The AAAI (Association for the Advancement of Artificial Intelligence) annual conference starts today. AI researchers get together to collaborate - discuss - learn about technologies that are helping to make machine work more like humans - a very difficult task. Humans are great at a whole host of capabilities that computers cannot even begin to perform. Sure, computers have recently solved the game of checkers and specially designed super computers can take on grand masters at chess - but these are example of brute force calculation - not intelligence. Games that cannot be solved simply by enumerating all possible combinations will continue to give computers a run for their money. Take for instance the game of poker.

The AAAI Man vs. Machine Poker Challenge is the attempt by the research community to take AI to that next level of “intelligence” by pitting two poker pros against a poker playing program. Embedding a winning poker strategy into a computer program is no easy task - especially when challenging experienced human poker players. Humans are extraordinarily adaptive and are able to adjust to their competitor’s strategies. Just crunching through billions of possible ways to play a hand will not be sufficient for a computer - it needs to be able to learn on the fly as a human does - this, I will suggest, is the essence of Artificial Intelligence.

This is not to say that some strategies cannot be used to help give the computer an advantage - especially with the benefit of not having emotions to fog its decision making. The art of bluffing in poker is a very human quality - so, can it be codified and used as a computerized strategy? Things like bankroll management are much easier to implement as part of system - and the computer has a huge advantage with its calculating capacity to keep ahead of a human in this area. The most obvious computer advantage is the ability to calculate pot odds - although most pro poker players are able to do the same.

Who is making the bigger gamble at the AAAI conference when humans play poker against machines? My money is still on the humans at this stage of the game - it is a real gamble by the AI research community, but the only way to eventually exceed the human capability is to learn by playing against humans.