A Computer Is the Champ — at Jeopardy

April 2011

Be humbled, be very humbled. First computers beat humans at checkers, then chess, and now the TV trivia show Jeopardy.

You probably heard that in February an IBM computer named Watson took on the two all-time most successful Jeopardy champs for three rounds and came out victorious: $77,000 for Watson, $24,000 for Ken Jennings, and $21,000 for Brad Rutter.

Though it’s a bit of a misnomer to call it a computer. More accurately, it was a roomful of 90 32-core IBM Power 750 Express servers working in parallel. Watson was fed the questions via text, and his mechanical finger pressed the buzzer when he knew the answer — which he spoke out clearly.

In round two Watson was the first to hit the buzzer on 25 out of 30 questions, getting all but one of them right.

So what? We already know that computers can quickly search massive databases. What’s new about this? The challenge was to understand natural human language.

Think about it. How often do your Google searches get exactly the right answer the first time? What if you could just speak out what you wanted to know, and 24 out of 25 times Google would come back with just what you asked for?

It’s been very difficult to get to the stage that computers could understand natural language, and IBM selected Jeopardy as the test of its new system because that quiz show typically uses a lot of word play in the questions.

Consider this one that Watson didn’t get in a subsequent round against some of the brainier members of Congress: the category was Presidential Rhyme Time, the statement was “Herbert’s military strategy,” and the correct response was, “What were Hoover’s maneuvers?”

Not easy, yet Congressman Rush Holt got it right (actually defeating Watson in their one-round matchup.) And it’s the sort of complicated question that Watson quite often got correct.

So why on earth would IBM dedicate 25 engineers and four years to creating a computer that could win Jeopardy? Obviously, they have other plans for this new software that’s adept at understanding natural language. They’re already marketing a system for doctors and nurses that will assist with diagnoses.

Think of it, a huge database of medical information that can give the latest updates regarding your maladies. The fact is, it’s impossible for medical professionals to keep up with all the latest scientific literature and state-of-the art practices and treatments. And often they make do with what they learned years earlier in medical school and what they might have picked up in their continuing medical education.

But now they can have a powerful assistant. And that was the point of Jeopardy: to capture the attention of the nation and show how computers can now be invaluable expert assistants in a variety of professions.

All of which brings me to my iPhone. What’s next for this gizmo? Thanks to my part-time job with iPhone Life magazine, I enjoyed an all-expenses-paid trip to Macworld in late January. There were two highlights: hearing a talk by computer legend Bill Atkinson, one of the early employees of Apple Computer, and getting to interview the director of mobile development for Nuance, creator of the leading voice recognition software for desktop computers and mobile devices.

Aktinson breezed through the history of computing, showing the remarkable trajectory from the first computer to today’s phenoms. And then he projected into the future.

What will your smartphone be like in the coming years? His answer, vividly described, is a Watson-like personal assistant: an earpiece with a microphone, speaker, and camera connected to the huge trove of information on the Internet. Have a question? Just ask.

Your personal assistant will record and remember every detail of your personal life. Someone walking toward you whose name you can’t remember? The camera in your earpiece recognizes the face, speaks the name into your ear, and reminds you where you met the person.

The Nuance exec told me that many companies are now starting to use speech recognition in the iPhone apps; that one company (Siri) has built natural language understanding into its app; and that Apple bought that company over a year earlier. Watson is coming to an iPhone near you.

All this strongly suggests that the “singularity” is on the horizon, the time described by legendary inventor Raymond Kurzweil when technological change becomes so rapid and profound, it represents a rupture in the fabric of human history. Kurzweil says it will be the year 2045, when computing power surpasses the brainpower equivalent to that of all human brains combined.

I’m starting to believe him.

This month’s hot tips:

Serif offers free website design software. Excellent and urgent article in the New Yori Times about serious WiFi hotspot security issues.


© 2011 by Jim Karpen, Ph.D.

E-mail Jim Karpen