Sergey Brin of Google at PC Forum 2003, impressionistic transcript by Cory Doctorow.

Reverse lookup from phone numbers are easy, we can detect them and return appropriate results. But looking up proper nouns is hard. Some combination of pagerank and other technologies will have to be brought to bear.

Some day, you won't be able to email a query to a friend and rely on her getting the same results -- that's the downside of customization. If you take the average Google Query and give it to a librarian, 80% of the time, he'll be able to find the answer without any further customization -- they don't need to know how old or young the questioner is, or what her zipcode is.

But that does mean that 20% of the queries could be improved through personalization. We need to figure out how to introduce that fairly complicated technology quickly and without it being confusing.

We do Google News. It's great, but we don't know how to make money off it. On the other hand, after acquiring Deja and turning it into Google Groups three years ago -- a purchase we thought would just give us great content and no money -- but now we're running ads targeted on the content of the message posts in Google Groups. That's good for the advertisers and for users.

And now we're trying this on hundreds of other sites, like

The Blogger opportunity came up -- a successful product, but not a successful business. We looked at it and said that Google could bring a solid infrastructure (like 10,000 machines, data-centers around the world) and the ability to run targetted textads atop Blogger pages.

Beyond that, dozens of ideas have been thrown around about where to go next, but the Blogger guys are currently occupied with getting the migration done, and they'll have to figure out which of those ideas to try later.

We'd be foolish not to consider and IPO, but right now, there are a lot of downsides to it. It's a lot of work (I'm inherently lazy), and the S1 form is very long (laughs). All of our financials would be disclosed to our competitors. Worse, it would focus the company too much on the short-term money picture.

But our employees and investors expect to get some money for their stock, so we still keep the possibility open. But laziness keeps on winning out. There are so many better ways for us to spend our time -- at the rate our biz is growing, the benefit of going public would be far exceeded by the opportunities we'd lose int he 3-6 months we spend working on it.

[Esther: What don't you want disclosed about your finances?]

There's nothing, really -- nothing a competitor could turn into an advantage. We probably worry too much about secrecy. Our employees know our financials today, but after an IPO they'd be distracted by all the analyst chatter and the stock-ticker.

The Semantic Web sounds really improbable. Asking everyone to write in a new language -- XML or RDF -- seems like an unreasonable burden on too many people, to make computer-readable info. We should use computers to understand humans, not humans to speak in computer-ese.

We wouldn't ignore RDF or XML if it were present, but we don't expect it to be in great profusion. We'll go anywhere that there's lots of information.

Re: a Google for your own hard drive. Well, Google's goal is to make everything searchable. Most of the products that do this suck. They index and use the disk when I want to and I end up turning them off. Google doesn't make client software, but we haven't excluded the possibility.

Re: Google's "immortal" cookie. If we set a one-year cookie, then all of your preferences would disappear at the end of the year, so we set a cookie that lasts indefinitely.

Do our algorithms embody values? I think if they do, we discover them retroactively. PageRank is a democratization of the Web, but we only figured that out some time after we invented it to solve a concrete problem.

I think that there's promise in the idea of not just linking people up with documents that experts have identified (viz. PageRank), but also linking searchers up with those experts. It's not in the pipe, but I like the idea.