Interesting interview with Pat Gelsinger of Intel (thanks to Zeev for digging it up and posting on IH).
A conversation with Intel's Pat Gelsinger
Steve Makris
Edmonton Journal
Monday, August 04, 2003
Pat Gelsinger - Intel's Senior VP and Chief Technology Officer
ADVERTISEMENT
Intel’s senior VP and chief technology officer Pat Gelsinger recently talked with Journal writer Steve Makris. Gelsinger joined Intel in 1979 and worked his way to his current position 10 years ago, making him the youngest VP in Intel’s history at age 32. He was a key contributor to Intel’s original i286 and i386 chip design teams and holds six industry patents and has more than 20 publications to his credit.
Founded in 1968, Intel is a leading chip maker for computers, cell phones and networking, with 78,000 employees and 2002 revenues of $26.7 Billion US.
Q. How would you describe your job?
A. Well, for me personally, my job inside of the corporation is to be the technology guy, to be the one who's really exploring and making sure that Intel is uniquely well positioned to take advantage of these technological innovations as they occur, both inside and outside of the company.
Q. What is Intel doing different today?
A. First, we are trying to create the future, a major shift with or R&D. There is no longer any innovation ahead of us and if we don't create it there is nothing else we can burrow from.
Second, we're reaching the abilities of silicon to go into numerous other areas in the future, beyond servers, desktops and laptops.
Q. Is there a growing divide between technology enthusiasts and the folks who have not been touched by new technologies?
A. I don't think of it that way, we are taking technology to more and more of the population. Before we had the geeks and the wannabe geeks. Now I still have them, but I see an increasingly large portion of the remainder of the population use technology as well.
Q. Do you have a personal goal?
A. To reach the entirety of humanity, every human on earth with our technology.
Q. Sounds nice but how do you do that?
A. Today computers largely require humans to fit with them. In the future computers will largely fit with humans. We have to make technology more transparent and visible, useful in more and more places with less and less sophistication.
Q. Is it still much cheaper to keep existing customers than to attract new ones?
A. To me the magic of the Internet is changing that, because all of a sudden in a homogenous network, anybody can deliver a new application to everybody. You could see these phenomenal growth rates, let's just take an example. Have you read the E-Bay business plan in 1992, let's say, and it required E-Bay to build a worldwide network that allowed anybody on the planet to be able to connect to their auction service.
Extraordinarily expensive, so much so that business plan would have never had any feasibility whatsoever. However, when E-Bay could assume that there was a worldwide network in place that anybody with a PC and a connection to the Internet could access these services, the business plan became not only viable but extraordinary and rapid in its potential for growth, and revolutionizing the online ability of training to occur and in fact fostered the creation of millions of small businesses as a result.
You can look at things like Amazon, revolutionizing distribution and all those kind of things. And to me that's really what is part of the magic of the Internet, when you have global connectivity you can in fact innovate in new ways that fundamentally lower the cost of acquiring new customers and users to your application service, whatever it may be.
Q. Does the word "profit" play in what technologies you push in Intel?
A. If technology can do it we will embrace it, even if it means eating our own children. For example, one of my own incentives, combining the then profitable, stand-alone floating point co-processor with the main 486 CPU got me a visit from two board members. They tried to explain the negative economics of that decision, but it was the right technology and turned out to be the best long term decision too.
Q. Seems technology is shying away from big powerful desktops, Intel’s cashcow, favouring small wireless pocket devices.
A. I'm not sure that actually holds. Have you ever tried to run E-Bay on your handheld, connected to the Internet wirelessly? It's dreadfully painful, isn't it? But if your handheld gave you auction alerts that extends and enhances the experience that you have from the PC. In fact, you might do more stuff on your PC because you have that extension of the service onto your handheld.
Furthermore Intel just doesn't happen to be building PCs, we happen to be the largest
market segment share of the processors that go into those small PDAs today, and we have the largest market share of the servers handling the data from those handhelds.
Q. But people can do more things that run on the Internet with smaller devices. How can you still push the desktop?
A. There’s a Ying and a Yang of that of course. If Internet bandwidth were cheap, free, secure and unlimited to all users everywhere on the planet, you can actually start to make an argument that you don't need any power on the home PC. But it is not, Today you have more powerful PCs but the lack of bandwidth, is like you're sucking on a really thin straw.
Q. So you think that's been holding back technology?
A. Oh absolutely. Lack of broadband is the biggest deterrent to growth of PCs in my mind. I actually have a policy paper that I'll be publishing shortly, that addresses this exact issue. If you go look at Intel's Q2 earnings results, our Asia business growth was faster than the U.S. I believe one of the contributing is factors is Asia’s embracing of broadband infrastructure and wireless data services.
