• Biography
• Threads

MYHRVOLD: We realized a number of years ago that in order for personal computing to move forward the companies that were involved in it would have to start doing research, and developing new technology very rapidly. But for a long time the technical agenda for little computers was set by big computers. In the early days of the personal computer industry, people would say well when are you going to have multi-tasking or multi-processors. There's a whole laundry list of features that first existed on mainframes or minicomputers. The workstation guy inherited that list and they ticked a whole bunch of the things off; the personal computer industry inherited that list and ticked a whole bunch of things off. It was clear even six years ago to us that that day would be coming to an end relatively soon, that we would have done everything big computers did. And, in fact, given the horsepower that was going to be available in microprocessors, our little computers wouldn't be little any more. Our little computers would soon have more computing power than a big computer of that era. Certainly if you take just the machines I have under my desk here, or take the hottest machine you can buy today - 200 megahertz Pentium with 64 megabytes of RAM, etc. - well that's vastly more powerful than workstations or minicomputers of five or six years ago.

We realized that we couldn't keep implementing somebody else's agenda; we had to develop new things. Many people in the computer industry did develop new things; I don't mean to say that the personal computer industry was all about slavish copying. But, you know, the graphic user interface is a great thing for Apple; it was a great thing for Microsoft, and others - so thank God there was a Xerox to invent that in the 70's. So Apple and Microsoft and others could spend the 80's implementing that cheaper, more efficiently, and extending it. At some point though we can't rely on companies like Xerox or the old line hardware companies. Most of the development of the computer that was done privately was done by two kinds of companies. It was either done by companies not in the computer business, like AT&T, the height of Bell Labs for many many years - during the height of their computer research - was at a time that it was against the law for AT&T to be in the computer business. Xerox is a company that wasn't in the computer business that did a tremendous amount of research there. Or, it was hardware companies: IBM, DEC and others. We thought that there should be research done by software companies - or a, done by PC industry companies, or by software companies. We started investing in it, 5 or 6 years ago; we've built a quite large and I think quite successful research group; we've got about 170 people in it. And over the course of the next few years we're going to dramatically increase the size of that and keep it growing with the company.

BROCKMAN: What kind of budget are you talking about?

MYHRVOLD: We're still working that out, but one of the things that's funny is that you've got to work hard to grow something as fast as the rest of Microsoft, so in the last couple of years our spending as a percentage of our overall spending, in research, has declined even while we increased research, we just didn't increase research as fast as the rest of the company. We can look at reversing that, and continuing to grow the research group.

The thing about research is of course the total effect is not directly proportional to the total amount of money you spend. You don't want to have warm bodies there, you want to have THE best people you possibly can. And that's always been a constraint on research expansion.

In areas like databases we have people like Jim Gray, who are one of the regional founding fathers of relational databases and transactional technology. In graphics we have arguably one of the best groups in the world; we have people like Jim Kajiya and Jim Blinn, Andrew Glasner, Alvy Ray Smith, and a host of other graphics researchers. In speech recognition we have a very strong team that has people from Carnegie Mellon and other leading universities. In natural language we have a very strong team. In programming tools and programming technology - which is very important to us because we use that technology ourselves - we have two or three very strong teams. So across the board - we also have a bunch of people who have been in the industry for a long time and have a tremendous track record, as generalists. Butler Lampson, who was there at PARC when the great things occurred and was a leader at DEC Research. Gordon Bell, who helped develop the Vax, and championed the it and Ethernet, championed the whole bunch of other major technological things at DEC -works for us. Charles Simonyi is another guy who has been involved in key things in computing at every stage � he has been a longtime Microsoft employee, but wrote the first graphical word processor at Xerox. And through a combination of having young people who've just gotten their Ph.D. and are out to conquer the world, and some of these folks who have been conquering it steadily, at least in the research sense, for the last 20 years, I think we're going to have a neat mix.

When I first started recruiting for Microsoft it was a real bitch, because we didn't have a reputation with academics for being a research place. Academics are often quite conservative; the places they like to go to have a history of 50 or 100 years like Stanford, MIT or Carnegie Mellon.

We had to start one of those without having that long history. At first it was difficult to get people, serious people, to be interested in us as a research place. As a development place we were well-known, but particularly five or six years ago the company was known in academic circles mainly for DOS, which didn't make lots of computer scientists want to come work here. But we persevered, and hired a few good people and then they helped us hire more, and those helped us hire more, so now it's such an interesting environment to be in that recruiting is a whole lot easier than it was back then.

One of the things that we thought was very important was establishing a research culture where we didn't have an us-vs-them attitude between research and the product. Or an ivory tower attitude - there's a variety of ways you can describe this - where you get a dysfunctional relationship between the people that have to ship your products, which is what ultimately pays the rent, and the people who are doing research. Part of our proposition was that that needn't be a big gulf between the two. Of course you're not going to force your researchers to go out and fix bugs in the next version of Office. They wouldn't find that very fun, and actually the people that work in Office would say, What, are you kidding? They don't know enough about it to go do that. You don't want to ask people who are fixing bugs on the next version of Office to say, Oh, and in your spare time we'd like you to do a whole bunch of research. They're different activities. But there's a tremendous amount of synergy that can come from having a good relationship between the people in research and product groups. We've had dozens of things spin off from research into the product groups, to the point that I don't think you could find a major product at Microsoft that doesn't have some code or some significant technology from research. They all do, at this stage.

BROCKMAN: Given its culture, can Microsoft ever be as open as a Bell Labs?

MYHRVOLD: We do that already. Having a lab which is closed and buttoned down and secretive is ultimately counterproductive. Because you're only going to have a small fraction of the world's smart people working for you; you're never going to hire them all. By being part of a larger research community you get the benefit of people criticizing - or supporting - the work that you do. We encourage people to publish - in fact at the last Siggraph conference, 20% of the papers at the conference were from people at Microsoft. If you look in other disciplines you'll find that people from Microsoft are not only key contributors to conferences and are very open about their work, they help organize the conferences. They're on the editorial boards of journals, and other things. We think it's very important to maintain an open environment.

At the same time, you can maintain an open environment and still get plenty of proprietary value out of things you invent, because you understand them best, because there's a variety of ways of protecting your intellectual property. So there's no contradiction between running research as something that will ultimately yield a great profit and doing great open research. There isn't a problem. People think there is, but there isn't. The main difference between research and one of our other product groups is not a profitability issue. Research is already very profitable for us, if you count the value of things that have moved across.

It's more of a time-scale issue and a predictability issue. Research is much harder to predict when it's going to be valuable to you, or, when you start a project, if it's going to be valuable to you. We have a variety of things in speech understanding and natural language understanding, and other areas, where the potential for it to be enormously valuable for us is there. But it's very hard to say, that'll occur in one year, two years, five years - or maybe never. If you have the ability to sit back and say fine, we'll take the risk on that, then you'll get a great return on it. But it's not something where you can manage by saying what's our benefit, our profit, next quarter, from research. It doesn't work that way.

The Chef talks about selected Digerati...