Communications of the ACM
Ibm's Watson Breaks New Ground in Artificial Intelligence
Watson competing on Jeopardy at I.B.M.s campus in Yorktown Heights, NY.
Carol Kaelson / Jeopardy Productions Inc., via Associated Press
One of the major challenges that the computer science industry has been facing is the development of a machine that can answer unmodified human questions. "We consider it a long-standing challenge in artificial intelligence to emulate a slice of human behavior," says IBM Research's David McQueeney.
One of the potential applications for IBM's Watson computer system, which became famous playing Jeopardy, is supporting physicians' medical decisions with data in the form of cited medical papers and with quick, real-time responses to questions. Though IBM has developed computers that can beat humans in chess, it takes much more scientific effort and computational work to understand human language, McQueeney says.
IBM used Jeopardy's format to provide a question-and-answer response platform with a high degree of precision in real time. Watson is able to discern meaning from the arrangement of words and their sentence structure. Watson also has a temporal reasoning system that allows it to find data through time and geospatial calculations, weigh evidence, and understand the meaning of rhymes.
Watson's underlying technology, DeepQA, is a massively parallel probabilistic evidence-based architecture, which IBM's researchers continue to improve. The success of Watson has demonstrated that computers can now support human interactivity tasks that they were unable to do before, McQueeney says.
From Federal Computer Week
View Full Article
No entries found