Jeff Hawkins is at it again…inventing up a storm. No less an authority then Walt Mossberg of the Wall Street Journal refers to Jeff as “…a key figure in the digital industry;” and his legacy as one of the great visionaries in the digital revolution is secure via the millions of devices running his Palm Operating System, and manufactured by the companies he founded (Palm and Handspring).
Not content to create the first practical PDA (the Palm Pilot) and first true smartphone (the Treo 600), Jeff decided to literally re-write the rules of what an intelligent computer should be, in his book ‘On Intelligence’. He founded a think tank to explore real-world development of his concepts, and a new company to develop the algorithms and software needed to bring them to life. Within weeks, Numenta (Jeff’s new company) will be offering free toolkits for his new computing paradigm. If successful, we could be on the road to intelligent machines that learn through sensory input of objects (visual, auditory, etc.), in emulation of the human learning skill.
As if that isn’t enough for one busy genius, Jeff will also be speaking at this year’s “D-Conference,” Walt Mossberg’s blue-chip gathering of the leaders of tech companies, venture capitalists, analysts, and major journalists who cover technology. This year will feature a joint interview of Bill Gates and Steve Jobs (live! On the same stage…), and something else: Jeff Hawkins dropping some very broad hints about Palm, Inc’s ‘3rd arm of business’.
When I asked Walt why Jeff was an invited speaker this particular year, he told me “I have known Jeff for 15 years or so, even before he founded Palm, and have covered all of his major products. He is a key figure in the digital industry, having developed both the first successful PDA and the first great smart phone. He has spoken at ‘D’ twice before, and introduced the Treo 600 at D. (Donna Dubinsky has also spoken at D multiple times.) This year, with all the new developments in phones and mobile, Jeff is a particularly appropriate speaker.”
Now, this all sounded suspiciously productive; so, I asked Jeff if he would get on the phone for an interview with Palm Addict. Jeff is one of the easiest interviews…he is so bursting with ideas, that I have only to ask one or two of the ‘right’ questions, and then write like mad while he speaks his mind. We spoke about Numenta, his take on the health of Palm, Inc. and if that ‘3rd arm of business’ will ever see the actual light of day. Read on for the Gospel According To Jeff.............
JEFF: Just to be clear, I have two jobs: I work at Palm, and Numenta. Actually, I’m an employee of Palm, and a volunteer at Numenta (laugh). I don’t want anyone to think like ‘Jeff’s left Palm’, that just isn’t true.
PJA: Tell me what’s going on at Numenta
JEFF: Numenta is about 2 years old. Prior to that, I was running a non-profit organization, the Redwood Neuroscience Institute, for about 3 years. While there I made a lot of progress on how the neocortex works. The last chapter of my book was about how to take the theory of neocortex, and turn it into a useful technology. At that time I didn’t know exactly how to do this, but I knew we’d be able to.
Right after I wrote the book, Dileep George (a co-founder of Numenta) read it, and came up with a mathematical framework to express the theory…then we could program it. He made a demo, which was very impressive, and it proved to me that we could build a useful technology. I never had any doubt that that the basic theory was correct; but how to turn it into a useful technology is a tricky thing.
So we decided to create Numenta. By putting it in a commercial framework, we can get more people working on this than if I did it in an academic or non-profit environment.
Numenta has created the first Operating System that works on these principles. We have created a set of software…a sophisticated toolset…it’s quite rich. It’s the first time that anyone has ever built a machine that runs on these principles. It’s different than traditional computers (although the software runs on Linux computers). Our software emulates a hierarchal memory structure (like the cortex). Anyone can use the tools and configure these memory systems; it’s like designing a brain. You don’t program it, you feed it sensory data, information like pictures, or from sensors on machines; any kind of sensory input.
Think of it this way, when you were born your brain didn’t know anything about the world; it had to slowly build a model of the world through your senses. That’s what our system does. It learns about its world…whatever it’s interfaced to, and it builds a model. It discovers the structure of whatever it is exposed to. You can use that model to solve all kinds of problems. For example, it can do inference (pattern recognition). We can say: ‘here is a new pattern/image…what is it?’ It can then do prediction in time (what happens next?). Essentially, it’s a memory system toolkit that works like the neocortex. We call it Hierarchical Temporal Memory (HTM).
PJA: So, what you’re saying is: if not right now, ultimately it will be able to look at things, and put them in context of what it’s experienced already?
JEFF: Yes. Essentially, you train the system on some sensory data. Then, new data comes in and it knows what it is. Let me give you a simple example:
Today, there is no computer in the world…none…that you can show a picture of something, a simple picture (a dog, a cat, a car, whatever), and it can tell you what the picture is of. Humans find this easy. I can do this with thousands of objects; I can just name them…its easy! But, there’s no computer that can do this today…not even close. It’s because brains aren’t programmed like computers; they work on different principles. Our technology can solve that kind of problem. It essentially says: “I’ve seen a lot of images and objects in the past; and, now when I see a new one, I’ve already formed a model of the world, and I know how this new pattern coming in likely fits into that model.
