The transcript of a 'philosophical' dialog between Microsoft Chairman and CEO Bill Gates and Canopus Research President Will Zachmann

20 July 1987

-----------------------------------------------------------

(The conversation took place in Bill Gates' office in Redmond, Washington. The third participant, Marty Taucher, was Microsoft's PR director at the time.)

 

Will Zachmann: "Let me start off by asking you if you would try to comment on a question that I think about a lot myself: What is information? What's the nature, the ontology, if you will, of information? Is it some sort of form of matter or energy or are we talking about something different here? Almost in a kind of cosmic sense, you know?

Bill Gates: Monday night I met with Doug Lenatt, who's doing this psych project and I talked with him about ontology, because in AI it's a big thing--how do you separate out knowledge. I hadn't thought about it that much in terms of office information.

I mean, there are some particular concepts that we're all pretty familiar with like sales transactions, budgets, forecasts, meeting schedules, trip reports. So we're all pretty familiar with the way of specifying information and certainly on the transaction side, MIS departments already know what that information is. They're just not providing the sort of ad hoc query or ability to play around with that data that ideally you'd like all sorts of managers at workstations to be able to have.

Now when you talk about making it easy for someone to move around in that data space to...let's take this corporation, you know, to move around and find trip reports that are interesting or customer data that are interesting, we have to organize it in a way that makes it obvious or makes it easy to get at that information in a variety of ways. Right now, it's pretty simplistic. You know the OEM data base is over here. You know the retail sales transaction data base is over here. You know, trip reports you sort of know, you sort of hope they're filed under that developer's name or something like that.

There's still some problems, and human communications sort of bridges that gap: to know where to go to get those things. We want to formalize all that stuff, and over the next year...right now we're totally connected with electronic mail so we're a reasonably good guinea pig in this thing. Our MIS stuff is pretty advanced in the sense that although it on VAX and is running sort of classic accounting software, we have the connection where any of these DOS workstations, using the Ethernet, can get up and get at that data, and in fact, they even do some special--instead of giving us the type of reports that usually managers get--they pull that data down in Excel spreadsheet form so any of us sitting on the network can get at that stuff.

I don't know if I'm answering the question.

Will Zachmann: Well actually, let me try to push it into a little different, uh, dimension maybe even. I had a conversation, I went to a thing, the Kleiner Perkins Presidents Conference about a year ago and I talked with Gene Kleiner. I noted that most of their successful investment are either in the area of information technology or biotechnology and after talking to him, it occurred to me that in a very real sense, those aren't radically different because it seems to me that really the breakthrough awareness that created modern biotechnology was precisely the recognition that the genetic structure had to be understood, not in terms of the conventional concepts of matter and energy in physics and chemistry, but as information; that the genetic code is precisely that--a code--and that it was the interpretation of genetic material, as information, rather than as the ordinary subjects of chemistry and physics that has opened the door toward gene cloning and all of these kinds of things.

And that seemed to me to evoke a much broader concept of information, or sense of information, than that of what we store in computers. I kind of get to wondering very fundamentally about the nature of information as a sort of basic constituent of what IS rather than strictly as something we manage business with. Does that make any sense to you at all?

Bill Gates: Well, actually, I mean, that's a long topic I could tell a lot about. I've spent a lot of time not related to my job looking at biotechnology, understanding, and I think you're absolutely right. A lot of good advances in biotechnology come now from taking an information theoretic point of view of how is this message sent digitally? How is this thing suppressed or reinforced as part of almost a programmed digital system.

And in fact, one of the most controversial and I think exciting, theories of evolution came from this guy, Doug Lenatt where he described...this is getting way off...but the rate of evolution in mammals is far higher than random genetic mutation could explain.

And of course there's all this information in DNA called introns which is not expressed as protein so it's not functional. Anyway, he came up with this theory actually about six years ago but it's just being explored seriously now. And it's clearly a right theory: that DNA has developed heuristics for knowing what sort of changes be adaptive as opposed to just random mutation.

Anyway, the idea that DNA is this so-called self-organizing system that's improved to a certain level and improving incredibly fast, I mean there's obviously something we can learn from that. Now, whether we learn it independently before we decode it from that, who knows? This is pretty distant from the specific stuff that Microsoft is working on.

Charles Simoni here was one of the original scientists at Xerox Palo Alto Research Center. He is just really in love with this thing and he thinks, "God, maybe in four or five years we can get some clues about learning software and heuristics." But I think not a lot scientists would hold that view of things, that decoding of heuristic techniques and DNA would proceed along that far.

But, you know, we're a pretty pragmatic company, I mean, our whole thing is office automation and the use of the PC workstation in office automation. I track this AI stuff very closely that was the Ph.D. work that I never completed or that I barely got started, you could say, when I was at Harvard, was in the AI field. In fact the IIIA was up here Sunday and Monday.

Will Zachmann: Were you an undergraduate at Harvard or...?

Bill Gates: I was both. I was an unsuccessful undergraduate and an unsuccessful graduate student.

