Archive for February 17th, 2011

Capsule Tech: Watson

English is an interesting thing.  It’s probably one of the hardest second languages to pick up, because of the massive numbers of irregularly conjugating verbs and bizarre, at times arbitrary, pronunciations.  And that’s if you’re human.  It can be that hard to pick up even if you already understand the general concepts of irony, sarcasm, puns, subtext, or any of the other things that go into communications and seem so natural to us as humans.

So what do you do when trying to learn English as a second language when your first language is machine code?

I watched all three of the Watson episodes of Jeopardy this week, and what I’ve seen I’m still processing.  It’s something so completely new and different that was on display that it felt like watching some step forward in history.  Will these episodes be our generation’s moon landing television event?  No.  Do I think that the episodes will have some long lasting significance?  Yes.  But when watching them there was something oddly familiar about what I was watching.

See, what makes Watson so amazing is that it’s easy to see it as not amazing.  On one hand it did what three people do five times a week: answer oddly phrase trivia questions that require not just a vast knowledge but also the ability to understand how the show uses word games and puns.  On the other hand it does what we see in science fiction all the time: it seamlessly parsed naturally spoken English and gave the anticipate responses.

But Watson isn’t a human.  And those three episodes were not science fiction.  And it’s oddly necessary to specify those, because what we did see on display was the first step towards a computer acting a little more human, or a little more like the computers that Star Trek has promised us.  The kinds of computers that run the holodeck, and can create complex simulations from just a few simple spoken statements.

Perhaps those will always be fiction, but this is still an amazing step in that direction.

Was Watson perfect?  No.  Any human playing Jeopardy would immediately not even consider Toronto as an answer in a category called “US Cities”.  And its knowledge seemed somewhat lacking when it came to the fine arts.  But these feel like such minor issues compared to how far the Watson team has come in creating a system that is capable of understanding human speech well enough, and responding quickly enough, that it took down the two best players Jeopardy has ever seen.

What was televised in syndication this week was a marvel, and I hope that anyone with the slightest interest in where technology is going to go over the next twenty years was watching closely.  Because what we saw this week?  Was the future.

, , , , , ,

1 Comment

%d bloggers like this: