Blog: IBM's Jeopardy-Playing Machine Can Now Beat Human Contestants

IBM's Jeopardy-Playing Machine Can Now Beat Human Contestants
Network World (02/10/10) Brodkin, Jon

IBM's Watson supercomputer can regularly best human contestants in the game Jeopardy, and within a year it will face a Jeopardy public challenge. Among the challenges the machine faces is understanding natural language, ascertaining answers to questions, and calculating the chances that its answer is correct in order to decide whether it should buzz it in--all without an Internet connection. Understanding a question is challenging to a computer, as words often have multiple meanings and can have different relationships with each other, and must be interpreted in the appropriate context. The computer then has to search within its own knowledge base to find an answer. Watson may be reasonably confident it knows the answer to a question, but will factor in the score of the game and the dollar value of the question before choosing whether it is worth the risk. In addition, the supercomputer requires a category selection strategy and a wagering strategy for Daily Doubles and Final Jeopardy. IBM researcher David Ferrucci says the research underlying the Jeopardy challenge could help address more critical issues, such as sifting through large volumes of data and determining the accuracy of a conclusion.

View Full Article