What does Watson know about birds? - Stephen Baker
Stephen Baker is author of Final Jeopardy. the definitive book on the IBM Watson computer that competed successfully on Jeopardy in early 2011.
I happened to see this heron in a Montclair pond a week ago, and it led me to wonder what my old friend Watson would make of that lovely water bird. Specifically, if Watson’s analysis indicates that a heron is a bird and it also has strong evidence that birds fly, would Watson be able to infer that a heron would fly?The answer to that is no. The reason is that Watson, for all of its achievements in Jeopardy, is incapable of generalizing, much less coming up with theories. Humans do this all the time. You see toddlers who mess up their irregular verbs, saying that Johnny “falled” or that Timmy “eated.” They’re creating a theory of language based on the patterns they’ve picked up. And when toddlers sees that robins and bluebirds and cardinals fly, they quickly generalize about birds. That’s a key aspect of human intelligence, and Watson doesn’t have it. Watson could find many references to herons, both as birds and flying animals, and it could correctly answer a question about what herons do with their wings. But that conclusion doesn’t inform any broader thinking on the subject. That’s not what it’s built for.So, as we look to the job market, humans who make a living by synthesizing information and coming up with theories are not likely to be displaced by computers anytime soon. Those who comb through data to find answers, by contrast, will be facing increasingly stiff (and tireless) competition.To be fair to Watson, I should add that its inability to generalize sometimes pays dividends. This is because generalizations usually have exceptions. If we come up with a theory that birds fly, penguins and ostriches can confound us. Watson, by contrast, would come across very little evidence of flying ostriches or penguins and—unburdened by theory—would sidestep that trap.It’s odd that I’m writing here about flying, because that’s the very metaphor that occurs to me. Our minds soar—and sometimes we lose sight of the reality on the ground. Watson never leaves the ground. It sifts ceaselessly, analyzing and crunches. Never distracted by ego, desire, theory, or hundreds of other human qualities, it just churns out the statistically most likely answers. In the cognitive realm, Watson’s our beast of burden.

 What does Watson know about birds? - Stephen Baker

Stephen Baker is author of Final Jeopardy. the definitive book on the IBM Watson computer that competed successfully on Jeopardy in early 2011.

I happened to see this heron in a Montclair pond a week ago, and it led me to wonder what my old friend Watson would make of that lovely water bird. Specifically, if Watson’s analysis indicates that a heron is a bird and it also has strong evidence that birds fly, would Watson be able to infer that a heron would fly?

The answer to that is no. The reason is that Watson, for all of its achievements in Jeopardy, is incapable of generalizing, much less coming up with theories. Humans do this all the time. You see toddlers who mess up their irregular verbs, saying that Johnny “falled” or that Timmy “eated.” They’re creating a theory of language based on the patterns they’ve picked up. And when toddlers sees that robins and bluebirds and cardinals fly, they quickly generalize about birds. 

That’s a key aspect of human intelligence, and Watson doesn’t have it. Watson could find many references to herons, both as birds and flying animals, and it could correctly answer a question about what herons do with their wings. But that conclusion doesn’t inform any broader thinking on the subject. That’s not what it’s built for.

So, as we look to the job market, humans who make a living by synthesizing information and coming up with theories are not likely to be displaced by computers anytime soon. Those who comb through data to find answers, by contrast, will be facing increasingly stiff (and tireless) competition.

To be fair to Watson, I should add that its inability to generalize sometimes pays dividends. This is because generalizations usually have exceptions. If we come up with a theory that birds fly, penguins and ostriches can confound us. Watson, by contrast, would come across very little evidence of flying ostriches or penguins and—unburdened by theory—would sidestep that trap.

It’s odd that I’m writing here about flying, because that’s the very metaphor that occurs to me. Our minds soar—and sometimes we lose sight of the reality on the ground. Watson never leaves the ground. It sifts ceaselessly, analyzing and crunches. Never distracted by ego, desire, theory, or hundreds of other human qualities, it just churns out the statistically most likely answers. In the cognitive realm, Watson’s our beast of burden.


Recent comments

Blog comments powered by Disqus