Watson on Jeopardy: Round 2
“Don’t let the pod door hit you in the butt on the way out, Ken.”
No, IBM’s computer system didn’t actually reference 2001’s HAL 9000 on yesterday’s exhibition match on Jeopardy against two of the quiz show’s greatest human players.
But I think it would get the joke.
This is really extraordinary to watch. It’s a computer really understanding human language. That’s been one of the barriers that many people thought was still years (if not decades) away. Rosie, the robot maid on the Jetsons, was able to give out advice in a bit of a wisecracking way. Jane could talk to her the same way she could talk to her teenager Judy…maybe even more easily.
It’s easy to miss how important and groundbreaking this is. We’ve all seen computers in movies that could carry on a conversation, and it may be hard to realize this is really happening.
Language has always been one of the things that we human owned. We’ve tried to get some other animals to get it: sign language with chimps and gorillas, dogs who can bark “I love you.”
But this is different.
It’s not that Watson is coming up with new ideas…that is probably still ours for a while.
However, Watson is able to retrieve ideas.
To quote author John A. Keel:
“To H*ll with the answer! What’s the question?”
That’s from memory, so it may not be exactly right…but it’s been key on the show. Phrase the question properly, let Google search the web, and you are likely to get an answer to most of the questions on the show.
But the questions (well, answers, technically) aren’t written in a way to make it easy for a computer.
Watson clearly usually gets what the question implies, not just what it says.
This will, if it can be done on a commercial level, really change the way we interact with knowledge.
A doctor will be able to say, “What was that treatment that male doctor did with the elderly patient for alopecia? It was somewhere in Asia, and used some kind of herbal extract,” and a computer will be able to answer that.
You’ll be able to ask a website things like, “There’s this new book by a celebrity that’s been pretty controverial…it had something to with game show in the 1980s,” and the computer will be able to tell you what it is.
The first round was a tie between Watson and Brad Rutter, with Jennings considerably behind.
The second round was a slaughter.
Spoiler alert! Watson ended up with $35,734, Brad was $10,400, and Ken had $4,800.
It also seemed like we started to see some more “personality” from Watson. He said something like, “Let’s finish off the category.” He made really odd bets…his final Jeopardy bet was $947…causing Alex Trebek to comment, “Oh. you sneak.” He also really wasn’t sure on the final answer, so he did one of my pet peeves…put a bunch of question marks at the end (I call people who do that “serial puncs”). He was also wrong…but won very handily (can you win “handily” if you don’t have hands?).
In tonight’s final show of the series, they will play a whole game. Unfortunately for Ken, the scores from the first game and this second game will be added together…that’s going to make it an uphill climb.
However, if Watson delivers another brain-kicking like he did last night, it won’t make much difference.
Have you noticed I’ve started referring to Watson as he instead of it? 😉
While this is huge, it still isn’t the test I’ve always said we’ve needed…a computer that can play Password…are you listening, IBM?
Or maybe I should be suggesting it to Watson… 😉
This post by Bufo Calvin originally appeared in the The Measured Circle blog.