Computer algorithms capable ofplaying the perfect game of checkers or Texas Holdem poker have achieved success so far by efficiently calculating the best strategiesin advance. But some computer scientists want to createa different form of artificial intelligence that can playany new game without the benefit ofprior knowledge or strategies. The software would face opponents after having onlyreadthe gamesrulebook. AnAI that can adapt well enough to play new games without prior knowledgecould also potentially do well in adapting to the rules of society in areas such as corporate law orgovernment regulations.
This idea of general game-playing AI has gotten a big boost from the International General Game Playing Competition, a US $10,000 challengethat has been heldas an annual eventsince 2005. The AI competitorsmust analyze the unfamiliar game at handsay,somevariant of chesswithina start clock time of 5 or 10 minutes. Then they each have a playclock of just one minute to make their move within each turn of play. Its a challenge that requires a very different approach to AI than the specialized algorithms that exhaustively analyze almostevery possible playover days or weeks.
This raises the question of where is the intelligence in artificial intelligence, saysMichael Genesereth, a computer scientist at Stanford University.Is it inthe program which is following a recipe,or is it in the programmer who invented the recipe and understands the rules for playing the games?
Geneserethrecently presentedthe latest advances in general game-playing AI at the29thAssociation for the Advancement of Artificial Intelligence conferenceheld from25-30Januaryin Austin, Texas. The latest champions of the General Game Playingcompetition represent the thirdgeneration of AI to have emerged since the first competition in 2005.
But the idea of general game-playing AIgoes all the way back tothe original 1958 vision of John McCarthy, the computer scientist who coined the term artificial intelligence.McCarthyenvisioned an advice taker AI that didnt need to rely upon a programmers step-by-step recipe to tackle new scenarios, but instead could adapt its behavior based on statements about its environment and goals. To paraphrase science fiction writer Robert Heinlein,such AI could behave more like an adaptable human who can write a sonnet, balance accounts, build a wall, set a bone,rather than just performa single tasklike a specialized insect.
Computer scientists usecompetitions based on games such as tic-tac-toe and chess as benchmarks of their progress. But general game-playing AI would not likely compete with specialized algorithms tofind the best solutions to the ancient game of Goor heads-up nolimit Texas Holdem. Thosespecialized algorithms are programmed to crunch all the information sets about possible moves made by game opponents at each stage of play. Such anexhaustive approach often requires intensive supercomputing resources.
By comparison, general game-playing AI can easily learn toplay new games on its ownby doing the equivalent of translating a games rulebookintoGame Description Language, a computer programming language it can understand. That means general game playing AI can rely upon just one page ofrules to learn games involving thousands of information states; chess, for instance, can be described through just four pages of such rules.
Most examples of AI tend to fall in the category of specialized algorithms following preprogrammed instructions. Some AI uses the popular approach known as machine learning to slowly adapt to new scenarios; they are, in a sense, virtual newborns that knownothing and must learn everything for themselves. General game-playing AI provides an alternative approach, incorporatingexisting knowledge rather than having to learn everything on its own.
I think there needs to be some balance between a machine that knows nothing to start and learns about world,Genesereth says, andamachine that is told everything about human knowledge and startsfrom there.
The first generation of general game-playing AI focused on maximizing the moves available to itself and limiting the moves available to opponents. Such an approach had onlylimited success;computer programs still struggled to beat humans during the first Carbon versus Silicon competition held alongside the General Game Playing competition in2005. Since that time,humans have never again beaten their silicon counterparts.
Read more from the original source:
Building AI to Play by Any Rules
