Back on August 13 2008 I interviewed Vint Cerf, Google's Vice President and Chief Internet Evangelist. It was a brief but rich exchange. Here is the text:
SV: Good morning, Vint. As you know, I am writing a book called The Googlization of Everything.
VC: By that I suppose you are using "Googlization" as a proxy for search in general.
SV: Well, yes, in a way. But that's only because of the ways that Google itself has moved itself into so many important parts of our lives. By all recent statistics, Google is the dominant search engine in the United States and a significant and growing presence in the rest of the world. Is Google becoming synonymous with the Internet itself? Is Google as an interface merging with the Internet itself, creating a unified experience?
VC: We are not as prominent as the statistics suggest. The use of Google varies significantly by place. In China, for instance, we are latecomers. Languages have specific properties. Chinese, Korean, Slavic - they all have different spacing rules and word endings that you have to know about if you are going to do search. That's why we need native language experience and have offices all over the world.
SV: What about the cultural power of Google and search in general? Jonathan Zittrain wrote a few years back that the battle to control domain names would not last too long as control over search terms and keywords would displace a good domain name. It's better to place high on a search result than to have a pithy domain name.
VC: I used to think that. I think that's not true now. Right now we are seeing an increase in top-level domain names. We can't dismiss the domain name world entirely. Domain names are important for us to get back to where we were before. They are important for binding sites to identities. Search is not everything, it turns out.
SV: Ten years ago, from the outside, did you see the importance of search to the Internet experience?
VC: Ten years ago we had Alta Vista, WAIS, and before that Gopher. All of those have very high conceptual value. But you had to know too much to use them. You had to know what you were looking for. The rise of social networks, blogs, and other sites has been very important. Recommendations still play a remarkable role in guiding me to remarkable places on the Internet. A good example is Google Alerts, for instance - having a tool that helps you observe what's going on on the network. It's a like a search that is always running.
As good as search is, it's oddly unsatisfactory. Web search still leads people to documents, but it's not necessarily leading them to answers. We don't have a good way of expressing the questions or the answers.
SV: You are talking about semantic search. A lot of librarians I talk to say that semantic search is not worth the investment. They say that we are semantic enough and that the textual and relational methods of Web searching work well when you know what you are doing.
VC: Certainly, semantic search is more an idea than a practice. People also say that about artificial intelligence, machine translations, and speech recognition.
Translation is surprisingly turning into something more pragmatic than ever. We are using statistical analysis rather than examining the deep structure of sentences. So we use things that humans have translated. We can then compare documents and say, "this phrase matches that phrase." In speech understanding, things are getting better. So I am not willing to give up on the idea that we can understand the semantics of the Web.
Look, we understand each other in this conversation because we share a common language and some common experiences. So the challenge is getting computers to achieve that level of common understanding. I am tempted to think that we won't have this kind of conversation with a computer until a computer can experience things - see, hear, smell, touch things in the world.
Look at language acquisition in humans. Babies learn by hearing certain words expressed in certain conditions. So they relate what they hear to what they see and feel.
SV: But why do we want to converse with computers? There are six billion humans I could converse with.
VC: Yes! There are things that computers can do that six billion humans can't do. Computers have the scale capacity to discover and analyze things. We can increase the total quantity of stuff that we can understand. We might want to discuss with a search engine what it is we are trying to discover.
Look search today is messy. Think about one of those big construction shovels, you know, like a tractor with a big shovel on the front. And you have to operate it by pulling and pushing a series of levers. It's big and imprecise. Using a search engine today feels like trying to move one of these Earth-moving shovels.
SV: So short of semantic search, what are the more short-term improvements to search that you would like to see?
Librarians always tell us that they would love more detailed control over the search mechanism. They would like to be able to manipulate the search better or search within intermediary results. That's really had to do on the scale on which we operate.
I would like to see increasingly effective translation. I would like to see us be able to search more types of digital objects - large-scale databases, documents, spreadsheets, presentations, games. We don't know how to index these things yet.
SV: I read today that Google was the search engine of choice for 70 percent of the searches in the United States in July. Assuming this number keeps growing at a high rate, what chance is there for diversity in the search market? How could another firm hope to take some of that market share and innovate effectively? It seems like Google is doing all the interesting research on these problems.
VC: Google does not believe it has a corner on creativity in this domain. That's why we open up our APIs. A good example is Google Earth, where we know others can add real value to the service. When we designed Google Docs we made sure people could get their data out of the service. And it's clear that lots of people are exploring different and better ways to do search. I don't think there is any one company that could have a monopoly on these ideas. That's one of the reasons we are so committed to keeping the Internet open - something I care deeply about.
SV: Let's talk about that. There is all this dueling lobbying and arguing going on at the FCC and in Congress between firms like Google and the telecoms. What can Google do to change the narrative to make this issue seem clearer or understood in different terms?
VC: That's a good question. We have to get beyond the bumper sticker slogans about all this. Look at it historically. When we all used dial-up, we had thousands of Internet service providers. Now that we use broadband there are at most two players available to subscribers - often just one. Competition is not disciplining that market. Unless you have enough competitors you have the potential for very uncompetitive practices. I am afraid that the system in the United States is just not set up to deliver the best broadband service.
Now in some places in the world - the United Kingdom, Japan, France, the Netherlands, New Zealand - carriers are compelled by the state to make service competitive. But we seem to be stuck in this odd loop that says deregulation always creates competition. It's just not true.
SV: So how can Google change the narrative?
VC: Well in the bidding for the 700 MhZ space, we made recommendations to the FCC. They took two of our four recommendations to make the mobile spectrum more open. The company that won that bidding is constrained to offer the spectrum with these openness conditions. We are also doing a lot with White Space in the television spectrum. And we are encouraging the implementation of broadband through municipal networks, something I am very interested in. It's interesting, the telecoms say they don't want the government to compete with the private sector in this area. But all the proposals would include a private company contracting to provide the service anyway.
In other areas, we are looking at better uses of spectrum and access to spectrum. We are looking at the potential of running fiber along power lines, although it seems that the idea of using power lines for broadband is not working out. We are looking at the potential of using satellites for broadband. We are encouraging people to explore or understand or articulate these technologies. We are excited about all these areas, but it's not in Google's best interest to go build infrastructure.
SV: What aspects of Google have made it such a success?
Well, we have hired some awfully smart people. And most of them have the benefit of being young. So they don't know "you can't do that." So they just do it.
Plus, the company gives its employees the time to explore new ideas - 20 percent of their time.
The management structure is very flat. And the company has very particular hiring practices and standards. We hire people who are both smart and responsible. They all have a "how can I help" attitude. So you can learn something from almost anyone in the company. You will notice an almost collegial atmosphere here.
The real brilliance of Google is the ability to monetize search through AdSense. This company uncovered the relationship between advertising and information. The old way of advertising had no direct interaction with the audience. But now the audience can click. So suddenly advertising is not a sales pitch. It's a response to an expression of intent. This form of advertising is narrowcast, personalized. It has very different properties than the old.
SV: Thank you very much for your time, Vint. I hope we can speak again soon.
Comments (1)
"I would like to see increasingly effective translation."
A condition for good translation is clear source text. Much text on the Internet is not clear. Often, it contains partial sentences, idioms, irony, rhetorical questions, anaphoric references such as 'this', 'that', and 'it', synonyms, non-standard grammar, and ambiguity. All these features cause problems for machine translation of text.
To increase the quality of translations, a good strategy is to educate people on how to write for machine translation.