• U.S.

10 Questions for Watson’s Human

4 minute read
David Ferrucci

Why aren’t you letting Watson speak for himself today?
Watson is trained to answer questions for Jeopardy! It’s not an interactive dialogue system, so it can’t conduct its own interviews. You can imagine giving it information so it could answer [impromptu] questions, but it would still be responding only from content it’s been given and analyzed.

What about aspirational questions? Could Watson respond if it was asked, “What do you want to be doing in five years?”
It would have to be given some information about itself as an entity in the world in order to do that. But that’s not IBM’s focus. Watson is designed to deeply analyze existing content and help people make decisions, not to be an independent entity.

(See pictures of Watson and its creators.)

People have said that Watson functions at the level of a precocious child. Do you agree?
When an artificial-intelligence system can perform a particular task, we have to be careful not to look beyond that task. Take Deep Blue, the [IBM] system that beat a grand-champion chess master. Few adults are smart enough to do that. But Deep Blue wasn’t a system that could go off and even approach a child’s ability to do language, to move, to think, to interact.

And yet Watson does understand natural language.
But only in a way that we call statistical machine learning. It gives you the answer that makes sense to you, but it doesn’t mean anything to the computer.

Do you worry that Watson could be misused–to game the stock market, say?
A chair can be misused, so it’s hard to answer that in the abstract. Watson’s not being put in control of anything. It’s not being hooked up to an environment where it’s independently making decisions. This just strikes me as the least thing to worry about here.

(See the top 10 man vs. machine moments.)

IBM talks about Watson’s being used to diagnose diseases. Can a machine make intuitive leaps like the ones Dr. House makes on the TV show?
That’s a tough question, because I wonder what intuition really is. It’s probably a process like connecting the logical dots, but we call it intuition simply because we’re not fully conscious of the process.

If I tell Watson a joke, will it get it? Could it tell me one?
One of the things we programmed it to do was recognize what humans would consider puns. It looks for word associations, for synonymy, for “sounds like.” But does the computer appreciate the humor? No, it doesn’t.

So no chance Watson could be an artist or poet either?
Actually, it could be programmed to have those features. You could train Watson to recognize modern art or classical art, symmetry, shape and color. You could tell it, This is good art, this is bad art. But again, that’s all.

Will Watson or its kin ever come to work in our homes?
There’s definite potential there. With speech recognition, you could talk to your TV, to your mobile devices, to your computer without having to type in a bunch of commands. I like to think about Star Trek, about a computer you could talk to while you’re walking around.

Artificial-intelligence pioneer Marvin Minsky once said that consciousness is like a simple memory chip in the brain. Is a conscious computer possible?
Over beers, I could talk for hours and hours about that. But it’s a question best left to a philosophical treatment.

Read TIME’s 1996 cover story “Can Machines Think?”

See TIME’s computer covers.

More Must-Reads from TIME

Contact us at letters@time.com