October 26, 2005Zeitgeist '05: The Google Partner Forum
Sergey Brin, Larry Page, John Hennessy
In introducing our final speaker for the afternoon, I'd like to try to put in context why it's important to hear from [him] and [his] guests. There are certain kinds of history that are more interesting to the world than others. We know that Google will be examined in the future. It's being examined already – and many people in this room may be part of organizations that are being examined in the same way. I suppose that everybody here hopes to have a story that will be looked at the way Google's is.
We also know also that the social culture, this academic and creative culture in Northern California over the last 40 or 50 years, [is worthy of examination]. In particular we must ask why this region has been so creative, so fertile in a technological and academic and commercial sense.
Our next speaker is positioned at the center of a number of the institutions that have been responsible for this achievement. I found out recently that he was the tenth President of Stanford, which makes it a more rarely-held position than Librarian of Congress – but then Stanford has had a shorter run than the United States. Stanford, as an institution, has been indispensable to the achievements not just of Google but of Silicon Valley in the last half century. And President John Hennessy has been personally involved as a Chairman of the Computer Science Department, Director of Computer Science Laboratory and a co-founder of several companies in this area.
He is going to talk to us about a history of the innovation he has seen here. And then he's going to bring in a couple of noted innovators to talk. So please join me in welcoming President John Hennessy, tenth President of Stanford.
Thank you. I think innovation is such an important part of what has made the Valley a unique place. And it's such an important part of what makes a university vibrant and exciting. One of the interesting things about universities is not only do they have great old things that have been around 100 years, but they have a key role in protecting and making sure those things are available in the future. And they must also look forward to the future – to build new buildings, to conduct new science and research. People often ask me, "How can a university president promote innovation within the university?" The thing you should understand about a university president is he's like a caretaker at a cemetery: there are lots of people under him, but no one's listening.
The important insight here for anybody who leads an innovative institution is that you cannot control it, you cannot dictate the pace of innovation. It has to bubble up from the innovators themselves. When we look at the Valley, people here have made unusual, incredible contributions. And new technology is often the driver for innovation.
In the early days of the Valley, there were orchards. Over time, it got built out, starting with Hewlett-Packard. What was Hewlett's invention? A very simple way to build temperature compensated circuits. That led to the creation of highly reliable instrumentation. That led, probably equally importantly, to a Disney movie called "Fantasia," which used that first audio oscillator. Innovation builds on innovation. And that's continued over the years.
At Stanford we've had people like Fred Terman creating innovation in how engineering is taught. We've had our folks at Apple, who got a lot of their inspiration not only from the invention of the microprocessor, which helped to make possible the personal computer. [Then there's] a place by the name of Xerox PARC, which in the 1970s was probably the best computer research laboratory in the entire world. I'm looking at my good friend Vint Cerf sitting here in the front, and remember the days when he was a Stanford faculty member and helped create the Internet, and that, of course, has changed all our lives.
So that pace of innovation continues. New discoveries and new technologies certainly drive that innovation, whether it's better search algorithms that enabled Google, new ways to build computers, the RISC technology that was worked on at Stanford and Berkeley, Windows and desktop metaphors that were worked on at Xerox PARC. These kinds of innovations created new opportunities. One of the most interesting kinds of innovation that occurs is when there's a collective breakthrough – a series of discoveries and inventions happens close enough in time that you create a tipping point, a discontinuous change [affecting] the way we live, the way we play, the way we work.
Think about the microprocessor, low-cost hard disks based on Winchester technology, bitmap displays, the mouse, the Ethernet. Those are the technologies that all came about in roughly the same period that enabled us to build workstations and personal computers and change the way in which we did computing. The Internet, the World Wide Web, a way to name locations universally, visual browsers: that all came together and created the modern definition of what we think of as the Web. Then Google and Yahoo! and eBay and Amazon came along and built their technologies on top of that. But it was the combination of those different technology changes at once that really enabled us to change the way we do things fundamentally.
I believe this pace will continue, that we'll continue to see great innovations in information technology. And here's a reminder that sometimes innovation takes a while. About a month ago, the DARPA Grand Challenge took place. What was the DARPA Grand Challenge? To drive the vehicle autonomously, given a set of points through dirt roads and real roads, to complete a 132 mile course which had to be completed in under 10 hours. One year ago, when this contest was held, no vehicle could go further than 7.5 miles before running off the road. This year, five vehicles completed an entire course that, in addition to driving through a variety of roads, included tank traps placed on the road. Again, no driver, no tele-operation, this car recognizes the tank trap and avoids it. Finally, a mountain pass with hairpin turns – it happened to be named Beer Bottle Pass – most of us would have closed our eyes driving.
This is fascinating because a whole set of technologies came together: New laser range finders, new planning algorithms. And the car that won, Stanford's car called Stanley, a modified Volkswagen Touareg, won by the use of machine learning. A set of machine learning algorithms were incorporated that allowed it to capture the last 10 percent, figure out how to avoid obstacles in the road. That's a really technology breakthrough. In less than a year, exponential improvement in the quality of building an autonomous vehicle. It's going to change our lives. It's going to eventually become a technology that we all expect in our cars, the way we expect anti-lock brakes and airbags today.
