holes in its own strategy is what makes it a ground-breaking project.
The program’s self-learning has also allowed it to build and beat skills that are mainly based on intuition and psychology in people’s case. Bluffing is one such skill. It is essential for the game of poker and it really takes time and persistence to master. However, through trial and error, an old-time yet perfectly effective approach, Libratus has trained itself to mimic decisions based on intuition and the psychology of its opponent.
The program’s creators explained that now and for now it is most proficient in heads-up No-Limit Hold’em games as more players at the table mean more possible outcomes and this makes it more difficult for the computer to go for the best possible moves.
Although trained to play Texas Hold’em, Libratus can be applied in other areas as well, including business negotiations, medical research, cybersecurity, military strategy, and many more that involve incomplete information."
https://www.cmu.edu/news/stories/archives/2017/january/AI-beats-poker-pros.html
http://www.casinonewsdaily.com/2017/02/01/ai-program-beats-top-poker-pros-ground-breaking-research/
http://www.pokerlistings.com/libratus-poker-ai-smokes-humans-for-1-76m-is-this-the-end-42839
https://www.wired.com/2017/02/libratus/
AI Program Beats Top Poker Pros in Ground-Breaking Research
"Artificial intelligence has been a scientific area broadly explored over the past decades. And its application in the game of poker has become the focus of research of two scholars from Carnegie Mellon University. Computer science professor Tuomas Sandholdm and PhD student Noam Brown may have developed an AI program that is able to beat some of poker’s brightest minds in No-Limit Hold’em heads-up matches.
Libratus, the name the program is known by, triumphed over four seasoned poker professionals – Jimmy Chou, Dong Kim, Jason Les, and Daniel McAulay – in a 20-day No-Limit Hold’em heads-up playing session. It had accumulated $1.7 million worth of chips by the end of the tournament.
The program had the difficult task to sort through imperfect information, while trying to detect when and whether its opponent was bluffing. Libratus was designed to deploy a rather different approach from the ones applied by other AI products in the past when learning poker.
Each night, the program analyzed its own strategy from throughout the day. Using a complex meta-algorithm, it explored the holes its opponents had spotted in its behavior at the poker table, prioritized them, and patched the three most obvious ones.
Libratus was thus able to self-improve with each day passing. Its creators explained that previous algorithms had been developed to detect holes in opponents’ playing style. In other words, Libratus’ ability to identify the presence and essence of holes in its own strategy is what makes it a ground-breaking project.
The program’s self-learning has also allowed it to build and beat skills that are mainly based on intuition and psychology in people’s case. Bluffing is one such skill. It is essential for the game of poker and it really takes time and persistence to master. However, through trial and error, an old-time yet perfectly effective approach, Libratus has trained itself to mimic decisions based on intuition and the psychology of its opponent.
The program’s creators explained that now and for now it is most proficient in heads-up No-Limit Hold’em games as more players at the table mean more possible outcomes and this makes it more difficult for the computer to go for the best possible moves.
Although trained to play Texas Hold’em, Libratus can be applied in other areas as well, including business negotiations, medical research, cybersecurity, military strategy, and many more that involve incomplete information."
https://www.cmu.edu/news/stories/archives/2017/january/AI-beats-poker-pros.html
http://www.casinonewsdaily.com/2017/02/01/ai-program-beats-top-poker-pros-ground-breaking-research/
http://www.pokerlistings.com/libratus-poker-ai-smokes-humans-for-1-76m-is-this-the-end-42839
https://www.wired.com/2017/02/libratus/