PokerStars Want You To Help Them Make Poker Bots

Artificial Intelligence seems to be undergoing something of a renaissance at the moment. Having been largely the purview of spec fic writers like Isaac Asimov and Philip K Dick, the vast leaps forward in computer processing power that allow even Jo Blow on the Street to set up a neural net on her home computer, have led to an explosion of think pieces ranging from CGPGrey’s YouTube videos (watch below) to Stephen Hawking’s claim that few things are more existentially threatening to homo sapiens than the rapid approach of the digital singularity.

PokerStars are not taking so alarmist a position. In fact they are hitching their wagon to the AI train. Recent jobs advertised on their website include Poker AI Research Engineer and Graduate Poker AI Research Engineer for any London or Dublin based programmers who will be “supporting, and often concepting AI technologies for poker, [they]’ll be helping with implementing and evaluating algorithms and providing software design and programming for projects.”

We may not be at the singularity yet, if only because some tasks seem rather beyond AI’s capability at the moment. Examples include AIs designed to name villages or create motivational posters both of which have proven hilariously inept, churning out responses that range from the obscene to downright sinister. But games, with their easily defined parameters, are where computers have always excelled.

Poker Bots and Deep Blue

Go always represented a holy grail for common knowledge games where everyone has full information because its large board of 19 x 19 spaces made brute force calculation impossibly long-winded. It’s been awhile since the glorified calculator that is Deep Blue beat Kasparov, but in recent years AI and neural nets have allowed scientists and engineers to create programs that can beat the best humans in the world at go.

Poker, with its hidden cards and hidden motives makes for a range of new AI problems. In particular learning exploitative strategies to take advantage of the suboptimal ‘greater fools’ whose play may not be perfectly game theory optimal. 

PokerStars Needs You

It’s a challenge that researchers in Alberta and at Carnegie Mellon, among other places, are putting time and funds into. And some of these AIs have done extremely well at both GTO and exploitative play against some of the best players in the world. PokerStars approach however appears to be aimed more at using the AI to help game development rather than creating poker bots.

The job descriptions on Stars’ website say:

“As part of the poker innovation team you’ll get to work on a number of high end products we wish to bring to market and how AI interacts with this to help us learn from and analyse strategies developed through self-play.”

So if you’re in the right area geographically and academically this might be your chance to change the game up for all of us.