I think this pace of innovation is going to continue. Alan Kay's quote here is exactly the right one, "The best way to predict the future is to invent the future." We're going to see dramatic breakthroughs in biology, this will be the century where biology changes our lives like no other technology. Why do I say that? Just look at what's happened in the last 50 years. [Back then] we didn't understand DNA; we didn't understand the cellular code. In 50 years, we're beginning to understand how cells actually operate, functional genomics, how things are translated from code into the actual cell operation. As we begin to do this, we'll be able to, as Ray Kurzweil said earlier, "reprogram cells, enhance their abilities, enhance the ability of a cell to fight disease and to repair damage." Understanding this is also key to insuring that stem cells can be used for important new applications. I think we will see a tipping point in the use of the biological sciences and their impact on our lives.
Forty years ago, there was no Silicon Valley. There was a place whose primary material import wasn't silicon. It was iron oxide, because the number one computer product made in the Valley 50 years ago was magnetic disks. Then it became silicon, and therefore, Silicon Valley. Today, there are more software engineers in Silicon Valley than there are hardware engineers. And that's going to continue. Information technology will continue to be important, and so will the biological sciences. And together, those two things will really change our lives in fundamental ways.
When I think about how you build and keep a creative culture, I'm reminded of a wonderful quote that my predecessor, Gerhard Casper, who was the ninth President of Stanford, once said. "Even after 100 years, each day in the life of a university must be a first day." And that's important for all of us who are involved in the creation of innovation to remember. Thank you.
I now have the honor of introducing one of our next two speakers who will join me. And after our two speakers have had the opportunity to make a short presentation, we're going to open the floor for Q&A and really mix things up.
Our first speaker is a man who really needs no introduction. [Sergey Brin] is one of the two co-founders of Google, and is still a Stanford PhD student on leave. Sergey has been an innovator in everything he does and has an incredibly broad range of interests. And we asked Sergey and Larry to talk about things that they were absolutely passionate about. Sergey continues to be an innovator in his thoughts and his directions, just as he was at the Stanford days. So please join me in welcoming Sergey Brin.
Thank you for coming here to join us in our event. I wanted to talk a little bit about how sometimes innovation happens slowly, but it's still important. Probably many of you have heard of Moore's Law. It basically said that we double the transistor count on chips every 18 months. [It has changed] over time, but basically it's an exponential improvement in computing performance. Google really benefited from that. When we were in grad school, doing our data mining experiments – because of CPU power, disk capacity, memory capacity, these basic hardware components improving so quickly – we were able to do what would have been thought impossible just a few years before in terms of indexing and making searchable very vast amounts of information on relatively small computing resources. [So] in that case, a well-known improvement in technology was able to drive new applications.
Sometimes improvements aren't as broadly visible, but they bring about significant change. [In that vein I will] talk today about clean energy. Most of the energy sources we use produce carbon or nuclear waste. There's a small amount of clean energy, which is mostly hydropower, and then a small amount of wind and solar. Hydropower, we're mostly using up in many parts of the world. So the parts that might be scalable are wind and solar. And I want to talk a little bit about those two.
These are the ones that emit no carbon, solar and wind, and at least in theory and according to a number of studies, that's good for the environment. If you look at the amount of capital costs you need to put in some of these systems, you can see that solar is quite expensive still, both photovoltaic as well as thermal. But as you go down, on the bottom there is wind. And wind is now within the band of fossil fuels, like natural gas generators and things like that.
[Here] is a cost-percentage breakdown for a wind project, like five megawatts. You can see most of the cost is actually the wind turbine. If you can lower the cost and increase the output of the wind turbines over time, just from better engineering, larger scale – you can lower the cost. And, in fact, that's what has happened.
So for the first time in history, wind power now actually makes a lot of sense. You can't use only wind power, because we want to do some things that require electricity when it's not windy. But I think wind power [can take up] a good portion of the energy that the world needs. And if you depreciate the capital costs, it works out in terms of per kilowatt hours [to be] very competitive with the existing utilities.
People that I run into at conferences say they're investing in wind – not out of altruistic intent for the environment. They're saying, "I just made this great investment. I'm going to make this back in so many years." Probably some of you have read about Berkshire Hathaway's investment in the really big 300-megawatt project in the Midwest. I have friends in Europe who have said they've put up so many tens of megawatts, and it's made for better business because there are some incentives that the government puts in place. But the prices are actually getting so low that those incentives aren't even necessary anymore.
And here's the consequence: the DOE [Department of Energy], which is probably being very conservative, shows a massive shoot-up in the amount of wind power that's available. And [use of] wind shot up like this because of cost factors coming down. Some people complain [wind turbines] are eyesores. I think they're pretty cool looking. I think they'll continue to go up, perhaps with improved technologies that look even different.
