A recent "experiment" on the Conan show had a version of IBM's Watson computer try its hand as announcer, leaving some to ask questions if computers (like Watson) will be the first choice for filling knowledge-based job positions. As noted in this report on USA Today, IBM announced a partnership with Nuance Communications to develop a "physician's assistant" technology based on Watson.
"How did Watson's technology allow it to sift through some 15 trillion bytes of information (one byte is roughly equal to one piece of information, say a letter like "h") to conjure the answers to oddly-worded questions ranging from Beatles songs to Familiar Sayings? Watson actually is a departure from more frequently-trod paths to artificial intelligence, Kelly says, which more often seek to give computers an expert understanding of a very specific area of knowledge (a famous example is a natural language program called Lunar that could answer questions about moon rocks). Such expert systems are 'a dead end,' IBM research executive John Kelly says, not able to handle the broad range of questions and knowledge that Watson or any of its real-world successors would face.
"Instead, the computer focuses on sifting through documents to find possible answers and then relates them to each other using methods borrowed from fields as diverse as linguistics and logic to develop confidence in possible correct ones. 'Watson is a triumphant celebration of thirty years of advances in efficiently indexing a large, broad library of texts, and fifty years of advances in statistical machine learning that can train on a huge set of past Jeopardy! example clues and responses,' says machine learning pioneer Doug Lenat of Cycorp Inc. in Austin, Tex."