Even people without televisions, like me, finally found out about the man versus machine competition on Jeopardy, with IBM's Watson. From this article in the Boston Globe, I had the mistaken notion that Watson already has voice recognition, but reading the PC World report on IBM Watson Wins Jeopardy, I understand now that Watson actually operates on text. They hope to introduce voice recognition in the future. What was apparently the big breakthrough was that the computer can not only understand complex natural language questions, but calculates the likelihood of any of 3 possible answers being correct. If any of the answers rises above the 45% threshold, Watson chooses that, or the highest of the answers rising above that level. Watson also has a complex algorithm for calculating how to place its Jeopardy bets. The natural language was the biggest challenge, and it goes beyond the natural language search capabilities we see in Westlaw and Lexis search engines today. The developers say they were inspired by the television series Star Trek and the computers on the star ships where crew members had only to ask a question and receive an answer or get a clarifying question in return before getting an answer. (See Wikipedia article, and IBM links Pressroom and Watson pages).
I see from the Globe article that the two television shows, Jeopardy and the Nova show that made a documentary about Watson competing on the game show, are in the nature of marketing for IBM and its partner Nuance Communications, Inc.:
The Watson intelligent computer system from IBM Corp. was a triumph, and not just because it trounced two human champions in the TV game show “Jeopardy!’’ Watson also scored a massive marketing coup: It became a celebrity and household name, and set the stage for new commercial products based on the smart machine.(Johnny Diaz,Boston Globe). They won't be able to add voice recognition and roll out the commercial product until 18 – 24 months from now. This article emphasizes the amount of memory, while stressing how much the reports have varied: somewhere between 4 terabytes and 16 terabytes of memory. 10 server racks with 10 IBM Power 750 servers and 2 large refrigeration units (well, of course! -- I'll bet it gets hot) housed in their own room at the IBM Yorktown Heights campus. The project to develop the health care computer system will be in conjunction with Columbia University and the University of Maryland. At least one article called the proposed product a cybernetic medical assistant.
Both wins were by careful design. On the day after Watson’s last Jeopardy!’’ game, IBM and speech recognition software maker Nuance Communications Inc. in Burlington revealed that they had partnered to produce a medical version of the computer system for the health care industry. The idea is to use the Watson combination of speech recognition, superfast processing, and a massive database to help doctors and nurses, who will enhance their diagnoses of patients by talking to the machine.
Media analysts said that a marketing strategy built on Watson’s new celebrity, which was boosted by the appearances on “Jeopardy!’’ and in a PBS documentary, paved the way to sell the technology.
“By putting Watson on ‘Jeopardy!’, it humanizes the technology,’’ said Geoff Klapisch, a Boston University marketing professor. “It easily demonstrates what it can do, and goes beyond the traditional route of introducing a service to a niche market.’’
Robert Thompson, professor of television and popular culture at Syracuse University, called the Watson rollout “a brilliant marketing ploy both on the part of the manufacturer and the part of ‘Jeopardy!’ . . . This is better than a Super Bowl commercial.’’
The Wikipedia article, under “Future uses,” comments that the general counsel of IBM, Robert C. Weber, has suggested that Watson may also be adaptable for legal research, and refers the reader to “Why Watson matters to lawyers,” an article written by Weber in the February 14, 2011 issue of the National Law Journal. In this article, Weber claims that Watson has 2 kinds of artificial intelligence. It understands questions asked of it, and it learns by experience. (I asked my computer scientist son if these two activities qualified as artificial intelligence, and he confirmed that they do. Bayesian logic (learning by experience), and natural language capability constitute forms of artificial intelligence.) IBM is calling the intelligence “Deep QA,” and Weber goes on,
...it's becoming clear that this technology will be useful in a couple of ways: for gathering facts and identifying ideas when building legal arguments. The technology might even come in handy, near real-time, in the courtroom. If a witness says something that doesn't seem credible, you can have an associate check it for accuracy on the spot.Well, I look forward to a computer system that can offer me a cup of Earl Grey, hot, as well as doing complex legal research.
Many types of organizations and professions could benefit from the increased insight that the Deep QA technology delivers. Think about the possibilities for medical diagnosis support, for better anticipating the energy needs of utilities or for protecting insurers, banks and governments from fraud. Deep QA won't ever replace attorneys; after all, the essence of good lawyering is mature and sound reasoning, and there's simply no way a machine can match the knowledge and ability to reason of a smart, well-educated and deeply experienced human being. But the technology can unquestionably extend our capabilities and help us perform better.
The image decorating this blog post is from the Globe article, but credited to Jeopardy Productions via Associated Press. It shows, from left to right, Jeopardy host Alex Trebek, champion contestant Ken Jennings, the avatar that represents Watson, a globe encircled by 42 lines which change color depending on the computer's certainty about an answer, and champion contestant Brad Rutter. The 42 lines on the avatar for Watson are a joke on the supposed answer to Life from The Hitchhiker's Guide to the Universe by Douglas Adams.