At the same time, solar costs are still pretty high up there, obviously because of the capital costs, printing solar cells. It's an expensive process. It takes energy in its own right. But it's actually also following a trajectory: even though it's starting up high, it's dropping. And I know companies in the area that are doing research. They claim they'll produce multiple companies that have different technologies, thin films, doing clever things by concentrating sunlight better. So that'll also make a big difference in the developing world, because these things scale down in addition to scaling up. And you can make village-sized solar cells, solar power generators.
I wanted to tell you a success story, though it hasn't completely finished yet - another example of somewhat slow progress by Moore's Law standards, but still significant. And that's in battery technology. Lead acid batteries have been around forever. This is the watt hours you get per kilogram and nicads for 30 or 40 years, liquid metal hydrates more like 20 years or so; lithium ions are relatively more recent. And there has been tremendous progress in battery technology. When you look at the issue of clean energy for transportation. You need a way to store the energy and if it's not going to be gasoline, batteries are a pretty good way to go. I think assuming we're going to rely on fuel cells ignores this point.
Now we have lithium ion batteries, and there is a car called T-0, recent versions of which can go 0 to 60 even faster than 4 seconds. It'll go 300 miles on a charge. It'll beat almost any other car out there in a 0 to 60 drag race. And the reason these things are possible is because this basic technology has finally crossed a threshold. And so I'm excited to both see what we can do to help technologies cross these thresholds as well as recognize when they have and see what new applications are possible, just like Moore's Law made Google possible today. Thank you very much.
Thanks, Sergey. If there's one more significant breakthrough in efficiency – which is the critical problem – how many of the photons do you turn into electricity, solar cells will instantly become competitive? Especially when oil goes to $100 a barrel, which will happen at some point.
I have this small difficulty in that the name Hennessy is associated with a fine brand of French Cognac. In the early days of search engines, I would type John Hennessy in, and I would get hundreds and hundreds of listings for buying this fine Cognac, which wasn't what I was looking for. And when they first showed me Google, I remember typing in John Hennessy – and lo and behold, it went to my website. And I saw that there really was a breakthrough that could be had in search. The president [of Stanford] at that time was Gerhard Casper – when he typed in "Gerhard Casper," he got hundreds of hits to sites dedicated to Casper the Friendly Ghost. Quite distressing to a university president.
When I saw that you could type in Gerhard Casper and it went to his university website, I realized that search could be done in a whole new way and that it was going to change the nature of how we used the web. So the other half of the team that helped create this search algorithm that was the beginnings of Google, Sergey's fellow graduate student in computer science, is now going to come out and tell you about his views on marketing in the future. Please join me in welcoming Larry Page.
Now, I must tell you what John didn't mention, which is that John came to me and complained that if you typed "Hennessy" into Google now – I think this is still true – you again get the liquor.
Now you get the liquor again, which just means that the web is being used more and more by commercial providers.
Anyway, we had an interesting conversation about that. Now, Sergey and I have worked together about for about ten years and so when he says he's going to talk about alternative energy, I say, "That's good, and I'll talk about marketing." You shouldn't take this as an indication of our relative interest in these things, although I think we're both very excited about these topics.
I'm having a great time here at Google and I just wanted to thank all of you for coming. We really appreciate all of your business. I think our strength has really been that we've been a place where all the authors and publishers and businesses and users can come together in one place and do things that are useful and efficient for everyone. And this, hopefully this is a way for us to start the conversation with all of you, and learn what's important to you and how we continue to make that work better for everybody.
Now, we've had a lot of interesting speakers already this morning. I just wanted to cover some things today that I learned. I would not have predicted an Internet company would end up sending two million pieces of first class letters a day. I thought that was pretty interesting. Do you know which one that was? Netflix, of course.
I also was upset with Malcolm Gladwell, because he couldn't tell us if we were a Picasso or not. He sort of hemmed and hawed and said, "Well, maybe you're both." And I said, "Well, that's good. And we should be both." And I was also really impressed with Craig [Newmark] of Craigslist, because he actually asked his customers how he should charge them, which I thought was really innovative. I wouldn't have considered doing that. And I think it's a really good idea. I think having those kind of conversations with your customers really is a good thing for companies to do. If some of the big companies out there started doing that, what people would say?
So, I'm really a computer scientist, so I thought I'd do a little escapade into marketing. I think I'll just give you my bias upfront. It's very sort of computational, sort of results-oriented. The thing that I always thought about marketing is, "they help you choose which kind of toothpaste to buy," which is really a bewildering problem, as you can see when you go into the grocery store and you have to pick your toothpaste, and it's not really obvious why you would buy one versus another.
Yet a lot of money is spent on [marketing] toothpaste and other kinds of consumer products. I've always been kind of confused by this: Why does that money get spent? There are a lot of smart people spending big checks. What do they know that I don't know? And I'm still not quite sure, but I'll try to explain it a little bit.
I would encourage a lot of you, you think about when you're doing marketing, how much is that really worth to the consumers, because ultimately they're ending up paying for it. The example with toothpaste – at Wal-Mart and grocery store chains, they use the same factory that makes your consumer brand toothpaste and make the same thing, and sell it with no-brand on it. And they subtract the cost of marketing, and the consumers get to decide which one they want.
