artificial-intelligenceAdvanced tech like AI has already made significant developments towards mainstream integration, with the gaming industry in particular embracing innovation with open arms. The poker world underwent a radical shift last summer, when an Artificial Intelligence program managed the unthinkable and triumphed over a team of professional human players in a tournament format. 

One year on from that landmark poker win, what effect has AI had on the poker industry, and what has it taught us about the applications of Artificial Intelligence and other advanced technologies in the gaming industry as a whole?

Where no AI has gone before…

For as long as the technology has existed, developers have pitted AI applications against humans in popular games like Chess and Go, and even computer games like Dota. However, poker remained a challenge for even the most advanced programs like DeepStack and Liberatus due to the game’s psychological contexts. Yes, like chess, poker is a game of strategy, but success in the game relies heavily on intuition, effective decision making based on hidden information, and psychological mastery – qualities that artificial intelligence simply doesn’t possess. 

Poker remained a very human game, until July 2019, when a program called Pluribus was pitted against five professional poker players in a tournament format. 

The showdown between Pluribus – built by scientists at Carnegie Mellon University in conjunction with Facebook AI – and the poker pros was contested following the rules of Texas Hold ’em – easily the world’s most popular variant. It was also the first time that AI competed against multiple players in a game, and couldn’t just rely on game strategy to win. 

In comparison with other programs, Pluribus was made “on the cheap”, costing $150 and built in just eight days on a 64-core server, but it was able to do what DeepStack and Liberatus couldn’t thanks to reinforcement learning. 

Lasting effects on the poker industry

Prior to the tournament, Pluribus played over a billion hands of poker against itself before competing against a single player. Over a relatively short period of time, the bot was improved to be able to evaluate the success of its previous plays and determine the efficacy of different moves, which it would then apply to future play. By the end of the tournament, Pluribus was able to do what humans do, which was determine its own style of play and adopt different strategies in order to beat the other players at the table. 

Now that AI has been proven to beat pro players in tournaments and single player games, it shouldn’t come as a surprise that players are turning to the technology to improve their own strategy. This is changing the way that people interact with the game as a whole. 

Previously, players would typically learn the game and hone their strategies by playing – and losing – against a wide range of poker players in live events. Then the online era took hold, and players had much more access to gaming resources, like “play money” tournaments that they could compete in for free, as well as a wealth of virtual educational materials. Here in the post-Pluribus era, contemporary players are now developing their strategies by competing against their machines. There’s even discussion in the industry that one day major events like the World Series of Poker could be open to AI programs.  

AI and home gaming 

The trend for fully utlising the application of AI as a developmental tool – machine learning – is also being echoed in the wider gaming industry. The video games we know today are a prime example of integrated machine learning, particularly as both hardware and software have become more sophisticated in the 21st century, but now researchers are developing ways to use AI to create brand new gaming experiences. 

A group of researchers at the University of Alberta, headed up by Assistant Professor Matthew Guzdial, are in the process of developing an AI program that would enable ordinary gamers to build brand new video games, in much the same way that other creatives like musicians are able to work with AI to produce new material. 

The program is still in its early stages, and a consumer-friendly version likely won’t be ready for a number of years, however Guzdial’s vision is that it will “help people make games” by minimising many of the barriers to entry in video game development.