PalmPilot Creator Models Computer on Brain
RENEE MONTAGNE, host:
Mondays, we focus on technology, and today, the effort to replicate the technology we were born with.
(Soundbite of music)
The man whose company developed a smart phone wants to make computers even smarter. Jeff Hawkins' products include the PalmPilot and the Treo phone. One of his other companies has been trying to produce artificial intelligence. He starts by studying the human brain.
Mr. JEFF HAWKINS (Founder of Palm Computing, Handspring, Numenta): The most interesting part, from an intelligence point of view, is the near cortex. This is the big wrinkly thing on top. If you take the skull off, you see the wrinkly thing on top.
STEVE INSKEEP, host:
That wrinkly thing can perform tasks that even the smartest computers cannot. Anybody whose computer can copy brain functions will be doing big business, and Hawkins' company is one of many that are trying.
Mr. HAWKINS: A scientist I know has proposed that the grand challenge, the million-dollar prize ought to be a machine that can tell two objects apart visually; literally, cats from dogs.
That's how far we are from doing what humans can do today. So, we're building a vision system that we believe will perform quite well, and it'll be very much like a human vision system that you'd be able to show it pictures of things in any sort of form and variation, and it'll say, and I know what it is. It'll very instantly say, oh I know, that's a cat, that's a dog, that's a car, that's a refrigerator, or whatever.
INSKEEP: Is a prototype already built?
Mr. HAWKINS: We built a small-scale prototype. It's not a practical system, in the sense it only recognizes line drawings, and it only recognizes 50 different objects. But already, I believe, and I think other people would agree, that it's pretty impressive.
An average person on the street might look at it and say, gee, well you're just recognizing a line drawing of a helicopter versus a dog, and they say, isn't that easy? Well, it is easy for humans, but it's actually a major advance in terms of technology.
INSKEEP: Does the computer ever have that very human moment of self-doubt, where it says, “Cat, no, wait, wait, wait dog! No! Cat!”
Mr. HAWKINS: In some sense. You know, first of all, these machines are not talking, or anything like that. The way they tell you the answers, they give you a little chart, we put on the a display of the computer that says, the probability of each of the things it knows. And so, when it's confused, it's basically showing a lot of probabilities that are equally high. So it says, well, a 30 percent cat, 30 percent dog, 30 percent table, or something like that. And that's how it's saying it's confused. It doesn't talk in the way a human would do.
INSKEEP: So, let's say that you get this prototype, which frequently recognizes line drawings of real creatures and objects, and you develop it to the next step, where it cold see a real cat, or a real dog, or a real person. What are the practical applications of that?
Mr. HAWKINS: Think about all the photos that you have on your computer, for example. If you want to find pictures of your, you know, Aunt Beatrice or something, there's no way the computer can do that. Someone has to go in and what we call “tag” those pictures. If you search for images on Google, the way you search for it is that someone has sit, and written in the name or description of each picture out there.
INSKEEP: They'll search the photo captions on the picture, basically.
Mr. HAWKINS: That's right. Yes, but a human had to enter those captions. There's no computer today that can look at a picture and say, “Oh, this is a picture of a boat. And this is a picture of a radio, and this is a picture of a house.” It just doesn't exist today.
And so, this would be a very practical thing just for personal, use or for search engines and so on. But it's really just the tip of the iceberg. If you understand what the technology is doing, it can be applied to many, many different types of problems.
INSKEEP: Hm. Such as?
Mr. HAWKINS: We've been approached by numerous car manufacturers, about a half-dozen of them, actually, who have research organizations, and they have cars which have sensors on them. They can easily put sensors around the car, like the new ones they have for backing up, or cameras, and so on. But they really, the car doesn't really know what's going on around it, or know what's going on inside of it. And you, if you were in a passenger seat, and you saw someone was very drowsy, you would be concerned about it. They would like the car to be able to do the same thing. The car would say, look, the driver's drowsy. Or, the car to be able to say, you know, there's an accident up ahead, I really ought to slow down, or someone's merging, this kind of thing.
INSKEEP: Can you tell me what pulled you, personally, from working on smartphones and devices like that, to this much broader and deeper question?
Mr. HAWKINS: What struck me is, when I was a teenager, I made a list of the really big questions, really interesting questions in the world, such as, you know, why does the universe exist? Nothingness seems more likely than somethingness. And then, going why are the laws of physics they are? And then the next one was, why is there life? And then the fourth big question was, well given that there's life, what is intelligence?
And, this was a problem that I felt we could solve in my lifetime, and that it would be a good thing. It would be just, it would be learning about the universe and learning about humanity, and that just couldn't be a bad thing.
INSKEEP: You were a pretty ambitious teenager.
Mr. HAWKINS: Yeah, I didn't tell too many people about it, because it sounded kind of crazy at the time, but…
INSKEEP: Doesn't get you a lot of dates.
Mr. HAWKINS: No, it doesn't, but I did work at it. In fact, I'm unusual, even though I've started three companies and one, three, four proper companies and one non-profit. Throughout all that time, I've worked part-time in the neuro-science field, just, you know, pursuing this goal.
INSKEEP: Well, Jeff Hawkins, thanks very much.
Mr. HAWKINS: Thank you, Steve. It's a pleasure.
INSKEEP: Jeff Hawkins is the founder of the company Numenta, and co-author of the book, “On Intelligence.”
You can hear a longer interview with Jeff Hawkins, and how, in his own words, his computers take a cue from the human brain, by going to our website, npr.org.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.