It was the stuff of science fiction.
The grandmaster of the television game show Jeopardy!, Ken Jennings, resigning himself to a crushing defeat at the hands of an IBM supercomputer, wrote "I, for one, welcome our new computer overlords." While IBM's Watson may not have correlated the statement to a 1994 episode of The Simpsons, the computer had handily demonstrated that in 2011, machine could beat man at a complex quiz show.
Watson, as you may have heard by now, was designed by IBM researchers as one of the company's frequent "grand challenges," efforts designed to push the envelope of science and computing technology. In 2011, IBM is celebrating its centennial year, and the Watson effort demonstrated just how far computing technology has come during that century.
In my day job, I am an IBM executive responsible for a software product designed to help people share information and collaborate. While not directly connected to the team of IBM researchers who conceived of and built Watson, the three-year project has become a source of much company pride.
The Jeopardy! series that featured Watson versus the top two human players in the history of the game resulted in the best ratings for the show in years. IBMers celebrated at viewing events all over the country, including a jam-packed auditorium at the Century Theatre Cinema in Chicago.
I had the opportunity to learn more about Watson during a conference a few weeks before the TV show. An IBM researcher told an audience of 5,000 about the project, how they had taught the computer to play Jeopardy! and the results. Some of what he described was quite surprising. For example, Watson was not connected to the Internet -- it asked Jeopardy! questions by analyzing data, not searching.
Another surprising aspect of Watson's play was that it had no pre-built way to dissect the answers given as Jeopardy! clues. Each and every time the computer was fed an answer, it had to analyze the words, pattern, potential double meanings, category of the game and many other aspects in order to figure out what bit of data was being sought. It had to finish all of that, and have at least 50 percent certainty of the answer, before buzzing in. Further, Watson had to buzz in using the exact same method as the human players: pressing a button wired directly to the computer.
The researcher's presentation was simply laying the foundation for the main attraction of the conference session -- a live Jeopardy! round, featuring Watson versus conference attendees. While Alex Trebek wasn't able to join us, actor Todd Alan Crain played host, as he had done for many months of Watson practice rounds. Crain injected side-splitting humor into moderating what he called "computer versus meatballs," playing off Watson's impressive skills and occasional wide misses. The human contestants, one of whom I had nominated knowing that she would at least put up a good fight, never had a chance, and I knew then that the actual TV show would result in an equally impressive slaughter.
With the Jeopardy tournament complete, IBM's Watson project is more than just a marketing effort. The technology invented to interact with humans in a human way has wide applicability to problems we are trying to solve today.
Writing in USA Today, Yong Suh of John Hopkins University stated, "Watson has the potential of addressing two pressing problems in health care today: deaths due to medical errors and shortage of physicians." Eric Nyberg of Carnegie Mellon University told ABC News: "I think the logical next stop beyond Watson is going to be systems that can advise you on selecting certain kinds of products that meet your personal needs."
Personally, this meatbag also welcomes our new computer overlords. What I saw in Watson was a computer not afraid to make mistakes and learn from them. It won big, not for being a perfectionist, but for being a calculated risk-taker. Isn't that a lesson we can all learn from?