A very interesting contest recently took place at the Rivers Casino in Pittsburgh where four of the World’s best poker players played against a machine. The machine, named Libratus was using the Artificial Intelligence (AI) technology developed at Carnegie Mellon University (CMU). The freelance gigs are happy to see new doors opening for using their skills.
Artificial intelligence (AI) is the intelligence exhibited by machines in general or more specifically, the computers. In computer science, the field of AI research defines itself as the study of any device that perceives its environment and takes actions that maximise its chance of success at some goal. The term AI is used when a machine mimics functions that humans associate with other human minds, such as ‘learning’ and ‘problem-solving.’ As machines become more capable, the capability becomes a routine job for the machine. The example is an optical character recognition, which is no longer an example of AI.
History of AI and Poker:
In a similar poker tournament last year, the machine, an earlier iteration of the AI lost, but just barely, to the humans. Team CMU is hoping for a different outcome this time, and as we know, their hope wasn’t baseless! This contest was a part of the university brain trust’s attempt to show that AI, which has already prevailed against humans in some other games can do the same in this game also.
Why is it difficult for a machine to play Poker?
In Poker, a single-player never gets a full view of the total game. For one thing, players have no idea who holds the outstanding cards, and it makes it difficult for the machine to understand what moves the other players will do. AI was used for a less difficult version of Poker earlier, but this time the task was harder.
For Poker, it is difficult because if you never bluff or if you always bluff, you are not a considered as the best player. Game theory tells you how to randomise your play in a way that is, in a sense, optimal. During last year, Sandholm led the development of a poker-playing program, called Claudico, which was beaten in a match against several professional poker players.
Why is the Machine Win important?
A win for Libratus is a great achievement in AI field. The Poker game requires reasoning and intelligence that has proven difficult for any machines to imitate so far. It is fundamentally different from checkers or chess as the opponent’s hand remains hidden from view during the play. The no-limit Texas hold’em was especially challenging because an opponent could essentially bet any amount. The latest version included new equilibrium approximation technique as well as several new methods for analysing possible outcomes as cards are revealed at later stages of a game.
Creation of this machine and use of AI
Libratus was created by Tuomas Sandholm, a professor at CMU, and his graduate student Noam Brown. Sandholm, an AI, and game theory expert says it is amazing that humans have been able to outplay computers for so long.
The advances in machine learning and AI have already seen some superhuman game playing programs Last year, researchers at DeepMind, a subsidiary of Alphabet, developed a program capable of beating one of the world’s best Go players. This achievement was spectacular because ‘Go’ is an extremely complex game. A few different research groups are focussing on tackling poker as well from different parts of the World, and this tournament has given a boost to all of them. You will also find freelancers working on AI today.
The techniques used to build a smarter bot could have many applications in other fields also. Game theory is already applied to research on cyber security, automated guidance for taxi service, and robot planning and it should open more doors of opportunity with the recent win of Computer against the humans. However, Libratus winning this game doesn’t mean that humans no longer deserve a spot at the card table. The multiplayer version of no-limit Texas holds’em cannot be mastered using the techniques employed by Libratus.
Libratus relies on three components. The first is known as reinforcement learning, an extreme form of trial and error. The Libratus developed its technique of playing wider ranges of bets than its human opponents by playing multiple games against itself. A second system called an end-game solver, allowed Libratus to learn from games as it was playing.
These two systems together were sufficient to defeat humans, but designers Noam Brown and Tuomas Sandholm of Carnegie Mellon University (CMU) in Pittsburgh added a third component to prevent Libratus’s opponents from exploiting patterns in the machine’s play. An extra program identified patterns, and the patterns got removed overnight! The human win was made more difficult by this strategy.
Why was the competition very interesting?
The human representatives were Jason Les, Daniel McAulay, Dong Kyu Kim and Jimmy Chou. They all tried their best but somewhere an incorrect strategy wiped out their gains and forced them to chase the AI for the remaining weeks. At the end of 20 days competition and after 120,000 hands, Libratus claimed victory with a daily total of $206,061 in theoretical chips and an overall pile of $1,766,250.
The Libratus creator Tuomas Sandholm in an email mentioned that it was a landmark step for the AI. More generally, it has shown that even if the machine does not have proper information, the AI ability has surpassed that of the best humans in playing a complex game. The event was a great excitement for students and researchers in AI.
The Libratus win was not a total surprise as from day-1 the Libratus was ahead of the human competitors, but still, it remained an interesting game to see how the machine is developing its thinking process against the humans using all their knowledge and intelligence to beat the machine. This win will certainly open innovation doors for everyone working in the AI space including those performing these tasks as freelance work.