Now, on the other hand, if you were buying medicine or vitamins where it's hard for you to tell what they're really doing, maybe you're willing to pay for that brand name, so that you know there's somebody with a lot of money at stake if they do something wrong. You see that maybe with condoms and other things like that.
[Meanwhile], people have a lot of choices about how to spend their time. One of the things that I've always been confused about when I watch television is that I see the same commercial maybe 20 or 30 times. And you know, it'll be really polished and it'll be kind of cool the first three or four times. And then, after the 20th time you get pretty sick of it. In fact, we learned today that 70 percent of the people with a TiVo skip the commercials, which I thought is a really interesting statistic. (I assume the other 30 percent are like in the bathroom or whatever.) That's my own experience using TiVo. And I don't think the solution is to make fast-forward buttons illegal or impossible – it's really to make the commercials better integrated with the shows and in a way that actually serves people's interests. That's what we try to do very hard [in our business].
And I'm amazed when I look at consumer products, like when I was buying a scanner. (At Google we have an interest in scanning lots of things, for example books and libraries and things like that.) So I was really interested in scanners that could scan things quickly. This may be eight years ago I got interested in this. So I went around and I went to the electronics store and I tried to buy a scanner that was fast. And all the scanners said was like 600 dots per inch and 1,200 dots per inch, even going to 2,400 dots per inch, which is small enough that you can pick up individual drops of ink or dust on your paper. [This] isn't terribly useful, because it takes a really, really long time to scan something. But none of them would say how fast they were – because they were all really, really, really slow.
So I think there's a big opportunity when doing marketing to redefine the metrics that people use. Imagine when you bought a coat if somebody would tell you, I don't know, maybe Google or someone else – about a thousand women have looked at this coat – say you're a guy, you're pretty young – and they had rated this coat as the most sexy for your body type. How much more would you pay for that particular coat if you really trusted that information, and it was really true? Those kinds of metrics don't exist yet, but I think that we'll start to see a lot more metrics for things as online stuff really starts to take off.
Now, I wanted to show you this [eye-tracking study] because this is an example of the kind of data you can really gather. On the left there is our website and the color refers to how often your eyes are likely to go to that part of the page. We actually have a lab here and we measure this. You can see that people like the search results, especially the first ones, the top of the page. The top left of the page is where you start reading, and it sort of falls of from there. But they actually do look at the ads as well, and the reason they do is because they're actually useful. You can see they're not quite as likely to look at those as the search results, but they're pretty likely to look. This is something we try to keep track of, and this kind of data can really affect how you do things online. You have the opportunity to present everyone something different, every person can get something slightly different and slightly better. And that's what we work to do.
Purchase data, of course, we use hugely. You can also think about using things like people's emotions. I read recently that if you listen to techno music while you're driving, you go like five miles an hour faster. Imagine if your radio station could measure the speed that all the people are driving. And then, it would be great for lawyers, I guess, lawsuits about people crashing into things. But you can imagine a lot of data that's inherent in the world that will start to be captured and be useful.
Now one thing I wanted to bring up, I know a lot of you guys are big advertisers and partners of ours, as we're really sending you customers. Imagine we send you a customer and they look at some low margin item – an MP3 player, say, and then they decide they really want to look at a diamond ring. And they don't actually buy that ring, but they just look at the ring. Now the fact that they've done that is probably worth money to your business. Currently we don't collect that kind of information – we only collect information really on who clicks on ads, and occasionally we get from you when people buy things, which is very useful. The thing is, people think about buying things a lot more often than they actually buy them. And that's very useful data for us to do the optimization, to send you better customers.
So I think you're going to see much, much more data of this type. Maybe you have too many of a particular item, or maybe you have some extra guys standing around in the shipping department that don't have anything to do, so maybe it makes sense to lower the prices a little bit while they're sitting idle. I think you'll see there are very sophisticated online companies like Amazon starting to do things like this with us – they have programs that talk to our ad system. You'll see much, much more like that in the future.
One of the big paradoxes for me has been: do you have peer accountability, like the kinds of things I've just been mentioning, or is advertising really about fun? Like when you're playing a video game, and you have a giant monster that's sponsored by the toy company, and you can buy the monster. Maybe that's a pretty good experience that's really about fun. I think you're going to still continue to see a huge mix of these things. I think people get a little confused about it being one [peer accountability] versus the other [fun].
I also think people make purchase decisions differently for different things. Almost everybody, when they buy a car, goes online and does a lot of research. Are you going to do that for your toothpaste? Or what do you do for a digital camera? I think the answers really change a lot depending on what it is. This is just a quick example the kind of data you can get. This [refers to slide] is for a recent movie, "Hustle and Flow." We worked along with [the studio] a lot to promote it, and it's basically the number of searches people are doing for that over time. You can see the release date and you can see the decay [in number of searches] over time. There's just a tremendous amount of data available that's useful for doing marketing.
