Hal, make room for Watson.
When it defeated two of the all-time champions of the television game Jeopardy this week, the IBM computer named Watson joined Hal 9000 in the ranks of intelligent computers in the popular culture.
The characterization of Hal as portrayed in the movie “2001,” embodied the dream of computer scientists 50 years ago to create autonomous artificial intelligence with characteristics comparable to those of the human brain: a field generally termed AI.
Watson, on the other hand, is the realization of the more modest goals of today’s computer scientists to develop systems that are complimentary to and that enhance human intelligence, an approach generally termed intelligence augmentation or IA. Google, which allows us to search and find valuable information among the millions of web pages on the Internet, is one familiar IA system.
“Very few researchers are focused on developing systems with autonomous intelligence,” commented Douglas Fisher, associate professor of computer science and computer engineering at Vanderbilt, who specializes in machine learning and artificial intelligence. “Today, most people are working on systems designed to enrich human experience and capabilities”
In that respect, Fisher considers Watson to be “a substantive advance in the state-of-the-art.” In particular, he can see such a system applied to the area of medical care. If it was filled with all the latest medical information, physicians could type in a list of symptoms and get back a series of suggested diagnoses and treatments, some of which they might not have thought of on their own.
Of course, before such a system is put into doctors’ hands, Fisher emphasized, IBM must come up with a way to handle Watson’s habit of occasionally producing silly answers, such as responding to a question in the US City category with the name of a Canadian city.
Sean Polyn is another campus scientist who was impressed by Watson’s performance. As an assistant professor of psychology and director of the Computational Memory Laboratory, Polyn studies the nature of human memory.
What struck Polyn was the vast difference between the way in which Watson and the human mind work. “A lot of what the computer is doing is brute force searching and correlating information in a huge database; including what IBM calls embarrassingly parallel processing, which consists of running a number of unrelated searches and statistical analyses at the same time.”
The human brain handles information in a totally different fashion. Within the brain information is stored within a highly interconnected network.
Take the case of celebrities. Polyn’s research shows that when asked to memorize a list of celebrities, the brain automatically categorizes them in as many ways as it can. It puts actors in one group and athletes in a different group. If two of the celebrities are linked romantically, they are grouped together. And so on.
“It is truly parallel. Drip in a little information and it spreads throughout the network activating all related memories,” he explained.
As a result, the brain can recall information extremely quickly despite the fact that the speed signals propagate in neurons is almost glacial compared to that of electrical circuits (25 meters per second compared with more than 200 million meters per second).
Ironically, it seemed that Watson’s biggest advantage over its human competitors was not in coming up with the correct answers to the questions, it was pushing the buzzer first. The buzzers are not activated until Alex Trebek finishes reading the question. If a contestant pushes the button too early, his button is locked out for a fraction of a second. Jeopardy officials insisted that Watson must also push a physical button, but IBM engineers appear to have given Watson the capability to respond in milliseconds after the buttons are activated, making it very difficult for the human contestants to beat it to the punch!