I have been using computers since 1982 and still recall a simple artificial intelligence program that ran on my Commodore 64. I was mindful of that this week as I watched IBM’s computer—named “Watson”—compete on “Jeopardy!” against two human “Jeopardy!” champions, Ken Jennings and Brad Rutter.
I was quite impressed by the computer, which easily beat the two most successful “Jeopardy!” players of all time. Some might see this victory of a machine over human champions as a cause for concern, perhaps thinking of Hal in 2001: A Space Odyssey. While there may be some cause for concern, it is not of the “Space Odyssey” variety. In truth, humans are firmly in control of any dangers.
Think about it: Technically, we did not see a machine beat human champions. We saw a team of IBM computer programmers and engineers—all humans—beat two human champions. Calling the computer “Watson,” we have a tendency to personify and humanize what is only a machine. Watson is not a conscious being. When watching Watson play this game, we witnessed a machine that follows rules that were programmed by people. The set of rules that Watson follows is incredibly complex, and the speed at which the rules are executed is nothing short of awesome. However, we did not witness a machine that thinks for itself.
Watson did not create any new knowledge during the exhibition. Indeed, it did not appear to me that Watson had access to more trivia than either of the human competitors. What was amazing about Watson’s programming was two things: First, there was Watson’s ability to search for the correct answer given the sophisticated questions posed. “Jeopardy!” questions often require contestants to make use of things such as humor and idioms, for example. To search for the answer to the questions thus required a sophisticated analysis of language. The second amazing thing about Watson’s programming was the speed with which the decisions could be made sorting through vast quantities of data faster than the fastest humans.
During the three nights of the show, viewers were given glimpses of how the computer’s programming skills might be put to practical use. One application I found particularly intriguing was medicine. A physician might soon be able to enter a patient’s medical data into a computer like Watson. This computer would be able to rapidly compare a patient’s data to all known medical resources. No one person can keep up with all new medical research. I know people with unexplained medical problems. A diagnostic tool like Watson would allow my friends’ physicians to essentially collaborate with every medical researcher in the world. More diagnostic mysteries could be solved.
Here is where the danger of a Watson-like computer lies: The final “Jeopardy!” question of the first game in the two-game match went something like this: What is the name of the U.S. city whose largest airport is named for a WWII hero and its second largest is named for a WWII battle? The answer was Chicago. Watson’s program came up with Toronto. Toronto? A U.S. city? A Watson-like computer has great promise for helping physicians, but physicians must still use their human judgment. The medical equivalent of a wrong answer to a trivia question could be deadly.
If we get to the point where people accept—or are required by law to accept—the decisions reached by our decision-making machines, bad things will happen. We have all heard the occasional story of the person who does something crazy like driving the wrong way down a one way street, or through a closed construction zone, because it is what the GPS said to do. As long as we remember that even amazing machines are still machines, and that human judgment is still needed, we have nothing to fear from Watson or his progeny. Education and human judgment will always be needed. The computers of tomorrow hold great promise as tools humans can use to improve our lives. Have no fear; we are still the masters.