IBM's Watson's ability to converse is a huge advance for AI research

Jeopardy-playing computer called by some the biggest computing advance in decades

IBM's Jeopardy playing supercomputer has been touted by some observers to be one of the biggest computing advancements in the past several decades.

The supercomputer, dubbed Watson, owes that significance to its ability to deliver more than calculations and documents. It can answer verbal questions posed by humans.

That ability, says IBM researchers and industry analysts, makes this machine more equipped than any before it to organize "thoughts" and verbally converse with people.

"I would say it's the largest computing advance of this century," said Richard Doherty, research director at Envisioneering Group. "I've been in computing since 1973 and followed technology before that, and this is the largest advancement in decades. This isn't an iPad. To reach [a computer] conversationally and have it respond with knowledgeable answers is a sea change in computing."

Watson's first appearance on the long-running Jeopardy game show aired the first episode of its man-vs-machine competition last night. Watson faced two Jeopardy champions - Brat Rutter and Ken Jennings.

Half of the first game aired Monday night, and Watson and Rutter tied for the top spot with $5,000. Jennings finished the day in third place with $2,000.

For IBM, the Jeopardy appearance by Watson represents the next stage in the long effort to develop a computer that can mimic human intelligence. Watson is a study in advancements in artificial intelligence and natural language processing that overcome the formidable challenge of conveying the same information in many different ways.

IBM scientists spent four years building a computer system that could rival a human in answering questions posed in natural language. Watson is based on IBM's Power 7 server hardware.

"I think Watson has the potential to transform the way people interact with computers ," said Jennifer Chu-Carroll, an IBM researcher working on the project. "Watson is a significant step, allowing people to interact with a computer as they would a human being. Watson doesn't give you a list of documents to go through but gives the user an answer."

IBM made this happen by going back to square one and analyzing the nature of questions and answers, said Charles King, an analyst at Pund-IT.

"Computers tend to be very good at linear and logical processes," explained King. "With Watson, researchers basically discovered that the entire question-and-answer process is riddled with ambiguity. Watson is designed to take a single question, run different algorithms simultaneously, produce multiple answers and then rank the likelihood of correctness, and do it all in a few seconds."

According to Doherty, another significant aspect of Watson's technology is that IBM's scientists gave the machine "confidence."

"I'd never before heard the word confidence applied to a machine," said Doherty. "That's something we all need in our daily lives. We've never heard of this being applied to artificial intelligence before. In Jeopardy, if you're not confident enough, you don't buzz in. I can afford to buzz in (if my confidence level is) above 81%. I can't afford to buzz in if I'm only 79% sure."

King said the computing advances in Watson are making computer scientists re-evaluate artificial intelligence and what can be done with it.

"I would think that across the world, computer scientists are looking at this and thinking of what the next generation of AI will be," he added. "We will be getting closer and closer to the measure of AI success that was imagined in the first Star Trek series -- a computer that could be queried without using a keyboard device, that would come back with the right answer in a few seconds."

So what's next for artificial intelligence?

Chu-Carroll said computers may be learning to use actual common sense over the next five to 10 years.

"I would like to see computers be more able to understand the world much better," she added. "We're trying to code rules of common sense knowledge. Rain is wet. Dogs are animals. We'd like to address this in a more scalable way. Maybe computers could learn some of this from reading books and use them to augment their knowledge."

Doherty expects a renewed, invigorating race among artificial intelligence researchers.

"It's going to turn up the wick on research at Intel, at HP, at Oracle," said Doherty. "[Watson] is great for the market and for society. Watson will accelerate a new race. Five years from now, we'll see [an AI] revolution and a variety of new solutions."

Sharon Gaudin covers the Internet and Web 2.0, emerging technologies, and desktop and laptop chips for Computerworld. Follow Sharon on Twitter at @sgaudin , or subscribe to Sharon's RSS feed . Her e-mail address is sgaudin@computerworld.com .

Read more about mainframes and supercomputers in Computerworld's Mainframes and Supercomputers Topic Center.

This story, "IBM's Watson's ability to converse is a huge advance for AI research" was originally published by Computerworld.

Insider: How the basic tech behind the Internet works
Join the discussion
Be the first to comment on this article. Our Commenting Policies