fbpx
The representation of a poker bot equipped with artificial intelligence

@Top1Percent

When will AI supplant humans in poker? How will we protect ourselves?

With the emergence of tools that seemed futuristic just a few years ago, artificial intelligence is the hot topic of the moment.

Everywhere you turn that’s all you talk about: Dall-E, Midjourney and the like can create fantastic images from a sentence,ChatGPT, Bing, etc. can write text, conduct research, and answer (almost) any question asked.

Many job sectors began to shake a bit, such as the carriage business when the automobile was invented. What if they steal our jobs? What will a graphic designer or copywriter be for in a year or two? Just to spice things up a bit, the images in this article were created with artificial intelligence!

With similar thoughts always growing, has anyone ever tried to extend the problem to the poker world, which affects us more closely?

Poker and technology – The importance of software

Today, technology is almost inescapable from poker. To learn, improve, and make money, the computer is an essential tool: one can play vastly more hands than live poker, use tracking software to analyze a gigantic amount of information about one’s own and others’ play, and the latest innovation is solvers, powerful programs that aim to calculate the most efficient move in each scenario.

Here comes the first point to make to the generalist media who create headlines such as “Artificial intelligence has taken poker as well” by talking precisely about these solvers.

Are poker solvers AI? How do the solvers work?

Are we sure that we can speak of intelligence with respect to a solver? In a way, yes. A requirement for software to be referred to as artificial intelligence is that it has the ability to learn, in technical terms machine learning. In solvers, this element appears in a cumbersome sense: the solver performs simulations of poker hands, sees which one turns out to be a loser, and adjusts the moves to minimize his losses (conterfactual regret minimization).

It learns what the problem is, fixes it and performs the calculation again on the other side, repeating this algorithm virtually indefinitely to reach a point where the differences are minimized. The famous balance point.

Pokker bots in the future

Why a solver can’t play poker

To solve one spot we can say it is smart, but all these calculations are not taken into account to solve the next spot. Every slight difference between one hand and the next requires that the entire calculation be done again, which for human players (perhaps mistakenly) is not quite the same.

For example, with A♣K♣ on board A♦K♠3♥ or on board A♦K♠4♣we will know that we should behave in virtually the same way. A computer will do two completely separate calculations.

Right now the computation time is too long for a GTO solver to find the best move of a hand in real time. But there are tools composed of a huge database of already solved spots that can be accessed quickly, without requiring additional calculations. Can software of this mold possibly defeat humans at poker in real time?

For the time being, the answer is no, precisely because the database though huge is limited, and it would be forced to make rough deviations, which can also lead to error. Relaunching 2.3bb instead of 2.5 may seem small to a human player, but it is enough to upset the balance of a mathematical resolution, and with the right accommodations this simplification can become exploitable.

Solver vs. AI, GTO vs. Exploitative

And speaking of exploitability, here is the difference between a solver and an artificial intelligence that wants to play poker against humans. A GTO strategy by definition aspires to reduce long-term losses to zero, that is, to have a game that is impossible to exploit. An exploitative strategy, on the other hand, aims to find weaknesses in the opponent’s game and aim to take advantage of these, purposefully unbalancing one’s strategy.

An exploitative game is on paper much more profitable than a GTO game, with the side effect of increasing the risk of making mistakes. As usual, to increase potential winnings, you have to increase risk. A law of finance that is the mantra for everything.

An artificial intelligence should learn to take into account the opponent’s playing style, weaknesses and tendencies in order to calculate the best game against the latter. He should, indeed, learn.

When an artificial intelligence faced human beings at poker

There are also examples where this has happened. The most famous was Claudico, an artificial intelligence developed by researchers at Carnegie Mellon University. This software learned how to play poker by playing an inordinate number of hands and learning from each one.
In 2015 they tested him against four poker pros, including the well-known Doug Polk, for a total of 80,000 hands.
On that day, the bot came out defeated in the overall total, but with an astounding performance nonetheless, which left researchers brimming with optimism.
For the record even Polk praised some aspects of Claudico’s game, particularly innovative and unconventional strategies, a solid game that is difficult to read, but also an overly passive and conservative tendency, which in poker is rarely profitable.

Will matches between humans and artificial intelligence be the norm in the near future?

Summing up

In the end, the time is not yet ripe to fear an AI that will send all poker players into retirement, and if that happens a solution will surely be found to ensure clean and honest play. And at worst, there will always remain healthy live poker, while AIs will remain as sparring partners to learn and improve.
Today a solver cannot play against human beings at the pace of human beings. A database of simulations will only get as close as possible to a solution without reaching it. An artificial intelligence will need more time and other small corrections before it becomes truly dangerous to a poker player.
The bottom line, however, is that without humans, poker is meaningless. Kind of like chess. Today there are chess engines (very similar to poker solvers as a concept) that can beat the best player in the world without difficulty, however, chess continues to exist and excite fans, and Magnus Carlsen continues to be on top of the Olympus even though Stockfish can defeat him.

There have been problems caused by this development of technology, sure, cheaters in chess seem to have landed in live as well, yet after the scandals a solution has always been found and the problem has always remained relatively contained. Chess did not die for Stockfish, poker will not die for Piosolver.

Top1Percent

We focus on value

© Top1Percent Jul 21, 2024 All rights reserved

Who we are

The poker school that provides you with all the tools you need to begin your path to professionalism.

Useful Pages

Home

Staking

Rakeback

Team