Q. Do you find yourself being more of an idea seller than a chip maker sometimes?
A. Well, we find being an idea generator and industry facilitator as a critical element that enables us to be a chip maker. And if we're not helping to innovate and create those areas of innovation for the industry, the adoption rate for our new chips and technologies is severely slowed as a result.
Q. Do you ever get frustrated with industry partners?
A. Oh yes, undoubtedly, let's be very specific. Do I get frustrated sometimes that Microsoft isn't doing the right thing in the operating system and I can't get a vendor like game publisher Electronic Arts to take advantage of this cool idea we have in hardware, and then get one of the equipment manufacturers to commit to a timely the product and put lots of marketing dollars behind it. Do we get frustrated about that? You bet we do.
Q. How would you compare the Apple vertical market to Intel’s horizontal?
A. There is no right vs. wrong, but we see the extraordinary innovative power of the horizontal industry. At this point, in Apple's 3% of the market share, people say they can't innovate when they (Apple) control the hardware or the software. In our horizontal market anybody at any layer can largely and independently perform independent innovation. Now when you add up those two models, at the end of the day, the horizontal one wins most of the time and that's what we've committed ourselves and our industry to.
Q. How much life does Intel’s traditional chip making silicon technology have?
A. Our belief is that silicon has no less than a decade and probably close to two decades and in fact we're expanding that canvas of silicon to make more and more types of devices. We plan to add new ingredients like carbon nano-tubes on top of the existing silicon superstructure.
Q. Seems like niche chip makers like IBM, boast of new ways to shrink chips, yet you are staying with silicon.
A. Make no mistake, we're doing a bunch of hypey, nanotechnology, like our atomic layer self composition formation. So if you want to go hype we'll talk about a number of those just to keep it fun. IBM makes it sound like silicon is nearing its end, but since their silicon business is trivial and ours is huge, of course they would be motivated to present it that way.
Q. Did Steve Jobs make the right chip decision, choosing IBM for his upcoming G5 processor, or will Apple be missing out on some pretty hot Intel technology.
A. I think Steve Jobs has made the wrong CPU choice for 20 years, he just added a few more years to the life of his bad decisions. Steve's not an illogical guy, he's passionate and opinionated about the directions he wants is a poor path for the company as well as a poor path for the users.
Q. What do Intel chips have to offer?
A. Our chips would help Apple could find ways to open up more applications for themselves, a broader set of products, we have Centrino mobile products that are stunningly good. I don't think it's a good decision for Apple or for their customers, but they've done a good job of turning the company back around at the same time so you can't discount all the things that they're doing and all the decisions they've made.
Q. Where does Intel get its ideas from?
A. We find ourselves uniquely well positioned on the planet to take ideas from other places, to collaborate and work with universities and venture capitalists, to find and understand the innovations that are going on in hundreds or thousands of people's minds, the real brain trust of the world, and not expect that we're just doing all of that cool stuff ourselves.
Q. How does this work?
A. So we've constructed a network where I have about 500 active research grants going into universities, we have five lablets that we built as sort of portholes into the university environment, we have ongoing researchers in residence who do internships, other programs like that where we have this churn and relationship in the academic environment. Intel has a
very active venture capital program under ICAP which has about 400 portfolio companies, many of those we call our eyes and ears deal, to keep our fingers in touch with what's going on in the venture community and that source of innovation. I also have a number of lab to lab research projects with the other major research institutes of the major companies in the world and I think we've uniquely positioned ourselves to do that in a global way, where we have the largest international investment portfolio on the planet, I've expended our research at the lab level to be very focused on the universities in China, Russia, eastern Europe, etc., India, where those other pockets of innovation. Yes we do those things internally, but what's really powerful is when we combine those up with the thoughts and innovations that are going on in the rest of the world.
Q. Can you give me an example of an idea that came out of left field?
A. I'm trying to think of a good idea, let me think about that one a second.
Q. Let's talk a little bit about Transmetta. They are a small competing company, that designed a small and battery efficient CPU that is getting into more small computing devices, especially in the far East. Their premise is to throw away hardware legacy from the CPU, running it if needed in software mode.
A. You can't ignore backward compatibility. People still run applications in corporate environments that were developed 15-20 years ago, in fact many of those applications, the people who wrote those applications are now dead and their children have no idea what they did. You're just not going to change those compatibility requirements, and that's just a flawed, it sounds good, it feels compelling, but it's wrong. Secondly, when they've looked at the implementation the benefits that it brings, yes initially there were some benefits, their chips had lower power at a certain performance level than Intel did. That was because we were asleep at the wheel. We weren't doing the right thing in our products, they were a great wake up call to us, and we now have superior products, the Centrino based products are just unquestionably better in every aspect of performance of power than theirs, all the industry benchmarks bear that out. So it was great wake up call for us, it did exactly that, we're not awake and we're delivering great products.
