Man versus machine, modern technology facing off against the human mind and spirit!

This scenario isn’t another sequel to The Terminator. It’s a series of poker challenges where humans bet and bluff against complex computers over the last several years.

Modern computer programmers have worked to challenge master players in games like chess, checkers, and other games. 

Tech companies, software developers, and university research programs see these challenges as opportunities.

They can better understand how to learn and develop various types of programs and technology advancements.

High-Tech and High Stakes

The quest to defeat humans in complex games has a long history. One of the first significant efforts came from IBM in the 1980s. The company was working to develop a computer that could defeat some of the masters in the game.

Those efforts culminated in 1996 with Deep Blue battling Russian chess grandmaster Gary Kasparov.

This time the human came out on top, winning four matches and losing two. Computer engineers got to work on the machine, giving it an overhaul after learning from the first series of games. 

Kasparov again battled the computer a year later, with Deep Blue coming out on top this time – 

  • Winning three matches
  • Losing two
  • Reaching one draw

The series was a breakthrough in the use of artificial intelligence. And some computer programmers and software developers soon turned to poker. Like chess, poker offered a game of many possibilities and decisions. Any computer opponent would have to “learn” and adapt.

High-Tech and High Stakes

Unlike in chess or backgammon, in which both players’ moves are clearly legible on the board, in poker. a computer has to interpret its opponents’ bets despite never being certain what cards they hold,” the New York Times noted.

Poker offered a game with endless possibilities and one that could be “beaten” by technology. The poker world became the latest group of players considering whether man or machine would come out on top.

Polaris Sets a Poker Versus Computer Trend

One of the first significant computer efforts at cracking human poker players came in 2007 at Canada’s University of Alberta. Researchers used a several poker bots and programmed fixed strategies. 

The computer could choose among these strategies when involved in a poker match.

Over two days in July 2007, Polaris battled against poker pros Phil Laak and Ali Eslami at the Hyatt Regency Hotel in Vancouver, British Columbia. 

The challenge saw four duplicate matches with 500 hands dealt per match. In an identical game, the same cards were dealt to the human players and Polaris, but with the seating reversed.

This setup ensured both players experienced the same cards and had the same opportunity to play both sides of a hand. It was also a way to reduce the variance involved in poker over a shorter term.

Laak had played an earlier version of Polaris called VexBot in 2005. He beat VexBot. But he had noted, afterwards, that his victory came with a bit of luck. 

How would he and his fellow poker-playing human come out in this challenge?

The humans found two wins, one draw, and one loss

Polaris Sets a Poker Versus Computer Trend

A stronger version competed in the Second Man-Machine Poker Championship in Las Vegas in 2008. This time, Polaris proved to be a much tougher customer. 

During six sessions, Polaris scored three wins, two losses, and a tie against the human players.

Each was an exact match of 500 poker hands against two different players. There were 6,000 hands, and Polaris finished up 195 big blinds in that span.

The computer’s success was a sign of things to come. Researchers worked to improve their AI efforts and see a machine become a steadier and much mor formidable opponent.

Claudico Takes a Crack

Carnegie Mellon University professor Tuomas Sandholm and some of his graduate students designed the Claudico AI computer program. Claudico’s team attempted to make the program learn independently rather than pre-programmed with many strategy options and alternatives.

That was easier said than done even in the 2010s. They needed a supercomputer with 16 terabytes of RAM to accomplish the task.

"Poker is now a benchmark for artificial intelligence research, just as chess once was,” Sandholm said of the project. 

It's a game of exceeding complexity that requires a machine to make decisions based on incomplete and often misleading information, thanks to bluffing, slow play, and other decoys."

In July 2014, Claudico won a poker tournament against other computers but would get an even bigger test against human players in 2015. 

Claudico Takes a Crack

The computer faced off against Dong Kim, Jason Les, Bjorn Li, and Doug Polk in a series of heads-up matches from April 24 to May 8. 

  • Claudico battled in two matches over 750 hands for eight hours each day. That worked out to 20.000 hands per player.
  • The goal was to provide a large sample of hands to take luck out as a factor as much as possible. 
  • 80,000 hands became the largest number of human-versus-computer hands ever reached.

There was also some cash on the line, with a $100,000 prize pool donated by Rivers Casino and Microsoft. The action was even streamed live on Twitch. Highlights from the match appeared on CBS Sports Network’s Poker Night in America.

In the end, the poker players won the match with a collected score of 732,713 chips. Dollar values weren’t in play. Overall, the group of human players proved dominating.

After the match, Polk noted that the AI programmers still had work to do based on how the program played the game.

Where a human might place a bet worth half or three-quarters of the pot, Claudico would sometimes bet a miserly 10 per cent or an over-the-top 1,000 per cent,” he told PokerNews. “Betting $19,000 to win a $700 pot just isn't something that a person would do.

Libratus Tops Human Minds

Early attempts at creating computers that could beat human poker players proved terribly unsuccessful. Skilled poker players could find shortcomings in their AI opponents in the same way that the best players do at real poker tables. 

Libratus Tops Human Minds

The AI and software opponents couldn’t adapt or change at the same pace as human players.

In 2017, the computer science department at Carnegie Mellon University looked to change that. They wanted to see if AI could, indeed, conquer human opponents. 

They recruited a group of four top poker players for the “Brains vs Artificial Intelligence: Upping the Ante” challenge at Rivers Casino in Pittsburgh, Pennsylvania. 