We're also looking at doing things like our offline ads. This is PC Magazine, where we're running some ads from our advertisers using an efficient auction system, like we do for the online ads. There's a very small URL you can follow at the top that lets you quickly get online and look at more information. So that makes print more trackable, and also more efficient.
One of the things that's interesting for us is the huge opportunities out there. Sergey and I used to joke: there were these 100 person search companies in the Valley, and we used to wonder what they did. We had about four people. And we asked, "What do all the people do?" As we've gotten bigger, we just see more and more things that we should be doing. And all of our people are busy doing the things that we're doing now. I don't think that's going to change, there's just so much opportunity. Maybe we're doing a couple percent of the things we see, and also that our competitors are doing.
We tend to be an innovative company. We're looking for new things to do. And we tend not to go into the exact space that a competitor is in, because it's easier to go into a space where no one is doing a good job. If you're a little imaginative about these things, you see many, many areas [needing attention]. And you've seen that with some of the things we've talked about, whether it be our libraries projects or book scanning, many things where we just saw big opportunities that weren't being addressed very well.
[In his talk] Ray Kurzweil mentioned smart computers. One of our big goals in search is to make search that really understands exactly what you want, understands everything in the world. As computer scientists, we call that artificial intelligence. I don't know if we will get there exactly on the year that Ray predicted on his exponential graph, but I do think that the power of computers as they're getting faster and faster have enabled us to do things like Google. It wouldn't have been possible a couple of years ago. I think our ultimate goal is to really make a smart search engine. And that'll be a big deal if we can do it.
Okay, we're going to open the floor up for a Q&A. We have two microphones in the aisles and we also have a roving microphone. So if you want to raise your hand if you're not near one of them, a rover will get you there.
Because this is a partner forum and many of us in the room have really good ideas that can help us all grow together. So what's the best way for us to funnel those ideas to you? I mean, do you have a conduit open right now that'll really listen?
Well, the easiest way, send them to firstname.lastname@example.org and I'll forward them. Also, depending on what kind of partner you are, be it in advertising, publishing, AdSense, they are also your respective representatives. I do want to apologize – there are a lot of people who send us really great ideas all the time, and sometimes we don't get around to addressing all of them just because we're a bit overwhelmed. We're working on improving that, so please don't stop.
How about over here.
Hey, Larry, you were showing some interesting things with trend graphs for buzz [about search topics]. Is that something you're going to be offering anytime soon? Because I'd love to see the ability to go back and do trending on the zeitgeist for specialized topics – to be able to know what's coming, what's increasing – as opposed to just snapshots.
I agree, but it's hard to do. We have had a Zeitgeist thing up for a while that gave you certain things we would choose, but obviously, I think this is very interesting information.
My question is a bit different. You gave us ways to find out information and information is power, information is knowledge. I have my personal source of information, my grandfather and grandmother. Can we develop something through, say, blogs for elderly people to leave to us what the wisdom of the world is and make it available for everybody so that we always get that information from the Internet? What do you think?
I think that would be wonderful. I've spent time interviewing my relatives myself just to record their stories. I come from Russia , which has gone through turbulent times over the past century or so. I just did that for myself and probably didn't record it in the best way possible. So that's really an interesting idea – to be able to both gather that information in some way, talking to someone and then typing up some notes might be a good way to go, or to videotape them if they feel comfortable. And then, if we were to provide some kind of infrastructure that maybe keeps it personal and private for a while, but eventually over time gets unlocked. I think that would be a great thing to do. Thank you.
I'm wondering if you're going to work on time-based search in the main Google engine. Let's say Karl Rove was indicted today, I could search on Karl Rove and see what just happened in his life rather than what got linked to last year or three years ago?
Well, we already do that to a large extent – we include news, for example, when you do searches on something like that. I imagine if an event had happened around him recently we would definitely include news articles that were timely. That's what search is about is getting people the things that they're interested in. And a lot of times that's timely information.
Right. The news comes on the top of the page, it's not in the main engine and there's no way to switch from the relevant-based view to a time-based view, and there's no way to find my blog, for instance – that might be relevant on some news that just happened. You have to go to a separate blog search engine, which is still starting.
Yeah, I think you're right. There are a lot of weaknesses. There are the advanced options that do give you some time choices, but they're probably not nearly as sophisticated as you would want and you suggest. And they don't allow you to do the inverse, like what happened to Karl Rove before this [news] came out, which is sometimes what you want to see too.
Ray Kurzweil here, and I have a quick question for each of you. I was interested to hear about your interest, Sergey, in energy and particularly in solar, which personally I'm quite optimistic about. If we captured one percent of one percent of the sunlight that falls on the earth, we could meet our 13 terawatts need. It'll be three percent of one percent by 2030. There are some new solar technologies that are much more efficient, lightweight, inexpensive. You alluded to a few of them. So I actually do think that we'll see that solar figure increase significantly over the next decade. Ultimately I think we can actually capture that one to three percent of one percent when we get to the 2020s. Are you personally involved in this type of technology – fostering it, mentoring it?
