Thinking Robot to Play Jeopardy!

In 1997, the “Deep Blue” IBM computer beat a human world champion in a game of chess. It made true artificial intelligence a likely possibility for the masses. Deep Blue used logic to outsmart a human being in a game of analytical skill and judgment. In 2010, IBM took the next step. A computer named Watson will take on two of the most knowledgeable humans, Ken Jennings and Brad Rutter, in a game of Jeopardy!

Watson may seem to possess a clear advantage in that, as a robot, he would have the ability to possess instant and correct responses by simply accessing data. But Watson will possess only the same skills and resources that a human contestant does. His “brain” consists of a small room with a few refrigerator-sized computers that house two key sections. The first and, seemingly, most important is Watson’s “knowledge.” Watson’s memory is simply a mass of encyclopedias, dictionaries, and other factual resources that any human could read and understand. Watson will analyze these documents and pinpoint what each sentence means and how it relates to his answer.

The second facet of Watson’s intelligence is his understanding.  Artificial Intelligence development requires an understanding of human inquiry. This necessitates knowledge of “natural” language. Natural language is a term that encompasses the grammatical elements of language as well as the idioms and nuances associated with how we speak every day. Jeopardy! requires contestants to answer questions that are sometimes phrased in puns or that include sarcasm. Watson is not merely programmed with “Question=Answer” logorithms; he must deduce the meaning of the question and then search through millions of documents, infer their connotations, and answer the question only if a definite level of certainty exists.

Watson is by no means perfect. He will sometimes misunderstand questions and give incorrect responses. He will not buzz in if his level of conviction for  the correct response is not above a certain threshold. Watson, however, is surprisingly self-aware and manipulates his strategy based on his own strengths and weaknesses during a given game.

Fiction has depicted artificial intelligence for years. Since then, we have developed robots programmed to perform set tasks and others to understand data. But Watson differs from these imaginary robots in that he simply understands. The thousands of wires connecting the computers that provide him with his abilities are akin to basic synapses of a human brain.

In February, Watson will listen to Alex Trebek speak, understand his questions, and provide answers of his own. Watson is the beginning of a new era in artificial intelligence. The public will marvel at what humans have created and will anticipate the day when A.I. surpasses Watson and exists symbiotically with each and every one of us.

Watson is truly a mark of our species’ progress in modeling a being that exists in our own image.