RSS

Jeopardy’s Watson computer still needs humans

15 Feb

If you didn’t tune in to Jeopardy last night to see IBM’s “Watson” supercomputer take on a pair of the show’s past human uber-champions, you have another chance tonight in part two of what is history in the making.

The first half of last night’s episode was incredible to watch – the computer literally kicked the crap out of human competitors Ken Jennings (who once had a 74-game winning streak) and Brad Rutter during the first half of the show, only to come back down to earth during the second half. After breezing through the easy questions, Watson actually fielded a number of wrong answers – gasp! Here’s a promo video explaining what it’s all about:

The cynic might think this whole computer-gimmick-on-Jeopardy thing is just a big commercial for IBM, and to some extent that’s correct. But there really is something far more important going on here. Just as Art Linkletter used the massive UNIVAC computer on his show in the 1950s to pair up men and women, so too is Watson intended to demonstrate just what is possible with today’s computing power. As the IBM engineers say in the video, Watson can ultimately be applied to all sorts of real-world applications, such as diagnosing patients. Indeed, the military is working on a similar program called the Personalized Assistant that Learns (PAL), which can interpret data and suggest courses of action to its human operators.

Watson and PAL are the products of Moore’s Law, where the processing power of computers doubles roughly every 18 months – a phenomenon that has essentially been true since the invention of the transistor back in the 1940s. As the Watson/Jeopardy hype built over the weekend, many people joked about the coming robot apocalypse – a time long foreseen in science-fiction when the computers become smarter than the humans and eventually take over the world.

The world domination part is unlikely to happen for a host of reasons, but computers surpassing us in intelligence is inevitable. Moore’s Law, or what futurist Ray Kurzweil calls the law of accelerating returns, essentially guarantees it.

Coincidentally, I happened to recently speak with Jim Osborn, the executive director of the Quality of Life Technology Center at Pittsburgh’s Carnegie Mellon University, a hotspot of robotics and artificial intelligence research. We chatted about a story on robots that I’ve been working on and, of course, the topic of robotic AI came up. He said that thanks to Moore’s Law and the law of accelerating returns, robotics problems – things like getting a robot to walk properly or grip an item correctly – are being solved with amazing frequency. However, despite that, the role of humans in determining the intelligence level of a robot is still extremely important because we’re the ones the give it the tools it needs to learn.

“It’s not just a matter of waiting for a faster computer to do the same algorithm just that much quicker, we still need fundamental improvements in the algorithms,” he said. “You don’t want to run a crummy algorithm on a better computer.”

In any event, if you want to see something amazing, tune in to see Watson in action on Jeopardy tonight.

About these ads
 
Comments Off

Posted by on February 15, 2011 in artificial intelligence, DARPA, robots

 

Comments are closed.

 
Follow

Get every new post delivered to your Inbox.

Join 5,218 other followers

%d bloggers like this: