Back in 2017, we highlighted the first major instance of an artificial intelligence program besting human experts in poker. Specifically, a program designed at Carnegie Mellon (dubbed “Libratus”) was deployed at Pittsburgh’s Rivers Casino, where it took on four professional poker players in heads-up poker over a 20-day span. After 120,000 hands, Libratus had accumulated $1,766,250 more than all four pros combined –– indicating a genuine advantage over the pros. As we recounted at the time, computer science professor Tuomas Sandholm concluded that the AI’s “ability to do strategic reasoning with imperfect information” had surpassed that of “the best humans.”

As it would turn out however, Libratus’s victory was only a milestone en route to an even more impressive achievement. Two years later, a follow-up program –– this one named Pluribus –– would accomplish even greater feats in poker. As explained in an article on Pluribus at ScienceDaily.com, this second program defeated six poker professionals decisively over the course of 5,000 hands. In a separate experiment, it also took on 13 poker pros who each boasted $1 million or more of earnings in the game –– playing five at a time and coming out on top after 10,000 hands.

In some respects, the AI programs and the experiments they took part in were set up similarly. Both programs were designed not just to understand and play poker, but specifically to engage in no-limit Texas Hold’em. As many readers will be well aware, Hold’em more broadly is by far the most popular style of poker –– but the “no-limit” distinction matters. Poker.org defines No-Limit Texas Hold’em as a variety of the game in which “players can wager up to the size of their stack as they’d like at any point in the hand.” This is as opposed to other types of Hold’em in which various restrictions are placed on how much can be bet at any one time; it’s why we’re aware of the concept of “going all-in,” as seen in films, televised poker competitions, and so on. And from a standpoint of AI poker performance, it makes the victories of Libratus and Pluribus all the more impressive. Fewer restrictions make for more possibilities and more variation; no-limit forces the programs to intuit the best plays consider pressuring opponents –– as opposed to simply making what they understand to be mathematically advisable decisions.

Where the Pluribus achievement sets itself apart, however, is in the number of players involved –– not just in total, but at a given time. As noted, Libratus defeated poker pros in a heads-up no-limit Hold’em construct. In poker, that essentially means playing one-on-one. The program outsmarted four poker pros consistently, but did so by taking them on one at a time. This makes for a somewhat simplified game, whereas Pluribus faced the added challenge of handling several opponents at once. And in multiple test scenarios, the newer AI proved able to outperform all of its opponents at the same time –– processing a great deal more information, and weighing strategies simultaneously against each competitor. It was not only a more significant challenge, but also a more accurate test of whether an AI could beat top human players at a standard no-limit Hold’em table or tournament.

What this actually means in the long run remains to be seen. In computer science and artificial intelligence research, there is something to be said for the idea of achievement for achievement’s sake; scientists wondered if an AI could master poker more effectively than a human professional, and proved that in fact it could. On the other hand though, Pluribus’s achievement could also pave the way for further innovation ahead. In a write-up on Pluribus at DigitalTrends.com, it was noted that co-creator Noam Brown believes that the bot’s “ability to handle multiple players, hidden information, and numerous possible outcomes” could lead to real-world applications “for the benefit of mankind.”

Only time will tell what those applications might be. But we could well look back one day to see that the victories of Libratus and Pluribus in poker paved the way for significant AI uses.