Robot AI Set to Conquer the Online Poker World

With artificial intelligence (AI) set to possiably move into the poker world it is time to test your knowledge and see if you can beat the robotic poker champions. The game below pits you aginst four robots in an online poker tournamentCan you indivually defeat each opponment and defeat the robotic uprising?  

DeepMind, the Google owned AI company, recently hit the headlines for creating a computer programme that defeated the world Go champion, Lee Se-dol, in a tournament that ended in 4-1.  Now they've set their sights on a bigger and potentially much more lucrative project: super computers that play poker to rival the pros.  So here at Ladbrokes, we've taken a look at this emerging technology and what it could mean for us mere mortal poker enthusiasts.


Researchers from University College London and a staff member of DeepMind have attempted to teach a machine how to play Leduc, an ultra-simplified version of poker using a deck of just six cards, and Texas Hold'em, a common favourite amongst many Ladbrokes online poker fans.


Using systems similar to those that enabled AI to beat Lee, the machine successfully taught itself a strategy for Texas Hold'em which "approached the performance of human experts and state-of-the-art methods".


These systems were called "deep reinforcement learning".  Firstly the AI would use the ‘neural networks technique' where the programme will use a huge array of data to make multiple individual decisions that will all combine to ultimately solve the end problem.  This technique works wonders in cases where lots of data is available BUT this is not always possible, such as with poker.  So in this case, reinforcement learning comes into play.  If the AI cannot solve the problem using data then the machine will keep playing poker and learn from its own mistakes, much like us humans do.  It repeatedly tries to solve the problem, learning by trial and error and incrementally improves from making mistakes, whilst enhancing its own training until it is successful.  What makes this even more impressive is that it will practise playing poker against itself- potentially over a million gamers per day- which would of course mean it automatically gets 1 up over any humans trying to hone their poker skills.


By learning in this manner the machine was able to calculate the mathematically best way of playing all by itself. All this even though it had no actual knowledge of the game of poker to begin with.


The battle's not over just yet…

When comparing Poker with the ancient game of Go it actually came across as a more difficult concept for a system to grasp. The AI poker programme has to do all its calculations despite having no knowledge of the cards present on the table or with other players. In this sense the machine has an advantage over a human player in that it can instantaneously calculate the odds of what hands each player may have and play according to those odds.


The biggest disadvantage of Today's AI poker machine's however is that they can not take into account the behaviour of the players. AI poker systems are still hindered when humans raise the stakes. It's difficult for AI's to interpret game play due to the irregularity and unpredictability of humans' betting.  So there's still time to take on the machines and emerge victorious!