Carnegie Mellon’s “Libratus” poker artificial intelligence would play the pros (Jason Les, Dong Kim, Daniel McAulay, and Jimmy Chou) for 20 days.

Libratus mowed down their opponents.

After 120,000 hands of heads-up No-Limit Texas Hold’em, Libratus had scored a collective $1.8 million off his opponents. The scientists involved in the challenge gained insight into much more than poker. They observed how AI could factor in situations with incomplete information - which is undoubtedly the case in a game of poker.

Using poker, the researchers helped advance the uses of AI in the process. The team outlined possible future usages as follows:

  • Business negotiations
  • Military strategy
  • Cybersecurity
  • Medical treatment and more!

Most players think of a poker bot as cheating and a negative aspect of poke. In this case Libratus proved to be a winner for the advancement of technology – as well as at the poker table.

“The computer can’t win at poker if it can’t bluff,” Carnegie Mellon head of the Computer Science Department Frank Pfenning told the university’s news bureau afterwards.

Developing an AI that can do that successfully is a tremendous step forward scientifically and has numerous applications. Imagine that your smartphone will someday be able to negotiate the best price on a new car for you. That’s just the beginning.

Pluribus and Poker

In 2019, Carnegie Mellon researchers were back at it, with Facebook as a partner. They were hoping for success in poker games involving more than one opponent. 

The goal was for AI to beat a standard game of poker featuring some of the top players in the world.

Pluribus and Poker

This time around, the university’s computer science team would be squaring off with players like the following pros – 

  • Jason Les
  • Six-time World Series of Poker champion Chris Ferguson
  • Four-time World Poker Tour champion Darren Elias

Pluribus couldn’t try to predict the endgame. Playing poker against multiple opponents meant it had to be able to reason in real-time.

Facebook AI Research scientist and one of Pluribus’s creators, Noam Brown, told the Wall Street Journal:

Multi-player poker is considered less a game than an art form that requires a multitude of skills, especially the ability to read human interactions and leverage that knowledge to exploit mistakes and weaknesses,” the Journal noted. 

Unlike predecessors, Pluribus developed its own strategy. It even used bluffs by “seeing” trillions of hands while battling against the five other opponents. 

Players couldn’t match the data set and analysis of the machine and software. And even the best poker players face physical shortcomings after long hours at the table.

A good AI has a ridiculously unfair advantage against humans: they don’t get tired,” Pluribus player and long-time pro Michael Gagliano told the Journal. “They don’t get hungry. They don’t deal with emotions.

In the end, Smithsonian noted that Pluribus “won an average of around $5 per hand, or $1,000 per hour, when playing against five human opponents.” 

Not a bad run at the tables.

MIT also Features Poker AI Research 

Poker AI Research

Carnegie Mellon and the University of Alberta aren’t the only universities to use poker to advance technology. Massachusetts Institute of Technology (MIT) hosts an actual poker bot’s tournament. 

As noted earlier, bots are not allowed according to most online poker sites’ terms of service. 

But these bots are for technological advancement and research involving programming, software development, artificial intelligence, and more. 

The annual event features teams of one to four student programmers. They play against each other as they see which creation has the best skills at the poker table.

As a game of incomplete information and uncertainty, poker is a prime application of the game theory concepts and decision-making skills essential to trading,” the tournament site notes. 

While traders make risk decisions based on the limited information they get from the markets, poker players make decisions based on hidden information as well, taking into account factors such as expected value and probability distributions.

Competitors also try “applying concepts in economics, mathematics, and computer science not normally developed together in academic settings in order to conquer their opponents and emerge victorious.

While this may not be the World Series of Poker, the competition can get fierce and has attracted several sponsors through the years. Each team has one month to program a completely autonomous poker bot. 

The 2022 competition featured more than $40,000 in prizes. It also garnered attention from quantitative trading and technology firms.

Do AI & Computers Have An Advantage: Does It Matter?

Do AI & Computers Have An Advantage

It seems technology may be gaining the upper hand in these mano-a-high tech battles. Computers have many advantages right out of the gate. 

  • First, there is no financial disadvantage for AI. 
  • A human opponent faces real monetary consequences in a game of poker. 
  • A missed draw for a computer doesn’t affect the machine or software. 
  • A poker player faces real-life implications for a losing session in the real world. 
  • They may also have to think long-term about their game rather than a single hand.

A lack of physical information from a machine is also an advantage for the machine. Several players who participated in the AI battles noted – 

A machine doesn’t give off any signals that it may be bluffing. Physical tells that may prove crucial at a real poker table aren’t available in these high-tech Hold’em sessions.

Obviously, table talk and chit-chat are out of the question as well. 

  • Players can’t pick up on verbal tells or audible words that might give away the computer's playing style. 
  • Players can be in more of a guessing game in more challenging spots, especially against some of the better AI poker players.

But in the broader picture, it’s crucial to remember that poker is a game for human players. Part of the fun of playing is using one’s wits and intelligence - daring to defeat other players. 

Winning is still a thrill.

Even when playing online, you’re battling other human players.

AI may be getting the upper hand as of now. But winning or losing makes no difference to a computer program. Whether AI can defeat a random player is of no consequence in the real world of poker. Heading to the cashier with more money than a player started with remains satisfying. 

Defeating other humans is always way more fun.

Sean Chaffin is a poker writer who appears in numerous websites and publications. He is also the host of the True Gambling Stories podcast