By Jennie Z. Rose

JEFF HAWKINS, inventor of the Palm Pilot, has acquired a reputation as a bleeding-edge leader in technology revolution. But with the recent publication of his book, On Intelligence (cowritten with New York Times science reporter Sandra Blakeslee; Times Books, 272 pages, $25), Hawkins's total immersion in the question of how brains work is now out of the closet.

Mainstream brain theory is data rich and theory poor, Hawkins says. It focuses on bottom-up circuitry. Hawkins is more interested in top-down processing, in which the neocortex makes predictions about incoming information.

The auditory system works similarly, which, he says, is why we can make out what someone is saying in a noisy room. Hawkins's mother twice took him to the doctor to get his hearing checked as a boy; it turned out he tended to hear what he expected to hear rather than what was said. This wishful hearing could point to Hawkins's tenacity, or his idea that intelligence is all about prediction. Or both.

Bay Guardian: What's the overall implication of what this book is telling us?

Jeff Hawkins: Everything we do is based on our brains, and if we want to understand why we do the things we do, how we make mistakes, why we are creative, what's our place in the universe, and all that sort of stuff, it would be nice to know how brains work.

When you understand how the brain works, you can start to understand how to raise kids, issues about disease, understand more about how brains go wrong.

And the area I'm very comfortable working in is building this stuff. It's not like building a human being, but you can build machines that work on the same principle that are undeniably intelligent in their own way. This is going to be a huge business. I think it's going to happen fairly rapidly. We're working on it now, not 15 or even 10 years out.

My entire book is about defining what intelligence is and how the cortex works. With this understanding it is now possible to start building useful machines that work on the same principles. We are just starting this work. In a few years we can judge how successful we have been.

BG: Your history involves some disappointments in the academic world – first at MIT and then at UC Berkeley. You've had more success as an entrepreneur. Does this affect your approach to science at the research institute you founded?

JH: Redwood Neuroscience Institute is in the science world. It is a nonprofit institute. The scientists there pursue their own research, publish papers, etc. There is a Stanford graduate student, Dileep George, doing his Ph.D. work at Redwood Neuroscience Institute. He is implementing a hierarchical memory system based on the ideas in the book and is making good progress. His system is solving visual understanding problems [recognizing images] but doing so with a hierarchical memory model. What makes it unique is that it can fill in missing pieces of an image and predict what it should be seeing. It is a good start. We are not trying to make practical things just yet. We are at the stage of understanding the technology, not building products. That will come later.

I do infuse some practices from the business world. I try to foster reaching consensus on ideas through various means. For example, at a conference we cosponsored I had theorists and experimentalists meet separately to discuss what the other group could do to help them. Then they presented the results. Normally these groups don't talk to each other. The idea of a breakout session was new to them.

BG: After researching brains for 25 years, what are the absolutes in your book?

JH: Intelligence is about making a model of the world and making predictions. The old idea is you have this computer in your head, the input goes in, you chunk on it, and then you do something. But that doesn't explain what creativity is, what understanding is. It's been a false idea. The new idea is the brain does not compute; it's a memory system. It's like computer memory, and it remembers the world the way you've experienced.

The whole idea is about prediction. The goal is to recall past events in a way that predicts the future. To understand something means that not only do you recognize it, but then you have expectations about how it's going to behave: what I'm going to say next, what word is at the end of a sentence, what will happen when I turn the wheel of my car, and so on.

BG: Is this what we call artificial intelligence?

JH: I don't like the term "artificial intelligence." Decades ago the A.I. community promised they could and would build truly intelligent machines. It hasn't happened, and I argue in the book that it won't ever happen until we first understand what intelligence is and how the brain works. So today there are no intelligent machines. Remember, however, intelligent machines will not necessarily be anything like humans or even do the same things humans do. I feel it is important to stress that point because most people's first impression of what an intelligent machine is [that it's] akin to a humanoid robot. That is not going to happen.

BG: What sort of data is this idea of prediction and memory based on?