A personal story, I met with one of the board of directors of Transmetta, a venture capitalist, and he said his response is we underestimated how quickly you would respond. In that case, it took us about a year to get good products in place that were competitive, about two years to have unquestionably superior products.
Q. Tell me about your new Centrino based laptops.
A. I was on the plane with a guy, he had an old IBM think pad, and his battery died after about two hours and I'm here cranking away. I use my computer when were sitting at the gate yet, as soon as it started I'm still using it.
He got up to go to the bathroom, came back and he said are you still going. I said yes, and I'm not even half way done with my battery yet. He was green with envy, it was great. It was a Centrino sales call, it turned into a wonderful experience, I think I sold half of first class on buying new laptops.
Q. Is Moore’s Law changing?
A. Well I think that the beauty of Moore's law is, you get to play it Multiple ways. I can double the number of transistors per chip every 18 months, or at the same cost structure I get to increase the amount, or I can take that increasing budget and deliver a lower cost chip that did the same thing it did before. Or it cold be a lower power chip, or it could allow me to take those transistors and integrate new functionality. So you get to play Moore's law, we the technology innovators get to play this card in multiple ways and that's why it's so powerful.
Q. Now coming to the core of your question, can people keep up with it?
A. Hey, geeks can keep up with anything, they're insatiable. Games, there's certain categories that are just absolutely insatiable in their demand for technology. The generic user, the answer is no, unless you find new compelling usages for that technology that actually deliver them fundamental new value without new cost or without new substantive training to go on to take advantage of it.
Q. Bill Gates told me the next big thing will be solid non-moving hard rives, while your boss, Craig Barrett said longer lasting batteries are on his wishlist.
A. I'll certainly agree with both of those. Computers 10-15 years from now will not have hard drives. You'll still have them for many applications but you will need for more and more of the applications, you want them more rugged, etc., you're going to want to be able to do that. So obviously batteries, unfortunately batteries are based on fundamental chemical properties, they improve about 5% per year, I don't think there's any particular optimism that that's going change in any fundamental way soon. So if I have the same sort of battery life, or the same sort of energy ??? hours available, I have to find new creative ways to use it, not demanding or expecting much more efficient battery sources.
But Moore's law gives me a tool to go approach those problems. Now a few other examples, hey we've been talking about speech recognition for 20 years now, I believe in the next year or two speech recognition becomes very, very cool and interesting because we're going to have moved beyond speech. We're going to move to audiovisual speech recognition.
Actually what we've found, and they just released a set of research in this area. Try, next time you're in a modestly noisy environment, I'd like you to close your eyes and hold one ear shut and try to conduct a conversation with somebody. Just try the experiment and what will happen is after about 30 seconds to 60 seconds, you will start to become disoriented and frustrated in the conversation. If there's a modest amount of background noise it will become very, very difficult for your brain to keep track of what's going on.
It's a very interesting experience if you try it some time. Well that's what we've been asking computers to do with speech recognition work for the 25 years, because they have one little microphone, one ear, and by the way that microphone's a lot worse than your one ear is, the acoustics of the ear are better than the microphones that we put on computers today. What we've recently done is what we call our audiovisual speech recognition library that we've released, which combines stereo colour vision so I have depth perception, I'm able to track people, I'm able to read lips, with array microphones, in our case we put six microphones, so your one ear is much better than our one microphone, we put six microphones in place, six ears.
The results of that is that we see, not improvements in speech recognition n quiet environments, those work pretty well already. A huge increase in speech recognition quality in noisy environments, with modest amounts of background noise. And we think that when you combine these things together, and it's called statistical processing when it combines the best of speech recognition with the effects of vision recognition and directionality, you ring those things together and we see enormous improvements in the human computer interface. And we think those are some of the breakthroughs that
will really leap us over some of these traditional problems. So in that case, you ask is it a hardware or a software problem, well the simple answer as yes.
So next year our platforms will start to include array microphones, six microphones in them, up to six microphones, so much better audio quality input, you will be able to statistically combine that with vision work such s the IBM work that you referred to. So the processors need to get fast enough to able to do vision. Vision's a very computationally heavy task and doing it in stereo is more than twice as heavy. Six microphone inputs, so that's computationally very heavy, and then these big heavy speech and reading algorithms. So I need to make the hardware better, I need to make the hardware faster, and now I've given new tools for the software guys to go to work on their algorithms. So that's what I'm uniquely excited bout, I think that's going to be pretty interesting for us to get over.