We have several major companies working with us: 4 last year plus 3 more this year. They’re applying (the software) to diverse problems. With the public release of our software we hope to get many more people working with us. It is a pretty exciting endeavor.
The best analogy I can make is to go back 50 - 60 years. At that time engineers understood the basic principles underlying programmable computers. However they didn’t yet know all they could do with it…they hadn’t developed compilers, and de-buggers, and CPUs, and memory chips. In the beginning, the machines were cumbersome and huge. We are kind of at that stage right now with HTMs. We understand an important idea: we figured out how nature represents knowledge in a brain, and how it learns. We are creating a toolkit that lets you do the same thing, and apply this to different problems. Yet, it’s at the very beginning. We don’t know what the first successful applications are going to be, what direction the technology will head, or what problems we’re going to run into. So, we’re at the beginning. We are close to making a public release (the ‘Research Release’) to allow other people to experiment with the technology. We call it a research release not because the tools aren’t done…the tools work, and are fairly sophisticated; but, we’ve learned that it takes time…months, maybe a year…for people to get good at it, to understand it deeply. It’s a whole new paradigm for computing.
PJA: It’s like taking someone who has only worked with C++, and then turning them into a JAVA developer. It takes them the better part of a year before they produce anything of real value.
JEFF: Right…but I think this is even a little bit worse. Imagine that you took someone who didn’t know anything about computers, and said: ‘OK, your job is to start writing a spreadsheet program.’
If you’re a competent programmer, it isn’t too hard. You can start pretty quickly. But, if you don’t know anything about computers you are going to go through some time, where you have to learn the concepts and techniques. We find that young people pick it up faster.
PJA: That’s pretty typical of any technology. My son is better with my Treo then I am!
JEFF: Exactly! We’re shooting for March 1st for the Research Release, although it might be a few days later. It will be a free release; anyone can download it, although there is a license. In the future, we may charge for the tools; but right now it doesn’t seem to make a lot of sense. With a lot of the toolset, the source-code is provided; because we think that people can add a lot of value, and create businesses around this. Our goal is to build a community of people working on this technology. Our goal isn’t to own it, or to build the applications ourselves. Our goal is to build a toolset and platform, and to get as many people working on it as possible. We’ll support them, and give them every opportunity to be successful. We’ll make money (eventually) by licensing the intellectual property. For example, if you embed this technology into a car, or use it to solve some major application, we’ll charge you for that.
At the beginning, everyone is just going to be learning and experimenting; so, it’ll be free.
PJA: I’ve been in technology sales since 1978, and I can see how this could solve the problems of ‘rigid rules’ that have hamstrung existing ERP and CRM systems. Systems with very inflexible structures are being used to address the very problems that call for the greatest flexibility and cognitive perspective. For example, this kind of intelligence would be a real boon to a supply-chain solution.
JEFF: You’re right, traditional AI systems are brittle. Rules based systems are only as good as the rules they have; those rules were put in there by a person, and they don’t change…they don’t adapt, at least not without constant and expensive re-programming. Last year, I was invited to talk at the annual AAAI conference, and one person called my work the “AI Spring,” as if to say that maybe AI is finally ready to do all those things they’ve talked about for so long.
A lot of people are excited about what we’re doing; we haven’t been in stealth mode. If you go to the Numenta website, we have white papers, and there’s the book I wrote (“On Intelligence”); but, we haven’t been promoting ourselves, either. Now, we’re going to get some publicity because we are releasing the toolset. Its not going to be a huge PR campaign…we don’t want to over-hype this. This is a challenging technology. The tools take some time to learn.
By the way, what we’re doing is quite biological. I have a presentation that I hope to get on our website by the time we launch the toolset. It has to do with the biological basis of this technology. This is a biologically derived theory. It’s not just ‘oh, humans behave this way’. No…this is the way the brain is structured, and the way the brain does these things…at least how the neocortex does. As far as I know, no one else has done this to the extent we have.
There is no doubt in my mind that this technology is going to rock the world…the question is what year will that be? Who will stumble and hurt themselves along the way and who will be successful? We’re 100% sure the technology is going to be big. However, if you look at the history of new technologies, a lot of early pioneers with the right idea don’t do so well. So, at Numenta, we have to make sure that we don’t over-extend ourselves, or get a major business issue wrong. We are being careful.
PJA: You have to watch out for the David Sarnoffs of the world.
JEFF: Yeah, I’ve heard that story. You know what? You just have to go forward, and try to make something happen.
ABOUT PALM, INC.
PJA: First of all, what is your current title at Palm?
JEFF: The title on my card is “Founder,” and I work part-time, half-time like I said before. I basically work on new projects…projects that aren’t in the public view as of yet.
