On a recent episode of his self-titled podcast, acclaimed computer scientist Lex Fridman sat down for a long discussion with Meta AI research scientist Noam Brown. Poker fans will recognize Brown as the man who made headlines in 2017 when his AI system Libratus defeated four of the world’s best professional poker players.

In the podcast, Fridman and Brown discuss Brown’s career and the process behind creating his game-changing AI. They also talk in-depth about the future of AI, the similarities between poker and other games, and much more. Here are our key takeaways from that chat.



One of the podcast’s most significant talking points is just how much influence AI has over the way people play poker. There was plenty of skepticism around Brown’s system before it was revealed, but its success has changed how players view the game.

AI didn’t learn to play by mimicking players’ behaviour; instead, it analyzed situations and came up with the best possible move to secure success. Some small tactics, such as upping the value of small pots or playing overbets, have since been adopted by top players. As AI advances, it will continue to find ways to secure victory in games.

Pros will no doubt study how these games are played and attempt to implement some of AI’s strategies. Like chess, poker players will also begin to play against AI and learn from them. In addition, they will use AI to identify their weaknesses and practice new strategies.



An exciting part of the conversation comes when Brown explains how AI has learned to master poker. First, he explains the concept of the Nash Equilibrium and how AI would need to find it to be a success. The Nash Equilibrium is the most common way to define the solution of a non-cooperative game involving two or more players. In layman’s terms, it’s a way to win where you don’t need to change your strategy.

AI will play against itself and then assess how the game went. After it has done this, it will look at “regrets” and analyze what would have happened if the game had been played differently. This process, called counterfactual regret minimization, allows the AI to learn how to play and eventually reach a Nash Equilibrium, even in a game of imperfect information like poker.

During a game, AI can search for possible hands and use neural nets to make decisions. Throughout the interview, Brown stresses the importance of these searches, not just in poker but in AI and its continued development.

Of course, AI is a lot more complicated than a few paragraphs, but these are some basic principles discussed in the podcast:



The discussion over whether chess or poker is harder to beat has raged on since time immemorial, but we finally have some sort of answer, at least from an AI perspective. Brown is hesitant to answer when asked in the video, but he then says that poker is a trickier game to beat.

He points out that chess is a zero-sum game where it is easier to find Nash Equilibrium than poker. He explains that the missing information in poker (the cards in your hand) makes it much more complicated to defeat with AI. While it is possible, the solution is much easier to find in chess.



One constant discussion throughout the podcast is the missing human element of AI and what benefits this might bring. The lack of human emotion is one of the reasons for AI’s success in poker. For example, Brown talks about how humans often raise their bets in relation to the size of the pot during a game.

When AI was adjusted and could bet whatever it wanted, it began to raise the pot significantly, often more than its value. From the bot’s perspective, it does the thing that makes it the most money. It isn’t concerned with putting players in an uncomfortable position, even though that was the end result.

Humans, however, are often much more restricted by their emotions. The connection we have with money and the stakes involved often lead us to make more emotional decisions. The two spend a significant chunk of time discussing this and how AI might learn to behave like humans in the future.

The old discussion about playing the cards or the player is redundant when it comes to AI. Still, there are plenty of benefits to AI becoming more human-like. Towards the end of the podcast, Brown discusses how, if they could mix human behaviour with AI, they could create bots that would allow people to maximize their skillset and prepare for players.

Using chess as an example, they discuss the possibility of programming a bot to play like a player with a particular Elo rating or programming one to play in the style of an exact player. The benefits of this, especially for top players, would be absolutely massive. We will likely see this technology created for zero-sum games before poker, but it’s definitely a development to monitor.



There is a lot of information to digest in the two hours that Lex Friedman and Noam Brown spend discussing poker and AI, but one thing that becomes clear is that this is only the beginning. Whether it’s the development of large language models or more human-focused AI, these developments will significantly change how poker is played in the future.

It is also evident throughout the episode that these developments won’t just change poker. We will see more and more AI in different parts of our lives, and its continued development will change the very way we live forever.

Did this article deal you a winning hand?

Jackpot! You’ve flopped a winning hand! This article has surely added some extra chips to your stack. Tune in for more valuable insights and pro-level strategies!

Looks like you’ve been dealt a bad beat. We’ll shuffle the deck and try again.

Paul seaton


Paul Seaton has been a professional poker writer and reporter for 12 years at the World Series of Poker, on the European Poker Tour and as part of the World Poker Tour team too. Passionate about interviewing many of poker’s best such as Daniel Negreanu, Erik Seidel and Phil Hellmuth Paul has been nominated for a Global Poker Award for Best Written Content.

More by Paul
Keep your game tight mobile Keep your game tight

Join Us On TikTok