Machine learns games 'like a human'
By Will Knight A computer that learns to play a ‘scissors, paper, stone’ by observing and mimicking human players could lead to machines that automatically learn how to spot an intruder or perform vital maintenance work, say UK researchers. CogVis, developed by scientists at the University of Leeds in Yorkshire, UK, teaches itself how to play the children’s game by searching for patterns in video and audio of human players and then building its own “hypotheses” about the game’s rules. In contrast to older artificial intelligence (AI) programs that mimic human behaviour using hard-coded rules, CogVis takes a more human approach, learning through observation and mimicry, the researchers say. The older approach is fraught with problems, as computers struggle when faced with situations that fall outside the remit of these rules and when new rules are introduced. “A system that can observe events in an unknown scenario, learn and participate just as a child would is almost the Holy Grail of AI,” says Derek Magee from the University of Leeds. “We may not have solved this challenge quite yet, but we think we’ve made a small dent.” The system was demonstrated at an event sponsored by the British Computer Society in Cambridge, UK, in December 2004. It went on to win the society’s Prize for Progress Towards Machine Intelligence. CogVis observed human volunteers playing a version of the game using cards marked with a pair of scissors, a piece of paper, or a stone. They were also told to announce when they had won or when the game was a draw. After watching for several rounds, CogVis was able to call the outcome of each game correctly. Chris Needham, another member of the CogVis team, says the system’s visual processor analyses the action by separating periods of movement and inactivity and then extracting features based on colour and texture. Combining this with audio input, the system develops hypotheses about the game’s rules using an approach known as inductive logic programming. “It was very impressive,” says Max Bramer, a researcher at Portsmouth University, UK, and chair of the British Computer Society’s AI group. He told New Scientist that CogVis could have many future applications. “You can think of lots of times when you’d like to be able to point a camera at something and have a computer interpret things for itself.” He suggests that machine’s could one day use this technique to learn how to spot an intruder on video footage or how to control a robot for important maintenance work. “It’s a very good start, and almost mysterious in the way it works,” Bramer adds. Stephen Muggleton, an AI expert at Imperial College London, UK, says CogVis combines several strands of AI research, from vision analysis to logic programming. “The result is an explicit plan-oriented theory, learned directly from visual and auditory perception,” he told New Scientist. But Muggleton says a key challenge will be to push the system to learn more difficult things. “It would be interesting to see if this approach will scale up to more complex games such as noughts-and-crosses or even beginner-level draughts,