PJA: Well, you’ve dropped some broad hints yourself over the past year about the infamous, legendary ‘3rd arm of the Palm business?
JEFF: Yes; and I did that on purpose, by the way. So, what’s your question?
PJA: So, you meant it, then? In that case, when? 2007, 2008?
JEFF: We’re going to be announcing something…but I won’t tell you what it is…we will be launching this year, not in the distant future. I’ll give you a much bigger hint: I’m going to be speaking at Walt Mossberg’s D Conference in May. It’s a high-end industry conference. Anyway, I’m going to be giving a talk there, and that would be something for the Palm fans to keep a close watch on.
PJA: Now, lets talk about the Wall Street ritual, where twice a year they say that Palm is a takeover target. The latest is that Motorola is in love with you.
JEFF: Yeah, I saw that, too.
PJA: I think I’ve heard a variation of that for the past 10 years.
JEFF: I wouldn’t pay attention to that. 90% of the time, there isn’t a shred of fact to those rumors.
PJA: I was curious, that Galleon Management (a hedge fund) bought a 6% stake in Palm.
JEFF: I didn’t know that. Believe it or not, I pay almost no attention to the Wall Street side of the business.
PJA: Too depressing?
JEFF: NO! It’s very distracting. I like designing good products. If I focus on that, then everything else gets taken care of. Why does Palm’s stock go up and down every 6 months? It’s a mystery to me.
PJA: Let me ask you this: ignoring all the external gossip and crap that revolves around Palm…especially now that the smartphone market has gotten so competitive. For people who are Palm Addicts…fans of the hardware (the OS issues would require a separate interview), what can you say? Are there words of encouragement, excitement or hope? There is a lot of nervousness in the Palm owner community about how viable Palm, Inc. is for the long term.
JEFF: OK, let’s put this in perspective. First of all, Palm, in my mind, is in the best shape it’s been in, in a long, long time. It’s amazing this company is still around. It was split up, the founders were kicked out, and they sold off the operating system. It’s nothing that I or Ed or Donna had anything to do with; but, this company has been operated in…many different ways.
Now, look at Palm today: we have Ed, practically a founder, running the business. Donna (Dubinsky) is on the board, and I’m still involved. We have our name back. We just bought the rights back to Palm OS from ACCESS; so we now have control of the Operating System again. One of the main reasons we did a Windows-based product is because we were worried we wouldn’t have access to the Palm OS. Palm is in complete control of its destiny again.
PJA: I think releasing WinMo devices was a smart move, in any case.
JEFF: It probably was; but here’s the thing: we didn’t have control of our own destiny for the past 5 years. We didn’t have our name, we didn’t have our brand, and we own our OS. We were struggling to get into the smartphone space, while we were losing our traditional distribution channels. It was a difficult time. While all this was going on, Palm kept growing; and, is still growing rapidly. Although there are lots of competitors, Palm is a very healthy company. We’ve been profitable for many consecutive quarters, years really; and, we’re making lots of money; and, we finally feel like we have control of our destiny back in our hands again.
Now, it takes a little but of time to turn that into products. It takes 2 years to design new products, and so you don’t see this control reflected immediately in the product line. In my view, this is a good time at Palm. In terms of viability, we have lots of cash, we’re profitable, and we’re in control of our own destiny. We haven’t been in this kind of shape in a long, long time.
PJA: Since the US Robotic days?
JEFF: No, earlier then that; not since Palm was an independent company, have we had this much control. People are saying: ‘…are they risky, are they going to go out of business?’ Are they kidding?! This is the best it’s ever been.
Go back to the mid-90’s. we were being heavily targeted by Microsoft, I don’t know if you remember. Their salesmen were saying that ‘…we’re going to kill Palm.’ I used to get condolence emails from colleagues, saying ‘sorry’. Others were saying ‘they’re a 1-trick company,’ ‘they’re going to be put out of business by Microsoft,’ ‘there is all this competition in the PDA space…they can’t survive.’ It was tough, but we survived and actually prospered. Today we are a leader in smart phones.
Now that we have competitors in the smartphone space, some people are saying, ‘how can they survive?’ A lot of our competitors aren’t doing anywhere near as well as we are doing. It always looks like its worse then it really is. We have a lot of momentum, we’re really doing great; and, we have some great new products, including the new stuff we just talked about.
JEFF: It’s not like we’re sitting around going ‘Argh! We have competition.’
No, we have good plans, and a number of great initiatives to move us forward.
You know, I wouldn’t continue working at Palm unless I felt there was a real opportunity to improve upon the state of personal computing. I think personal computing is important, and can we make better computing products. I feel great about Palm. The smartphone is starting to take off. I’ve been predicting this for the past 6 years… and, now it’s happening. Smartphone adoption is going to keep growing rapidly for some number of years, and we’ve got some clever and cool plans that you’ll see unfold over the next year. I’m not worried about the company, and Palm’s fans shouldn’t worry about it either.