Artificial Intelligence And Poker – The AI is coming
Artificial Intelligence And Poker
Poker is a skill-based card game which requires strategy, reasoning, and intelligence to win. It has been discussed at length whether AI’s can beat humans at poker because of this. Human beings have the benefit of having these features naturally while Artificial Intelligence can find them hard to imitate. But with technology advancing as much as it is every day, could Artificial Intelligence end up beating humans in a good old poker game? Thus perhaps even become so prevalent in our future?
There are numerous different poker games available to play. You can expect that professional poker players would be well versed in the majority of them. As its impossible to teach reasoning. Can Artificial Intelligence manage to gain the upper hand over these professional players? They would certainly have to learn every single different poker game, and then they would have to attempt teaching reasoning.
There are many games that Artificial Intelligence have been able to beat humans at. But poker has been proving quite the challenge. There is not one move that they can’t complete. Instead of having to randomise what they do so that their opponents do not know if they are bluffing. Researchers use the mathematics of strategic decision to try to find the best strategy – but this is often approximate.
While just a few years ago the idea that AI’s beating humans at poker was laughable, advances in technology and machine learning have seen a number of programs appear designed to hopefully beat human players of even the most complex poker games. In 2015, two researchers used an AI to challenge some top human poker players and lost. In the past month, they challenged four top players at Texas Hold’Em with an AI called Libratus who ended up winning! No AI had ever beaten the best at such a challenging game, but Libratus managed it. Does this mean we can start to expect plenty of AI’s to beat humans at poker?
Designed by Carnegie Mellon professor Tuomas Sandholm and Noam Brown, a grad student, Libratus (which means ‘balance’ in Latin), played across twenty days and managed to top its four human competitors by more than $1.7 million. Libratus also left all four players with a negative number of chips. The players who lost to Libratus said that the machine happens to do a little bit of everything well. The AI managed to bluff at the right times and fool them; it knew when to change its bets to confuse the other competitors and also knew when it was best to bet low. They also believe that the machine’s strategy changed from day to day over the twenty-day playing period. If they thought they’d worked it out, it would have changed by the next day.
Some could say it was unfair that the AI seemed to change strategy day by day. However, isn’t that what humans do too when it comes to high stake poker games like these? All players know how they play and their own personal strategies and if they thought that Libratus was getting wise to them, they would, of course, change their tactics up. Each player was playing their own one in one game with the AI, and they shared their strategies. Ultimately they wanted to beat the other human players too, but they all also wanted to beat the machine.
But what benefits would an AI who could beat humans at poker bring? We know that Artificial Intelligence can find the reasoning that us humans are born with difficult to imitate. But, if it is possible as shown by Libratus, how could this impact the world for the future? Some people say that a machine that manages to master no-limit Texas Hold ‘Em mimics the kind of human intuition that strategies such as auctions, financial trading, and political negotiations require.
It had been previously impossible for Artificial Intelligence to master the skill game of poker in all of its forms, especially the more difficult poker games such as Texas Hold’Em. However with the knowledge that in 2017 an AI named Libratus managed to not only master the game but beat four of the top players. Topping them by over one million dollars and leaving them with negative chips. Surely it must be envisaged that more and more AI’s can begin to learn the strategies of the game and also start to beat humans at it.
Technology is advancing so much, and with Artificial Intelligence being able to master even the most challenging of skill games, surely this shows that the future holds so much more for AI’s. It was something that was proving ever more challenging for AI researchers, and now that a team of researchers has cracked the game theory. It can surely go towards other situations just like those mentioned before like financial trading or political negotiations. Artificial Intelligence definitely holds a place in society, and they can be in use for many different things. With them solving challenging situations, just what can they be in use for in the years to come?
It has taken quite a few years for researchers to finally program an AI that can master poker against the very best of human players. So, it is inevitable that we will start to see more AI’s starting to master the game. With other difficult situations arising every day in all aspects of life, perhaps we can also expect Artificial Intelligence to be able to solve those problems too. Despite not having the benefit of natural human intuition, could we also perhaps see a time where Artificial Intelligence can solve a problem before humans? Or even solve something that we as humans find impossible? It seems that only time will tell, but the recent exciting advances in AI capabilities should definitely have an applause.