Libratus, a name for an artificial intelligence bot that made headlines a month back after defeating a team of accomplished poker professionals, is proof that computer programs are now intelligent enough to beat even the best of humans in the tricky card game of no-limit Texas hold'em poker.
The highly publicized AI defeated its opponents in 20 days. However, as luck would have it for the Carnegie Mellon University team, another AI bot in the name of DeepStack, managed to accomplish the same feat before them - back in December 2016.
The Alberta team, led by Michael Bowling, has already made headlines in 2015 with his Limit Hold'em bot "Cepheus" as it was able to solve the simple hold'em format. In 2016, they claimed the No-Limit Hold'em game with their DeepStack AI, as the bot proved itself against 11 top NLHE players with over 3,000 hands.
Overall, the University of Alberta recruited a number of poker players from the International Federation of Poker and were each asked to play 3,000 hands against their AI program. A total of 11 players completed the challenge and DeepStack beat 10 of them statistically, and by a significant margin. The 11 player though was able to achieve a "statistical tie."
The feat of DeepStack was only briefly mentioned in the press as the Alberta team chose to not publicly share their findings until they had been "peer reviewed."
As of last week, the achievement made by DeepStack was published in Science, a prestigious and highly respected peer-reviewed scientific journal.
There are key factors to keep in mind though when comparing each AI bot. Libratus played 30,000 hands with each player which meant statistical results are more accurate. Another is that DeepStack played with good players, whereas Libratus played against four of the best in the game.
Source: https://www.cardschat.com/news/czech-deepstack-v-cmus-libratus-fight-for-top-ai-dog-status-40483
