Jeff Hawkins thinks he is on the verge of two breakthroughs. As chairman of the Redwood Neuroscience Institute (www.rni.org), where he spends half his time, Hawkins has helped prototype a novel computer vision system using an algorithm based on his theory of how the neocortex works. That algorithm could serve a wide array of applications and spawn a new direction in electronics, Hawkins believes.
The other half of Hawkins' time is spent as chief technology officer at PalmOne ,where he designed a widely praised smart phone, the Treo 600. Hawkins is now working on what he says could be the next big thing in mobile computing a smart-phone-related gadget that will debut in less than two years.
The 47-year-old engineer still sports the boyish looks and enthusiasm of a grad student, even though he has helped found three companies and designed handhelds, such as the Palm Pilot,that established the category of personal digital assistant. Long fascinated by brain theory, he recently published a book on the topic (see www.onintelligence.org). Hawkins sat down for a chat not long ago with EE Times editor at-large Rick Merritt.
EE Times: You say the next big thing will be machines that work like brains. Why?
Jeff Hawkins: It was clear to me 25 years ago that the brain works on a set of principles different from our everyday computers. Those principles produce amazing behaviors. Flies can fly better than any machine we ever made and are just that big [pinches fingers together]. So I felt that if you understood how the human brain works you could build machines that work the same way, and they could do things humans could do, like recognize speech.
But as always in technology, when you come out with a whole new computing paradigm, it will blossom into a gazillion things people never thought of before. People who invented the transistor and microprocessor had no idea what was going to come from it. This [brain theory algorithm] is going to be bigger than all that stuff, I promise you. It is the ultimate way of information processing. It's a fundamental capability that has so many ways of extending what we do as humans that it will be incredible. It will be right up there with the invention of the digital computer and electricity. It will transform our lives in a major way.
EET: You say in your book that some of the Microsofts and Intels of this new industry will be founded in the next 10 years.
Hawkins: I'll be bolder than that. . . . Now I think it will be two years. I am already spending time thinking about how to commercialize this. For starters, we can solve some major outstanding problems in computer vision that have no answer today.
EET: Like what?
Hawkins: There is no computer in the world that [can] tell cats from dogs or people from cars. They only work in very constrained environments. But the way the brain works, you can look at patterns that are completely different and know they are the same. I see this as a Treo [moving phone around] and that as a Treo and that as a Treo. How does that work?
Computer vision guys have no idea how to do any of that. I think I know how to do it. I'm virtually certain of it, because we have it working now.
EET: Describe your prototype.
Hawkins: It's basically solving the invariance problem, which is: How do you recognize objects in all different variations? And it's doing it, I believe, in the same way the brain does it.
It's doing pattern recognition on 90 different symbols, including black-and-white only and line drawings.
It's not doing anything practical yet. It's not in the state where the average person can look at it and say it's amazing. But a computer vision person might look at it and say it's impressive.
EET: Why did you focus on computer vision?
Hawkins: Because the sensor issues are easier to deal with, it's a well-understood market and it's an academic problem that people can relate to quickly. I've already talked to some vision researchers about this, and some are pretty excited.
There are 10,000 applications for this. The point is, the algorithm doesn't just work on [vision], it works on anything. It will work on speech recognition anything where you are trying to understand the world.
EET: What's the next step?
Hawkins: All I have to do is show you can do something really hard that no one has ever done before, and show how simple it is using these tools, and then I will get people's attention.
That's my mission over the next year. Maybe we will have a conference to bring a lot of people together and have a demonstration.
EET: And the larger goal?
Hawkins: I am on a mission to get an industry started and create a movement. I want a movement that's both academic and industrial, much like what happened with artificial intelligence, but based on the right science.
It may take a while to get all the academics to understand this, and I have to prove it. But I have a lot of people excited about it now.
EET: Let's talk about mobile computing. How do you see your job as CTO for PalmOne?
Hawkins: I'm really a product guy. I tend to focus on the next big thing.
I was very involved in the Treo 600. I was almost not involved at all in the Treo 650. [Now] I am working on something else which is completely different. My value is trying to figure out how to get the next thing going.
It's hard in any company to start something new, because there is a lot of resistance to it. People don't understand it. It takes a lot of force to make things happen. That's the role I play.
EET: So, what is the next big thing in mobile computing?
Hawkins: I can't talk about that, of course.
EET: What's the time frame of this next big idea for mobile?
Hawkins: Somewhere between a year and two years. I'll give you this little piece: I actually had the idea for this thing almost three years ago, and I got really excited back then. At [former PDA startup] Handspring, I couldn't do this idea. We didn't have the resources. But at PalmOne we can do it. That was a big plus for me when we formed PalmOne.
EET: How does this killer idea relate to today's smart phones?
Hawkins: The smart phone is clearly going to be a major driver in all multimedia computing. In a few years we'll have 3 billion cell phones out there, and they will all be smart phones eventually. They will all have lots of memory, processors and operating systems.
This is a huge sea change in the cell phone industry. You will see an upgrade cycle you could compare to the PC world going from character-based to graphical-based computers.
Clearly, smart phones will be people's first and foremost browser, their first line of defense on e-mail. There will be more and more things no one has really talked about yet.
EET: How do you innovate in this space?
Hawkins: I make the analogy that you want to scope out territory, like settling the West. You want to scope out the right piece of land. You don't actually know what is going to happen there. But you want to pick the land with the right attributes. If you are right, people will move in next to you, and you will be in a great spot.
EET: What are some of the big problems in mobile-systems design today?
Hawkins: It's a very complex product design and sales space. We have all these carriers, each one has their own idea of what a smart phone should be, what the technical specs should be, and what e-mail and the Web-browsing experience should be.
It's very complex and daunting. We have chip makers like Qualcomm or the GSM guys who control their deals and have their own ideas about using Brew [software], or this or that.
There are [wireless-carrier networks such as] GPRS and Edge and EV-DO and 3G, and different flavors of them. For a developer these are very expensive and complex, and difficult to integrate. The radios and software are costly, and we have to do them all for unclear advantages in different markets. We could be building more and better products, but we spend so much of our resources doing this stuff.
The channel is complex too, because it is a different set of people at the carriers who have their own ideas about how they are going to market these products, and you have to deal with them.
It's frustrating at times. We can't control the process. We have to make 20 little teams, one for every little carrier. That's an obstacle in general for moving the market forward. It slows things down so that people can't innovate.