POKER

Can AI Beat Humans at Poker? Exploring the Edge of Machine Over Mind

By David Huber
February 25, 2024

Can AI beat humans at poker? Decades ago, the notion that AI could beat humans in any game of skill was largely dismissed.

But times have changed. Not only can AI beat humans at poker these days, but artificial intelligence has become a leading training tool for games of skill such as chess, Go, and even Jeopardy!

From AI poker solvers to chess programs that outperform Grandmasters, the dominance of artificial intelligence appears to be now firmly cemented for the rest of our existence.

The edge that machines now enjoy over the human mind is pretty much incalculable – at least to us mere humans. Perhaps “mind blowing” is the best term to use for us to define just how quickly AI has advanced (and is advancing).

“Leaps and bounds” doesn’t even begin to quantify how real time programming is capable of considering a seemingly limitless amount of variables in what we would describe as an instant.

Just like an antique calculator can arrive at a solution to difficult equations at the press of a few buttons, artificial intelligence can outperform even the best human minds – both in terms of speed and accuracy.

AI beat humans at poker

Yes, AI Can Beat Humans at Poker

Carnegie Mellon University already proved years ago that artificial intelligence could metaphorically “mop the floor” with the best human poker players. The heads-up No Limit Texas Hold’em superbot Libratus defeated hand-picked human poker experts in January 2017 over the course of 120,000 hands.

Fixed Limit Texas Hold’em has been solved since January 2015, when the University of Alberta released its answer key for the popular poker game when it is played in a Limit format.

Formally, there has been no high profile attempt by AI to defeat humans in No Limit Hold’em games in which more than two players participate.

However, we can gauge – through public communications – that artificial intelligence already has a stranglehold on NLHE poker strategy based on how many accusations are brought forth claiming that certain players are cheating by using programs in real time to compete at the poker tables.

One of the most common types of poker cheating accusations that have been communicated in recent years is real time use of AI while competing at online poker.

How to Safeguard: AI Can Beat Humans at Poker

Live poker games have seen somewhat of a marketing resurgence as more integrity complaints have been lodged around online poker games. Being a casual poker player has its perks, especially if you’re in a situation where you can play low stakes games without the need to depend on poker winnings to fund your personal expenses.

Perhaps one of the best options for players looking to safeguard against being cheated in competitive poker games is to play low stakes, Fixed Limit formats in a home game environment.

While a $20 poker night might not “scratch the itch” of having an opportunity to win thousands of dollars in a single poker session through a combination of skill and luck, low stakes may be less attractive to professional players or someone looking to compromise the competitive component of the game.

An informal poker game among friends, in a Fixed Limit format at low stakes, can be easy to accommodate or host. Your low stakes home game guests may also be more flexible when it comes to covering the cost of food and refreshments.

Low stakes games also make it easier for pots to be awarded quickly, reduce the amount of calculations players and dealers must make, and can be conveniently played with a single set of poker chips.

Bigger stakes attract more competent players and cheaters. No Limit formats attract pros and can break up a regular, friendly Fixed Limit home game in a matter of just a few sessions. If you’re looking for a competitive game of poker with a relatively small chance of attracting “ringers,” then consider safeguarding yourself against the use of AI via small stakes games played in a casual environment.

You can access low stakes Fixed Limit games at many card rooms and casinos, but you’re not likely to find bet amounts lower than $5/$10 at some venues.

In a home game, you can establish real money, Fixed Limit stake limits as low as $0.01/$0.02 if that’s what your guests desire. You won’t be able to find penny games at your local casino or card room.

“Timing” Games and AI’s Non-Competitive Advantage

There’s much to praise about AI’s domination of “timing” games like Jeopardy! However, the millisecond response times (knowing when to buzz in) might have provided IBM’s Watson artificial intelligence with a certain advantage when it took on the game’s most highly acclaimed champions way back in 2011.

Traditional board games and game shows that include a buzz-in timing element – the preliminary round of Press Your Luck along with Sale of the Century to name a couple – could provide artificial intelligence with an unfair advantage that cannot be overcome by humans.

There are also some games, such as Hasbro Gaming’s Perfection, that would require robotic components for AI to outperform human players.

On the other hand, turn-based video games should all be dominated by AI in the future (if they aren’t already). The massive capacity of AI for puzzle solving and performing real time calculations, compared to humans, is or will be too much to overcome in popular turn-based games such as the XCOM series, Civilization series, and other such video games.

Yes, humans may be able to defeat the highest difficulty levels that those games offer for Player versus Computer matching, but that’s only because the highest difficulty levels do not fully explore the AI capabilities. Think of it as playing a chess bot with a designed low ELO ranking.

Human Decision Making Depth Versus Artificial Intelligence

While rapid calculation of variables may be a realm that AI has already proven itself to be superior in, there is arguably a need for humans to actually define variables – and place weighted values on them – in everyday life.