I'm very optimistic too. I made tiny investments a long time ago before I really had the resources in at least one solar company that's doing interesting work in making it cheaper. And I assume you don't mean capturing that solar energy the easy way by releasing greenhouse gases.
For electricity. I'm also quite optimistic. And the next thing to overcome, if you can do that, is just to store it overnight, which I'm sure is also doable through nanotechnology and all of the things that you advocate.
Yeah, better batteries and fuel cells, because we do need to store the energy for wind or solar.
Now Larry, when I was here three weeks ago, I was talking to some of your fellow engineers about your language translation project, which is actually quite impressive – using data-driven approaches to these self-organizing paradigms to discover the rules. And apparently there's one pair of languages where nobody on the team knew one of the languages at all and actually got results that compared reasonably favorably to human translators. So is this an area that Google is going to continue to explore and make available to its users?
We haven't announced anything there yet, but we definitely have a great team that's working on that. I think if you think about doing search as really understanding information, then the kinds of techniques you'd need to do language translation [for] are also needed to do search. So we view that as a basic thing we should be working on. Our spell checker was written by a fairly small team of people and it works in very many languages that they didn't speak either, which is really an amazing thing. You can use some of these learning techniques to do things [where] the programmers don't even speak the language but the same techniques work.
Yeah, I mean it's actually very encouraging because Turing based his whole definition of AI at the human level on a language-based test. So language really embodies human intelligence. I think that's the greatest asset that Google has, is this database that you can mine to create intelligent systems.
Our Research Group recently won the machine translation competition by NIST for both Chinese to English – I forget which direction it was – and Arabic to English. So we're very excited about the progress we've made there. It'll take a little bit of work to make it "productionized" and allow people to use it directly, but we're very optimistic.
We're also very excited about Arabic being chosen also, because there's a very small number of books translated into Arabic every year. It's on the order of 500 or something like that. If you think about the issues that the world faces, imagine if you had really good access to English and other languages from Arabic. I think it'd make a big difference.
I'm curious – as you expand into other markets, are there some things that you've seen or learned about the way people use search that is something you want to introduce to Google globally or just in the U.S. ?
I would say probably the markets that we gain the most learning from most unlike the American market would be the Asian markets. And if you look at Korea , China and to a certain extent Japan , the language usage is very different, the cultural differences are there. South Korea has the highest broadband penetration in the world and they have Internet usage that has completely baffled us. There's a lot of learning internationallym and that's why we also set up our research labs in Japan and in Zurich and Brazil , all around the world. India also.
Speaking of news and the news site that you have, which is fascinating, can there be better technology that differentiates between different news sources so that sources of news that more people find reliable get a higher ranking than information that is less reliable – based on a number of factors, including the fact that John Q. Public just wrote about it in his blog, and that's the only way he wrote about it. Is there some technology that you can apply to news and to other aspects to increase our understanding of the bona fides of the information?
We get covered occasionally in the media, so we're pretty familiar with a lot of the issues. Google News causes us to like think a lot about [this]. I'm really optimistic that with the advent of online reporting, and more of the news coverage moving online – and I think this is something people really overlook between bloggers and sort of professional journalists – is that the bloggers actually link to each other and excerpt each other and things like that and the professional journalists do not do that or do it very, very rarely. That's actually a big issue. The nature of the online news reporting can be very fast. It's not really necessary for someone to write the same story as someone else. They can just link to it. It's less work and you could write just the unique parts of that story. And you could also comment on maybe what's valid and what might be suspect about the story, rather than just rewriting it but changing a little bit of the content. So I'm really optimistic that if we get some better systems in place for online news that the existing, I think, very high quality, professional staffs can be a lot more efficient, and get people news that's a lot more interesting and a lot more correct.
But it isn't just who links to your story, because many people can link to all kinds of things – if thousands and thousands and thousands of people go to the New York Times because of the authority that the New York Times brings, or CBS, versus some more interesting or bizarre story, maybe people are linking to them. In the public view, they don't have that same bona fides. But using the metrics you have, they [are] kind of even…
I guess we would ideally like to see not just trying to label the authoritativeness of the source but of each individual article and every individual work and like Larry said and even perhaps sanitations as to what might be the issues here or not. I think these are hard problems, which is why we don't pretend to have solved them today or even necessarily have a solution in the future, but it's certainly something we'd like to try to solve and are working on and hope some of our experiments might be successful.
I'm from Brazil , and I'm very curious about the big phenomenon that is happening in our country. It's about Orkut. In one year and a half the audience grew up from 0 percent penetration to 50 something, close to 60 percent penetration. And I think [now] more than 70 percent of all the people in Orkut [are] from Brazil . What happened?
