Two artificial intelligence (AI) programs have recently showed they have what it takes to “know when to hold ’em, and when to fold ’em,” after grabbing the headlines following triumph over human professional card players when playing Texas Hold’em.
So what does the future hold for poker and AI?
First it was chess – the great symbol of intellectual competition – that was the mountain that artificial intelligence would have to climb if it wanted to really worry humanity.
But in the end Deep Blue beat Gary Kasparov back in 1997 without breaking a sweat and without resorting to anything that could be mistaken for intelligence.
Instead it brute force calculated its way through every possible option with a little help from a database of human games than helped narrow down its massive field of calculation.
For a while poker has been seen as an alternative yardstick for AI. Unlike chess and go poker is a game of incomplete information, with what information there is frequently muddied by irrational opponents.
By applying constraints such as limited bet sizes, game theory has allowed mathematicians and their machines to solve certain types of poker. Some short stack situations have been solved for a long while.
And more recently a computer came up with a game theory spread limit heads up strategy that will turn a profit consistently against any opponent.
But the best players don’t play a game that guarantees a profit against any player adjust their strategy to exploit specific players.
They also tend to play no limit with its greatly increased number of possible bet sizes and factors like pot and effective stack size that make every decision almost unique.
By setting up an AI to analyse a huge database of hands these AIs have passed certain thresholds ‘in the lab’ with the Liberatus bot beating top level human players over a significant number of hands.
But for the moment Liberatus remains an expensive university gewgaw.
Poker Alfie on the other hand, might not be able to beat Daniel Negreanu heads up but odds are it’ll give you a run for your money. Under the rubric of a training tool you can play PokerAlfie for free. But it is the nature of these things to trickle down and fast.
What would an affordable version of Liberatus mean for online poker?
There are pretty much only only a few possible outcomes. The online poker providers will have to stay ahead in the arms race against the bots. Or online poker will have to come to an end… for humans.
The question that will decide all this will be simply whether or not a profit can be turned.
If an AI can beat other AIs by more than the rake, then online poker will live on as a place where very soon only robots are good enough to compete. Your average poker room will be fine with that, they still get to rake each hand regardless of how organic the players are.
Once the last human realises its a fool’s errand to play against the robots for cash, we’ll all either decide to go play in our local cardroom or set up a home game circuit, or we’ll become the mechanics, bankrollers, and fine-tuners for our digital surrogates.
New algorithms and more experienced neural nets will rise to the top to be our new Isuldur’s and Tom Dwan’s, no doubt taking their owners with them.
Meanwhile more basic AIs will try and eke out a living for their human masters in the low stakes games, perhaps at those stakes even human brains will still be able to compete unaugmented.
The End of the World as We Know It
On the other hand if the AIs are too closely matched and no one beats the rake, or people don’t put the effort into the arms race and the ecology of bots stagnates.
Then perhaps online poker will peter out leaving university departments to play against each other for academic kudos.
With no humans online for fear of bots and no bots online because there’s no profit in it. The lights will go out across the online poker-verse and people will head back to the live action alternatives.
Depending on how difficult the access to bots becomes, and on how prevalent the moral decrepitude becomes it may be that there is a tolerable number of bots that won’t kill the ecosystem.
When simulations are run of animal populations where populations are made up of either honest-players or cheaters there tends to be an equilibrium struck in the digital populations.
If everyone is a cheater, there’s no one to cheat and the cheaters starve being replaced by honest players.
When there are very few cheaters the cheaters gain the advantage again and their population swells until it hits the equilibrium amount.
Perhaps in the end, AIs will just be part of the cost of playing online.
Want to try an online casino?
Choose an approved casino from our carefully selected list. VIEW CASINOS