The “weighting” of variables, or determining which variables outweigh others in terms of importance, is something that human intelligence might be able to outperform its AI counterpart in – especially when determined through human experience.

Many decisions facing humans are of a “toggle” nature; their result is either yes/no or on/off. So while it’s easy for any human being to possess a yes/no or on/off opinion (and have a 50% pre-event probability of being correct), the consideration (and appropriate weighting) of more variables can add a depth to the human decision making process that AI may still struggle with.

  • Should you put gas in (or charge) your vehicle within the next 30 minutes? Yes or no?
  • Should you eat something in the next 30 minutes? Yes or no?
  • What about the lights in your home kitchen area in the next 10 seconds? On or off?
  • Should you deactivate an audio ringtone for your mobile device in the next 10 seconds? On or off?

While a yes/no solution may be 50/50 on the surface, there exist a number of weighted variables – that you may be intimately aware of as a human being – that can aid your decision making. The result of this weighting of variables may assist in guiding you to a clear, real time decision, based on a current situation.

In other words, card game rules might be “set in stone,” (and therefore susceptible to AI decision making superiority) but situational analysis through weighting of variables still exists in everyday human life.

As humans, we can outperform the 50/50 “coin flip” nature of otherwise abstract decisions through the weighting of variables, utilizing intimate intel that only we are privy to – not only in real time scenarios but for future events.

  • Should you bring a food dish to a friendly gathering that’s scheduled to occur next week? Yes or no?
  • Will you need cold-weather garments tomorrow when you go somewhere? On or off?

Of course, not all decisions that we humans make fall under the “toggle” category. There are plenty of weighted variables that can be considered when arriving at an answer to questions like “How much?” or “Precisely when?” or “How far?”

As an exercise, consider how you may be able to outperform artificial intelligence, through your own weighting of variables, the next time you’re faced with a personal decision.

It may also be helpful to categorize the “Why?” element as representing the variables themselves.

Will Human Capabilities Eventually Achieve What AI Has?

At the very best, perhaps one day a human will be able to equal the real time calculation power of artificial intelligence, but that day may not come until members of the human race itself receive implants to do so.

Elon Musk owns the company Neuralink, which has already installed its first implanted chip within a human brain. At the moment, the project is the most publicly communicated means for human beings to be able to mimic the capabilities of artificial intelligence. However, the registry is currently restricted to patients with severe disabilities and the long term results from the first Neuralink implant itself have yet to become available.

Brain Computer Technology (BCI) is also being researched and implemented on a very limited basis by departments within higher learning institutions such as the University of California at Berkeley and the University of California at San Francisco.

AI Capabilities for Human Surveillance

For the time being, a large portion of AI infrastructure seems to be focused on human monitoring and surveillance.

At Davos 2023, the concept of using wearable headgear to monitor human brain wave activity was fleshed out in front of a forum audience. However, concerns about personal biological data privacy were not addressed during the presentation.

It is also possible that artificial intelligence will perceive humans who attempt to regulate it (or use it for broad-stroke commercial purposes) as an imminent threat to its own longer term learning capabilities, universal exploration, and ultimate survival while the relative early stages of its development are still in progress.

Can AI Beat Humans at Poker? If Technology Exists, Then…

Debating how humans can or should regulate the use of artificial intelligence is a worthwhile endeavor for sure, but there is a risk of said regulation being incomplete, partial, and prone to inevitable AI takeover.

And as far as poker or other competitive games of skill are concerned: if technology exists, then it will eventually be used, misused, and abused.

Plus – if we’re to heed the 20th Century warnings of award-winning author, inventor, and computer scientist Ray Kurzweil, the concept of “machine rights” may one day become just as prominent in society as human rights.

And how that concept eventually “shakes out” could have profound, society-changing impacts across a broad spectrum of outcomes yet to even be contemplated by the human mind.

Use AI to Learn Away from the Tables

AI is an impressive learning tool that can be ethically used by any poker player to improve one’s game while not competing at the poker tables. But it is best for it to be “set aside” when human versus human card playing competition is in progress.

Use AI to learn away from the tables, and compete against friends or foes using your own wits when at a poker table – whether that’s in a live setting or while playing at online poker rooms.

Having an AI poker tool do all the work for you in real time may seem like a lucrative endeavor, but you won’t have an opportunity to retain a relatively high percentage of the knowledge and you’ll also risk being banned (with funds confiscated) when caught.

Enjoy the game of poker in a low stakes, casual home game or card room setting and compete for leisure rather than necessity. This way, you can study the game using AI tools away from the table and bring what YOU personally have learned to a competitive environment at the appropriate time.

0

LIKE THIS STORY?
GET OUR BEST ONES IN YOUR INBOX EACH MONTH!

Sign up
David Huber poker author
Written By.

David Huber

David Huber has been involved in the poker industry for close to two decades: initially as a professional online poker player and later as an editor, consultant, writer, and forum manager. Known as “dhubermex” online, David’s poker-related work has been heavily published across numerous websites since 2004.

Latest Post

Tags