Shouldn't we be asking you? That's a great point to the previous question about what can we learn from other countries. Many of you might not have even heard of our product Orkut, but it's phenomenal in Brazil . And the next country after that, I believe, is Iran , although they shut it off from time to time. And after that is Pakistan . But you're right. I think it's like 80 percent, a huge percentage is Brazil , and growing at an incredible rate. We cannot keep up. We deploy a lot of big products that get used a lot and we are struggling to get those machines in there. It's something that we study and we'd love to reproduce that success in other markets beyond Brazil and Iran and Pakistan . But also to a certain extent that reflects how different the markets are and perhaps there are different products that we should have in different markets.
You never quite know when a market like that is going to take off. I remember the thing that convinced me the web was going to take off was when a local pizza parlor put up a web page. Because all the techies were using the web, but what convinced it that it was going to break into the consumer space. And that was it. When the pizza parlor put up a web page, I said, "This is going big time." And that kind of thing, you see that kind of thing happening around the world now.
You guys have a great search engine. You guys have a fantastic company ethos as well that I think many people, end users and even competitors admire and like. It's a good democratization engine in many ways. You're doing a lot of good for the world. One of the things as a fan that concerns me or at least I'd like you guys to comment on is the pace at which you guys are introducing different ideas that go against different established oligarchies and monopolies and whatnot. And these are all big industries. It worries me a little as far as – the question I guess is how do you guys decide how fast and furious, and how many people to take on and to not spread yourself too thin, so that you can continue to provide high value services in every area that you go into and continue to disrupt at a good pace and not overextend yourself?
I was trying to address this point at the end of my talk. It's kind of strange for us, because we really don't see it that way, but I agree this is part of the way it gets perceived. What happens is a lot of our competitors, they get really focused on what we're doing. And as a result you see a lot of stories like, we're taking on somebody. Like I said, that doesn't make strategic sense for us, and if you see companies doing that, they're usually doing something sub-optimal for themselves because every company has different strengths and weaknesses. You should choose what's optimal for your company to do. For us, we would deliberately look not to do that. People telling stories about this say, "Oh, that's really a sexy story that Google's going to compete with us, or Google takes on this company." That's usually not what we mean to do, those things are often exaggerated. There's been a lot of stories like that about us, and there's a lot of other companies out there that are doing just fine. And I think that will continue to be the case.
We run Google in some ways very much like a university, like Stanford. And while we have some important strategic directions that we discuss at executive management, we also have all these 20 percent projects that individuals can do pretty much whatever they want. And it sometimes has its costs. I mean, sometimes we unnecessarily agitate certain industries and it sometimes distracts our attention from one of the principal things that we really care deeply about. We so far, however, have benefited from of the successes that have come out of such projects like Google News or Google Groups or Gmail – small, individual projects that people have started up as 20 percent or as a small project. And some of them just haven't panned out. It's something we need to be careful about, but I do certainly like everything we've managed to create that was one of these little things that we didn't really plan on or push. As Eric likes to say, we don't want to say, "thou shalt not" to our employees and our teams – not do this, not do that." We'd rather try to figure out well how can we help these industries, how can we help these partners. That's the philosophy we have.
Google as a network service provider – that's not software, that's infrastructure. Could you give us a little bit more insight about that initiative, those initiatives?
Are you referring to the WiFi kind of access or…?
WiFi, and also I understand from your – from sources that you're buying dark fiber and rights to dark fiber in various transit. And I mean, you're becoming a network service provider, a global network service provider. And that's very different than your core business. And if you could give us a little bit of insight into it, I'd sure like to know.
All right, I think that's people connecting the dots kind of and sometimes incorrectly. We have lots of dark fiber and we need it. It connects our data centers, which are large and spread throughout the world.
Yeah, actually, yes, we get it initially as dark fiber. We light it pretty quickly and we're often under capacity, thanks to the many Orkut users and Gmail users, Maps users, all those products which actually use a lot of those resources. And we have experimented with WiFi. There's a pizza place I think down the street that has a Google WiFi and a gymnastics gym. And also we did spend I think about $1,000 and bought some access points and hooked them up on Union Square in San Francisco, and provided some I think in Bryant Park in New York also.
We are also putting in a proposal, as many of you probably read, in San Francisco , to try to WiFi the city. To us this is an experiment. It may seem like a large experiment. We've kind of looked at financial costs and think that they're fairly modest. We don't have a particular end goal to take over the world with these WiFi things. We'd like to see how cheaply we can provide this service, and are there revenue opportunities behind it. Particularly, we're interested in local advertising opportunities for our Maps product, for example. We'd love to partner with other companies for it, because for one thing there's a lot we have to learn – and also we have no intent to do this sort of thing like worldwide, to take that on. I think that would be way too expensive and far afield from our core.
I'm Vint Cerf, and since I'm a new Google employee, I may be breaking the rule by asking a question here in this public forum. Well, actually, I'm going to ask you a question, John, first, OK? And I have questions for Sergey and Larry. The question for you, John, has to do with innovation and how it works at Stanford and upon what it depends. Terman introduced a fabulous amount of resource in the university by drawing on government research funding and expanding the engineering faculty dramatically. I'd like to understand further from you in the 2005 period, how heavily dependent is Stanford for its innovation resources on federal government grants?
