IBM has created a supercomputer named Watson who just beat the two greatest Jeopardy! champions of all time in a warmup match. PopSci has the story:
Watson, named after IBM’s founder, is one epic supercomputer. To handle the formidable task that competing on Jeopardy! presents, IBM spent years constructing a computer with 2,800 Power7 cores. That power is absolutely necessary—a single-core CPU, like in many modern computers, takes about two hours to come up with an answer to a standard Jeopardy! question, rather than the three-second average Watson currently boasts.
A lot of the challenge in creating an algorithm that can answer Jeopardy! questions lies in the questions themselves—the language used in these questions is hardly ever simple, often incorporating wordplay, riddles, and irony—but there’s an additional problem in the addition of risk. In a split-second, a competitor must assess confidence in the question, weigh that confidence against the penalty of getting it wrong, and decide of the question is worth answering based on those factors. That’s an intuitive effort for a human, but Watson had to be programmed with some incredibly complex reasoning to be able to do the same thing.
Go read the whole thing. And you should note that Watson isn’t connected to the internet, which means “he” is kind of doing this with one hand tied behind “his” back. This is so much scarier than Kasparov Vs. Deep Blue. Chess, basically, is a numbers game. Jeopardy! is a language game. I bet we’re not that far away from a computer writing a John Grisham-like bestselling thriller.
Humanity doomed. Film at 11.
2 humans on 1 computer isn’t really fair. Some questions will be generally easier for humans, others will be easier for computers. With two humans, there’s only half as many such questiosn for each to answer. It’s like having two candidates with similar views each lose to a third candidate whose views aren’t as popular as either of the other two, but doesn’t have their votes split.
I thought we already had a computer writing Grisham’s thrillers.
I’m not really impressed. Sure, a computer can be programmed to do any one thing extremely efficiently. This is not news. When you show me a computer that can win a chess, win at jeopardy, hold a conversation without it being obvious that you’re conversing with a computer and train a dog then I’ll be impressed. Until then it’s just number crunching and word searching.
Put this computer onto the Hulk robot body. You have a Skynet-type killing machine that can make ironic James Bond-styled comments over your crushed bones.
What is “The end of the world as we know it?” Alex.
Are you telling me that Dan Brown isn’t a computer program…
I need proof of this.
Also, I think Stephen King has a computer program where you input various elements and a page length and it creates the book for you.
That which is “intuitive” for us humans is just the application of programming we’re not consciously aware of.
@4: excellent.
AI is this generations flying car. Or rocket pack.
They should use an AI to finish the Wheel of Time series.
@6:
Are you telling me that Dan Brown isn’t a computer program…
I need proof of this.
Your average Commodore 64 could write better shit than that hack with one byte tied behind its back.
So I’m the only one who actually thinks this is pretty cool and impressive? I guess so. Everyone else must be a supercomputer specialist in AI research, completely jaded, or most likely, as Catalina puts it, “This is Seattle, where everyone thinks they are smarter than everyone else.”
iwatson
i want one
Anyone who’s watched a fair amount of Jeopardy knows that knowing the answers is the smallest part of success. It’s hitting that damn buzzer, and then not blanking out. It’s not a game of knowledge, it’s a game of psychology.
When I watch, I typically know over 90% of the answers, but I doubt I could play the game well enough to pass the first round (especially because half the time the first word out of my mouth would be “FUUUUCK”, which is against the rules).
Since the computer has no psychology, it has a huge advantage.
It should read, “….. takes about two hours to come up with a question to a standard Jeopardy! answer, rather than the three-second average Watson currently boasts.” and “…..A lot of the challenge in creating an algorithm that can produce questions for Jeopardy! answers lies in the answers themselves—the language used in these answers is hardly ever simple, often incorporating wordplay, riddles, and irony…..”
The writers at Popular Science should watch more Jeopardy!
It’s doubtful that a live internet connection would help.
I’m also impressed, and I used to do AI research. In terms of language though, there’s a big leap from “What is X?” to writing paragraphs, let alone having a conversation. A big part of language is pragmatics: figuring out what the conversation is about and where it’s going, and then using your turn to get it there; this extremely complex part is removed in a game like Jeopardy!
Still, a journey of a thousand miles etc. “Just number-crunching and word-searching” is a hard problem too, if you have to do it competitively and in real time.
There was a much better and longer article about this in nytimes magazine in June 2010.
http://www.nytimes.com/2010/06/20/magazi…
PopSci is shit.
Oh please. My dog wrote a John Grisham-like bestselling thriller. And she’s been dead for two years.
And Will in Seattle has been automated for at least as long, and he’s apparently a viable candidate for something.
We already live in the future.
@14–that’s right, a big part of Jeopardy is simple reaction time, hitting the buzzer first.
Does the computer hit the buzzer with its index finger or thumb?
There’s no way this thing beats Frank Spangenberg.
John Grisham or John Updike, there’s not much terror in that.
Besides, the humans were holding back. The top prize is $1 million, 2nd prize only $300,000. Would you tip your hand before the big match?
Watson’s toast, his chips just don’t feel it yet.