Deep Blue was an amazing achievement in the application of compute power to an extraordinarily challenging but computationally well-defined and well-bounded game. By searching and evaluating a huge space of possible chess board configurations, Deep Blue has the compute power to beat a grand master.

[...]

Watson faces a challenge that is entirely open-ended and defies the sort of well-bounded mathematical formulation that fits a game like Chess. Watson has to operate in the near limitless, ambiguous and highly contextual domain of human language and knowledge. Ultimately Watson’s scientific goal is to demonstrate how computers can get at the meaning behind a natural language question and infer precise answers from huge volumes of content with justifications that ultimately make sense to humans.

Rather than challenging the human to search a vast mathematical space, Watson challenges the computer to operate in human terms. To understand and answer human questions and to know when it does and doesn’t know the answer — to assess its own knowledge and ability — something humans find relatively easy.

Like Deep Blue, Watson may not mimic human thought processes to get the job done, but unlike Deep Blue there is ultimately no guarantee for Watson that an answer exists or that it can even be inferred from its sources, no matter how long it may search.

In other words, winning at Chess uses an entirely different skill set than understanding everyday human language. Sentence formation, especially as used in Jeopardy, is more of an art form than a science and Watson cannot simply rely on a database of possible word combinations in order to understand English.

Moreover, much in the same way that Deep Blue “sat down” in front of Kasparov to play chess with no help from computer operators or programmers other than to simply start it up, Watson will be pitted against Jeopardy players using only its own “senses” to buzz in and answer questions. It will not be networked or controlled, but rather fed each question electronically at the same time that the human players receive them. It is then up to Watson to interpret what the question is asking for, buzz in, and answer appropriately before other players do.

So it it like Googling the answers while watching Jeopardy? Not quite. Have you ever actually tried to look up Jeopardy answers online while watching? Google responds with so many results to every query that it’s nearly impossible to find the right answer in the few seconds that it takes a person who already knows the answer to ring in and respond. So even if Watson were connected to Google, it wouldn’t be able to sift through the hundreds, if not thousands, of search results it found for each question.

Making Watson understand language is part of a project called the Open Advancement of Question Answering initiative, founded by IBM and Carnegie Mellon University. Accurately understanding human language is one of the keys to naturally interacting with any machine. While voice recognition technolgoy exists today, it is often inaccurate and requires the user to recite specific pre-determined commands in order to get a computer to perform a task.

But imagine if you could simply talk to your computer as if it were another person and it would, in response, do whatever you asked. Rather than opening a Web browser, typing “Google.com” into the address bar, and searching for “lasagna recipe”, and hunting through hundreds of Web pages for the best one, a computer could just be told, “Get me a good recipe for lasagna,” and it could bring you a high-rated recipe without showing you all of the steps in-between. This is one of the boundaries between the mouse/keyboard paradigm of human-computer interaction we know today and the “space age” interaction we so often see in science-fiction films.

Unfortunately, it is not known if the battle between humans and machines on the Jeopardy set will be aired, or even taped. My guess is that if the computer performs poorly, we won’t see much more on the subject until it is improved. But if Watson rises to the challenge and looks like it could beat Ken Jennings’ records, I sense a whole new series of “Are You Smarter than a Supercomputer?” reality shows to come.