Well, like all U.S. universities, [Stanford] is still incredibly dependent on the federal government. The federal government provides the lion's share of research, especially in the biomedical area, where they are by far the largest funder. What the federal government doesn't necessarily do well is to start up some of the most important new activities – to find that group of individuals that is going to create a breakthrough, going to create a whole new radical technology, but they're going to do it in some area that's not currently well-funded. That's what the universities have to find money for, so that that young assistant professor who is going to do something great is not inhibited by the fact that they don't have a gray beard yet and a long history of accomplishment. Because that will be where the innovation [comes from] – so much of the innovation comes from young people, and it's absolutely crucial.
So I would conclude from that that we might give some serious thought to how those government funds are allocated and what thinking goes into identifying the people that get them.
Oh, absolutely, absolutely. The great secret of academic institutions is that they have this hidden talent that works for almost nothing. It's called graduate students. They're smart. They don't see boundaries. They're willing to do new things. They're willing to upset the apple cart and discover something brand new. And that's absolutely crucial to why universities work so well.
And actually, John, that sounds like 95 percent of the Googlers I know, too, especially when they're doing work on that 20 percent, because it's their idea
Okay, so here's a problem for Larry and Sergey. The search engine today that we run, and others run, gets a snapshot of the recent Web. What it doesn't do is capture the ephemeral nature of the Web, something that [Librarian of Congress] Jim Billington brought up, in a sense, when he talked about American memory, the capture of materials that are out of print and so on. What would it be like if we had a kind of temporal knob that allowed us to see the Web as it evolves, to look back into it? I know there is a project that Brewster Kahle has worked on related to that – is that something you think would be of value from the standpoint of being economically supportable, or is that simply impossible because of the huge amount of memory that we already require just to keep track of the current system, let alone what's gone on over the last dozen years?
Well, we don't like to use the word impossible. And especially not when there's a proof of existence such as Brewster Kahle's Wayback Machine. For those of you who don't know, if you go to archive.org, that's the Internet Archive, and they've been storing the Web for a number of years, and they have something of a search feature on it. It does not quite provide you all the capabilities of like a Google Web search over all data, but it's pretty good and it's something I've used from time to time. For example, you can see the home page for Google that Sergey designed in like 1998 on there. We have many embarrassing things on there.
I think that among other challenges we need to find a way to deal with that embarrassment. But I think that's exactly the kind of thing I would expect people here to undertake and to use some of their time on. Just because we don't see exactly how to pay for it or whatnot today, that doesn't stop us from doing projects. And Google News and Google Image Search, by the way, still don't have ads. So I'm sure we'll finally figure out how to pay for it in the future and I hope we manage to accomplish that.
You also are challenging a number of different industries, and so partnership and cooperation might also transform into competition. You are innovating at a very fast pace. Well, if you have to partner, you have to build trust and confidence in parties. So can you give us some underlying ideas, guidelines, you have in developing the way you want to challenge the different industries, on what grounds do you think to compete in, and on what grounds to cooperate, because at the end of the day you have to make money from what you do.
I like to think that we are not setting out to challenge existing industries. We obviously have a number of products that have bred competitors, as Larry said. And in some cases, like Gmail – there were other webmail providers. If any of you have had a webmail account, did it go up significantly in capacity over the past year or so? They mostly have, and I think that's been really beneficial for users and for the webmail industry.
Many of our projects sometimes are controversial. Like I know Google Print for Libraries has been – there have been some objections raised. That's the project where we're scanning library books to make them searchable and to make a card catalog for the world. I know there are some legal debates, but on the whole we really care about publishers, and we want to support their business. I think that the Google Print projects are going to sell more books for publishers and authors. And while there may be a debate to one particular implementation detail in the library part of the program – exactly how the permissioning system works, opt-in versus opt-out – on the whole we'd love to grow the publishing businesses, and that's what we're setting out to do and that's what we're measuring our existing small implementation of Google Print is doing.
We should say too Google has really thrived from having a huge network of partners, maybe unlike a lot of how a lot of tech companies got started. I mean, we had, for example, a Netscape partnership really early on, and we actually had a misunderstanding about the amount of traffic they were going to generate. And we actually took down Google.com – this was a long time ago – in order to serve their traffic. And they were pretty happy with us. As a company I think we've been very good to our partners. And we've had a very large number of them, both on the ad, serving search to companies, also serving many different companies advertisements …
Even companies that compete with us on search, like Ask Jeeves for example, that we're the primary revenue source for. We like to serve companies and make them successful, even in some cases when they're competing with us in certain product lines.
Question to President Hennessy. What will be your advice to two young, really smart grad students that as we know are the most important secret source to a great university – [when they] will come to you with big eyes and fire in the belly and tell you that they want to quit their PhD in order to go and execute their dream?
Well, I have a reputation for doing what Fred Terman did many years ago when those two graduate students were Billy Hewlett and Dave Packard. I think when you have a great technology that can change the world, you should go fulfill your dream. That's certainly what I told these two guys, and I think they've done a marvelous job of it. So thank you all for joining us this afternoon.
end of session