Watson’s victory impressive, not revolutionary


By Calder Phillips-Grafflin

Jeopardy! has stood for years as one of the most public intelligence contests between humans. For the developers at IBM, Jeopardy! was the perfect target; not only would it offer an extremely public demonstration of their new technology, it would allow them to demonstrate the incredible advances in computing power.

Even after chess fell to computers with IBM’s Deep Blue in 1997, questions remained about when computers would be able to compete with real human “intelligence.” Deep Blue’s victory wasn’t a product of ‘deep’ intellect; raw computational power allowed it to analyze all possible moves and act accordingly.

IBM asserts that Watson’s victory in Jeopardy! is a revolutionary advance in computing, but is it? All the victory in Jeopardy! proves is that a powerful enough computer with enough available information can search for answers more quickly than a human.

Watson’s claimed strongpoint is its natural language processing abilities, but its performance on Jeopardy! leaves many questions about that. Watson’s answers were much less a subtle understanding of the question and much more an amalgamation of the results you’d get it you used a normal search engine on the ‘keywords’ of the question. Natural language processing is certainly a difficult task, but it would be unwise for IBM to rest on its laurels with Watson.

IBM believes technology like Watson will be used for medical diagnostics and other fields where  large amounts of data must be processed by people who may not know what they’re looking for. However, Watson’s performance on Jeopardy! really doesn’t carry over to that sort of cooperative work with humans; if the humans depending on Watson aren’t smart enough to convert their questions into easy-to-search keywords, then they won’t be smart enough to rule out the incorrect answers Watson will inevitably provide.

Watson’s victory at Jeopardy! is certainly historic, and IBM’s engineers can be suitably proud of their work. We should be careful, though, before we believe systems like Watson are ‘intelligent’ enough to make decisions that affect people in the real world.


Leave a Reply