A team of Facebook researchers has developed an algorithm using Artificial Intelligence (AI) that can play Poker heads-up Texas Hold’em no-limit better than humans.
ReBeL is a system that uses regular knowledge in Poker to make better decisions, in less time, within a real game environment
This project is called Recursive Belief-based Learning (ReBeL) and its performance is optimal using less knowledge compared to other existing poker AI systems. For example, other algorithms use a combination of reinforcement learning and a series of behavioral patterns. However, in the investigation carried out by Facebook, it is established that this mixture of technologies minimizes the performance in games such as Poker where it makes assumptions that are not in tune with the real world, because in a real game the probability of execution of the plays and the strategy applied by the participant.
Facebook seeks with ReBeL to improve this by allowing the algorithm to work in a more real poker game, using common knowledge that it acquires from other participants, where the search for results is carried out with greater simplicity and flexibility, which expands the possibilities of defeating the best poker player of the world.
What ReBeL does is create a mirror-like game identical to the original game. With PBS it generates different possibilities of acting in a poker game and tries to anticipate possible answers and see how the game would evolve with each option. All this reinforcement learning adds it to your knowledge store, creating a network of value.
At the moment the Facebook ReBel research has been published without the source code, to prevent it from being used on other platforms and to encourage cheating and misuse of it against online players.
No Copyright infringement intended. Video/Photo Unknown Direct for credits/issue or text us +57 3606412 ... | Respect to Photographers & Influencers
