The Who2 Blog

Watson vs. Humans: How’d We Do?

How did we do with our predictions last week about the big Watson vs. Humans showdown on Jeopardy?  Let’s look.

Totally Wrong

Our prediction on final scores: $49,800 / 36,000 / 28,300, with Watson in third.

Actual result: $77,147 / 24,000 / 21,600, with Watson in first.

Our prediction: “I see either Brad Rutter or Ken Jennings winning the match through the all-too-human stratagems of crazy aggressive betting and random moments of genius that turn the tide.”

Actual result: Watson not only won, but dominated. All-too-human stratagems were useless.

Partly Right

Our prediction on the buzzing-in breakdown: 41% / 32% / 27%, with Watson leading.

Actual result: 57% / 26% / 20%, with Watson leading.  He dominated even more than expected.

Our prediction on Final Jeopardy: Watson may have trouble with the quirkier questions and may be tripped up by human wagering.

Actual result: He had no trouble with wagering, but Watson did indeed have trouble with the quirkier questions.  (“What is Toronto?????“)

Totally Right

Our prediction: “The buzzer will play a colossal hidden role in this match, as it always does.”

Actual result: The buzzer played a colossal role, as it always does. Watson’s incredible run in Double Jeopardy on the first game — 23 right answers, to 3 for Ken and 2 for Brad — showed that when he wanted to get in first, he did.

Probably there is no “fix” for that.  If you’re going to have the computer buzzing electronically, it’s just going to be faster.

So this match didn’t show me that a computer is “smarter” than humans at Jeopardy! — just that it can be AS smart as humans and much faster when wired to the buzzer.

That shouldn’t sound like sour grapes. Watson was amazing! And Jennings and Rutter said all the right things

But I’ll bet privately they’re muttering to friends, or to themselves, that with a more equitable buzzer system they could still take down the computer.

Related Biography

Share this: