Humans VS Artificial Intelligence in Poker: who is better?
There have been numerous competitions organized between humans and AI to see who is better at poker. While machines and software are smart enough to beat humans at other board games like GO and chess, poker is a lot different because the game comprises inadequate information, unlike the other mentioned games. With GO and Chess players can view the entire board while in poker you don’t see your opponents game.
This article will bring information on poker games between humans and artificial intelligence and who emerged victorious between the two.
No doubt there are poker pros around the world but can they beat AI in a fair game of poker? This article will answer this question as we go back in time to bring you what went down between humans and AI in the past.
AI (artificial intelligence) is a significant breakthrough in the world of science and they are becoming important in everyday affairs, for example, Siri and Alexa are artificial intelligence assistants famous around the globe for their ability to answer questions, book appointments, shop, set alarms, and numerous other functions. Tesla’s have come up with AI that can self-drive a car which was impossible to do some ten years ago. The question remains can AI play poker? Several poker sites say yes!
Humans VS AI in Poker
Let’s look at instances where Humans and AI played poker and who emerged the winner.
On the 30th of January 2017 at a casino in Pittsburgh, after playing for 20days an AI designed by a duo of Carnegie Mellon researchers won four experienced players at a complex form of a poker game, the no-limit Texas Hold’Em whose tactic involves a knowledge of long-term game theory and betting strategy. While this defeat came as a shock to the players and numerous viewers who watched with keen interest. The AI which was designed by Tuomas Sandholm and Naom Brown is called Libratus which is a Latin word for balance. At the end of the game, Libratus won its four competitors with
over $1.7million leaving them with negative sets of chips. On observation, a University Of Michigan Professor Michael Wellman suggested that the game strategy showcased by Libratus can be beneficial in political negotiations, auctions and even financial trading.
The four human players, McAulay, Jason Les, Jimmy Chou and Dong Kim concluded that the machine changed its game pattern each day which made it difficult for them to find a loophole in its gaming strategy while it studied their gameplay which it used against them every time. Although the researchers kept on reprogramming the machine each day but so was the case with the humans as they were using every tactic at their disposal to increase their chances of success.
Also sometime in 2017 University of Alberta researchers developed an AI known as DeepStack defeated its human opponents in a fair game of poker. Before it was set to play a game DeepStark underwent a series of intensive training which involved deep learning by playing countless randomly created poker simulations against itself allowing its neural network to develop general poker mindset to help it conquer situations it may not have encountered in the past.
After preparing it was time for DeepStack to compete with a human opponent where it played 11 human players in a space of four weeks and a total of 3000 matches. At the end of the period, DeepStack defeated all 11 opponents by a large margin. The neural network of DeepStack allows it to disintegrate each game into smaller bits and can read between the next two to ten step. It can work through hundreds even millions of possible outcomes in less than three seconds and use the information to select the best move to ensure its success.
In 2015 three out of the best four poker players in the world defeated an AI known as Claudico in a game of Heads-Up No-limit Texas Hold-Em after 80,000 hands. The game lasted two weeks. Although the winning wasn’t scientifically significant as the human players won a total of $732,713 over the artificial intelligence competitor. The players Bjorn Li, Doug Polk and Dong Kim left the table with $559,033, $213,671, and $70,491 respectively. Claudico whose development was directed by Tuomas Sandholm who 2 years later helped developed Libratus which defeated its human competitors at a game which lasted 20 days was actually a prototype that was tested to run simulations for a later design which saw the birth of Libratus which shocked the science community and poker world as a whole.
AI is by far different from a human opponent, for one they cannot be overwhelmed by the psychological confusions that human face when faced with a challenging scenario like in a game of poker. There is no doubt that the AI market is on the rise as studies show that by 2022 the poker market will gross over $16billion. From the scenarios outlined above, we can say that AI will be a favorite choice to beat humans at a game of poker unlike chess or Go. Reasons for this conclusion are:
- Poker Artificial Intelligence is a lot different from Chess and Go. In chess and Go players can view the entire board as there is no hidden information. Humans and computers can easily see the game pieces and know what to expect at all time. Whereas in poker you cannot see your opponents game which makes it difficult to read the game.
- Poker is easy to Automate. It is true that poker is a game of strategy or an imperfection game. It involves a high mathematical process which AI can easily automate faster than humans.
- AI is programmed with different forms of learning. AI like Libratus utilize reinforcement learning to tackle Nash equilibrium algorithms. While DeepStack utilizes a neural network designed with deep learning to tackle its problems.
- AI learned by playing against itself and not by playing human opponents. This makes it easy for it to come up with strategies that human cannot employ which give AI an advantage over humans.