Earlier this year, in the wake of the Libratus AI bot soundly defeating a team of top heads-up poker pros, I wrote a lengthy op-ed called In Defense of Humanity. In the first sentence of my conclusion to that article, I wrote “The moment we start caring more about the poker accomplishments of machines and bots than flesh and blood players, all hope for our game is lost.” Not that I’ve wavered in my feelings about the issue, but it’s impossible to deny that poker AI is nonetheless quite impressive, especially when you consider how quickly its capabilities have advanced in such a relatively short amount of time.
It was back in 1984 that the first efforts to create an artificial intelligence machine dedicated to winning at poker were successful, when Mike Caro debuted the Orac software. Since that time, numerous companies and universities from all around the world began researching the potential for poker AI advancements, as it deals with imperfect information, similar to humans in real life.
The detailed, interesting infographic below, created by the good folks at poker sites, is followed up by a short article outlining five interesting facts you ought to know about AI.
1. People of all ages are well in favor of AI
Not many people are aware of this, but research has shown that 35% of people over the age of 55 trust AI and 62% of millennials aged 17-24 do as well. Moreover, 71% of people over 50 strongly believe that at some time in the future, intelligent, artificial intelligence-based assistants will help simplify their lives.
2. The extent of the imperfect information dealt with in poker
In Limit Texas Hold’em Limit, there are over 316,000,000,000,000,000 different game situations that can be attained. In other words, that means that this form of poker alone presents more unique situations that the number of atoms present in the entire universe! A human would need a total of 10 billion years, with a play rate of one second for each game, to finish up all of the possible situations.
3. The advantages of Poker AI over the human mind are astounding
As I mentioned in the op-ed I references at the outset of this article, AI cannot get tired, therefore it does not make poor decisions once it is fatigued. Additionally, an AI cannot feel the value of money therefore when it comes down to bluffing or folding, it will not get scared- rather, it’ll simply calculate the odds, due to its lack of emotion. Not only this, but an AI is also capable of identifying various types of weaknesses in players competing against it.
4. Online poker bots aren’t really “good at poker”
Although using bots is illegal on all gambling and poker sites on the market, it’s important to point out the fact that no bot would be able to beat a table full of poker players. The average win of a poker AI bot is around 5bb per 100 hands, which isn’t really much. The thing is that the bots are programmed to face off in a specific poker discipline (limit hold’em) against one player. It’s relatively helpless at other poker variants against multiple opponents.
5. Online poker bots can quickly be identified
While they do not pose a real threat to the online poker community these days, it’s important to be aware that repeated use of identical bet sizing, identical timing, using uncommon lines, and failure to complete CAPTCHAs during the game all point to a bot playing, rather than a real human.
What does the future hold?
For poker AI bots to progress any further, programmers would need to address what I noted above in #4 as well as extensively in my aforementioned op-ed. In other words, AI developers would need to create an AI machine capable of playing against multiple pro opponents (at the same table) and beating them. As an AI is capable of doing over 20,000 years-worth of human work in a single week, the possibilities for AI in the future are basically limitless.