JH: It's data collected from all sorts of brains. All mammals have a cortex, so a lot of research is done on rats, some on rabbits, and a lot of data is gathered on humans too. What I've done is put all the pieces together in a cohesive framework. That really hasn't been done before. You can't pick up any book or paper and read the step-by-step logical analysis that On Intelligence has.

BG: Does this book convince us that of all the structures in the brain, the core of intelligence lies with the neocortex?

JH: The majority of neuroscientists would agree with that. It's not a controversial statement in and of itself, although you will find many scientists who study other structures will argue about it. They'll argue that you won't be able to understand the cortex without the other structure. No one disputes that the cortex is where all these high-level functions reside. The issue is can you understand it without understanding the basal ganglia, or the cerebellum? I don't know about convincing people, but there's a good portion of the world I don't need to convince. The others, well, we'll see ...

BG: You predict a new industry based on machines with hierarchical memory. How will this change society?

JH: It's hard to predict the future, as I say in the book. The best you can do is to look at certain trends. All I can say is the technology will have the same sort of pervasive effect as the digital computer. It'll work its way into a lot of things. Machines that can understand their world, make predictions about it, and be intelligent will be common. We won't be sitting around the breakfast table and talking to them, but they will be a tool. We'll make cars that drive themselves.

BG: You write that true revolutionary intelligent machines will be "able to experience genuinely exotic, alien worlds of sensation." What are the chances these machines will outsmart us?

JH: When people invented the computer, they were afraid of it. The same is true for the steam engine. What actually turned out with computers is that, well, they have some downsides, but a lot of people have benefited. Computers didn't turn out to be a terrible, dangerous thing that took over the world. They turned out to be clunky machines that don't always work all the time, but they really have been very useful.

In this case, people might think that building brains is like building robots. They'll think about I, Robot or the Matrix movies. This is all made-up confabulation. I'm talking about building machines that understand the world on the same principle that you and I understand the world. They won't have emotions, but they can build models of the world and understand it. They might understand the weather better than you and I do, genetics, various social problems. But they'll be tools, not autonomous entities.

BG: I think the fear is that once you have an intelligent species, that maybe it could actually control its own evolution.

JH: Well you used the word "species," and I didn't. A computer is not a species. It does not replicate. Computer viruses do, and those are bad, but in general, computers don't do that. They just sit on your desk, and they're a bunch of plastic and sand. I talked about this recently because of the article in Wired by Bill Joy called "The Future Doesn't Need Us." He confused self-replication with intelligence. They're completely different things.

BG: But is it possible to create something that is intelligent without any possibility of evolution?

JH: Really there's no connection between intelligence and self-replication. Computer viruses, man-made viruses that attack humans – no intelligence required. Self-replication is required. In this case, intelligence has no necessary relationship to self-replication. If someone wanted to build a self-replicating robot, which I have no idea how to do nor is it in my field of work, but if someone wanted to, I suppose they could also try to make it intelligent. But the danger is the self-replicating part.

BG: Is it possible you're wrong on this?

JH: Maybe I'm wrong; maybe all along the desktop computers have been silently plotting to revolt. And waiting for the moment. But it's not going to happen. These are like memory chips or hard drives. There's no replication. I could be wrong, but it's the same chance of your car, eyeglasses, or computer replicating. People are confusing the issue of being human with intelligence. It's in the movies, books, going back to Alan Turing. It's all wrong. I'm not trying to make humanlike things. There's no genes; they're just pieces of plastic.

BG: Do you believe people are ready to accept the premise?

JH: I think the ideas in the book will take time because there will be a lot of people who don't like it. They may find religious implications they don't like about it, or they don't want to believe the brain can be understood. It'll take time. Remember, until the 1950s there were whole societies of people who thought the world was flat. They didn't believe it until they could get on a plane and fly around the world.

BG: We have an existential curiosity about our brains, though.

JH: It gets into the core of what humanity is. To talk, ask questions, to be able to query anything – it's the brain doing it. We think about ourselves as this body, but it's the brain doing everything. You are your brain. The body is going along for the ride.

Jennie Z. Rose is a San Francisco writer.