Q. What will next generation laptops do?
A. I have this vision of the 111A laptop - one pound, one inch, one day and
"always" connected anywhere, running for 12 hours on batteries...but that is three or four years away.
Q. More connected stuff on laptops, where do all the wires go?
A. It's all going to be built in, four antennas and six microphones, fully wireless. All you will have to do is plug it in the power outlet at night.
Q. We make fun of sci-fi shows like Star Trek and cartoons like the Jetsons. Does their technology affect real scientists?
A. People who envision sci-fi breakthroughs of the future actually helped inspire researchers of today. I had William Shatner join me on stage last year during a speech. He brought along one of the stage mockups for the communicator, and compared his Motorola cell phone to it...they look sort of the same. My observation was that in fact people like Stephen Hawking and the other researchers, are inspired by Star Trek.
I think these visions of what the future might hold actually inspire people's innovation of today.
Q. Are you inspired by sci-fi?
A. I'm not a big Trekkie fan or a big Jetsons fan. To me, interactions with the people, innovators, is what turns me on. Just sit me down with one of these people for a few hours to get me thinking new thoughts and to try to synthesize new things to think about.
I like testing other thoughts against my own thoughts or knowledge of an area. That's really hat excites me personally.
Well two things. One is sometimes we consciously plan those things, we know hat the case, for instance Pentium 4's going into laptops and people doing hat, we knew it was going to happen in some segments. It was like, well, if we don't do it they might use somebody else's chips. So of course we'll support it. In other cases, let me give you the second one, I think that's actually more compelling, in other cases we don't forecast it, it happens, we didn't think of that particular innovation or market application or something else, but if it's our chip hey, that's still wonderful, and that gives learning to go apply to the next generation of what we're doing.
Q. You are starting to talk more about silicon radio, a new way to send and receive wireless information. Will it change wireless as we know it today?
A. That's absolutely my goal and my firm belief, what we're trying to do in the Radio Free Intel silicon chip technology.
Q. What does it have that today's technology does not?
A. I'm going to make more radios, smaller radios, faster radios, cheaper radios, fully integrated radios. We are trying to apply Moore's law, that chips and power double every two years, into silicon radio. The vision is very simple, make every chip into a radio capable device. Everything like cell phones, PDAs, music players and appliances start to communicate with each other.
Q. How small can these appliances get?
A. Literally, cell phones can be as small as the buttons on your shirt or your earring you wear at a dinner party.
Q. What other applications are there in Silicon Radio?
A. When my kid leaves for school in the morning I'll put a little radio on them that allows me to know where they are. I'm going to have location services based on those. I will be able to attach one to my child at night and I'll be able to literally eliminate things like SIDS (sudden infant death syndrome) because you'll know and be able to monitor and interact with them from anywhere. I think the medical applications of these things are going to be enormous over time.
Q. Can everyone afford new technology?
A. We make these things cheap and ubiquitous, it allows us to connect every human on the planet. We still have 5-plus billion to go who've never been touched by technology or connected to the Internet.
Q. How will technology change our lives from now on?
A. The opportunities here are enormous. Does it change everything? You bet it does. Does it provide opportunities? Yes. Will it be disruptive? Highly so. Sometimes we get criticized that you're taking away peoples' business and my response is we're doing what technology allows us to do and that really presents new opportunities. Intel is a building block supplier. We do nothing ourselves. Everything we do is then added to, combined with something else.
Q, What is your business model?
A. My fundamental business model is this strict requirement to partner with and work with all people in the industry. It's not like, I might enable a software guy to take advantage of our new chip. If I don't enable it, nobody's going to buy it.
Q. How many arguments do you win with Intel board members?
A. I don't argue with the board that often. I personally have never been in a situation where the board just shot down a proposal. But they are at a different vantage point and see things that I don't.
Q. Have you had a personal experience that got you going on new technologies?
A. I'm sitting in the plane seat, they're just about to close the door and I get a call from daughter who just got lost in downtown Portland. She had no idea where she was, and was crying, and I could not help her. This fuels my passion for location services, being able to have computers or phones that know where they are and talk to each other.
Q. Would you be interested in running the company, or do you think you're more effective where you are now?
A. Well that's a very personal question so if you'd like a very personal answer I'll give it to you. yes, I want to run the company. At the same time, I have four kids, three of them are still at home, I think I'm in a role now where I can be highly effective for the company and the industry and have a reasonable balance of my work and family life. I'm in no hurry to change, I'm 42 years old, I have 20-plus years here yet, so I'm not in a big hurry.
Q. OK, we know Moore’s Law that says chips and power double every two years. What is Gelsinger’s Law?
A. My law, we will build everything into silicon that allows us to touch and deliver technologies everywhere. |