Will Zachmann: Really?

Bill Gates: I was there 2.5 years, went and ran the company for seven months, went back for 4 months. So I got a total of 3 months [Bill apparently meant to say "years] of schooling. I was in a very unusual thing where I was taking graduate courses and undergraduate courses, but I didn't finish either course of study.

My undergraduate stuff was more economics, history, psychology and my graduate stuff was actually more the theory of computation those first couple of years because I wanted to learn about that stuff even though my intended thesis was the AI stuff.

Anyway, you know, AI breaks down into a lot of things. One of the key things is learning. Can you take a body of information and detect patterns in that information? Can you generalize from it? Humans are amazingly good at this stuff. We still can't figure out how humans categorize things so quickly, make predictions so capably. I actually think there are applications of that to office automation but then there's this whole problem of how much knowledge does the machine have to have of high-level concepts, like communication management, sales, budgeting. You need sort of an expert system knowledge of those things before you're even at a level where you can detect these patterns, and since we can't have the computer like go to school to have all these experiences, maybe we need to prebuild in a knowledge of how things work in the office and what the key office concepts are.

Anyway, in 3 or 4 years, that type of understanding on the part of workstations software will start to become important. I actually think that for the next 3 to 4 years that the things we've embarked on right now--large address spaces, graphic interface and serious networking, you know, corporate-wide electronic e-mail--I think the productivity benefits to that--once you get people convinced to use it--are pretty incredible.

But then in that next wave, some of the techniques of AI for letting you deal with the information in a high-level form, which does get into this ontology thing, now you got all this information, how is the human supposed to find his way through it and how is the machine supposed to talk to you about it in some form other than here's a field in a table, here's the nth character in a document, here's cell 813...

Will Zachmann: One thing struck me a long time ago the first time that I tried to get my wife to learn something about DOS. She had some experience with Adventure and Zork and games like that When I tried to get her started with DOS, what she immediately started to do was to type in commands to try to do things in an effort to get it to do what she wanted it to do, very similar to the way you would in one of the adventure games.

That formed a very strong impression with me which, it occurred to me that those games very much involved doing things that are in some ways like real life. You have to learn in that environment how to succeed in that environment, what works, what doesn't and so forth. but the games like that always seemed to me to be basically like many games are, there's oriented toward the artificial world created by that kind of game.

One of the thoughts that occurred to me about that was whether it wouldn't be possible to build games like that that actually taught people real things. In fact I tend to think of that as maybe one of the most intriguing possibilities I can think of what to do with software, even personally, in the 1990s. Does that sound like a possibility: to create learning experiences for people that are entertaining in the way that a game like Adventure or Zork is, but that actually teaches people something about something is perhaps useful to them to know or to learn about?

Bill Gates: Sure, it's very possible, I mean take, well take physics. There's this one textbook if you just sit down and read it, you can learn it. So why waste someone's time going to classes? Well, it's just not interesting enough. I mean i happened to be in the mood during one period and I learned a lot of physics just sitting reading the thing. But in general, it's not a good way, handing people textbooks, isn't a good way to expect them to learn a lot of stuff.

So, fine. you gotta make the stuff more interesting and having some guy standing up at the blackboard is a little more interesting and maybe having the program that interacts with you using images and sounds and building a model of you--you know, what does this person know, what don't they know?--okay let's illustrate that better...

Sure, that's possible and that's why Microsoft--take the work we've done with RCA on this Digital Video Interactive and the whole area of CD ROM where you want very large storage. We need to get the machine to the point where it can provide a rich experience and where's there's tools to set that stuff up.

But I always thought, you know, people said that they had educational software on a little cassette. Well, if a piece of educational software was short enough so that, you know, I could like memorize the thing in two minutes, it can't provide enough variety of response to keep a human interested. It would have to be extremely rich.

There's some good work in this area going on at Xerox Palo Alto Research Center in John Seely Brown's group on how the machine is a tutor. I mean they create these simulated lands, Mathland, where it sees the types of errors you're making and it's very visual in giving you feedback. I mean we're still quite a ways away on this stuff, and to some degree, you've got to be pragmatic. You've got to say, "Hey, what is it that counts in office automation?"

Well, awesome electronic mail, data base access, scheduling, project management, presentations, things like that, and how do you make those thinks compelling which is very directly relevant to us? Well, having a co-worker who is greatly successful or having some outfit that you look at it and you say, "How did you do that?" that sort of creates motivation in the environment of most importance to us, and so we have to use those things.

The experience that we can create, the software that's distributed on floppies isn't quite what we'll be able to do with CD ROM and yet floppies will be the primary distribution media for the next 4 years certainly.

Will Zachmann: If it seems like sometimes we're getting far afield, that's partly because I'm deliberately trying to lead us as far afield as possible. [laughing, Bill Gates laughs, too]

Marty Taucher: Well, it's sort of an interesting position. I mean, were also a pragmatic company and what we're focused now is really on sort of on getting people from the DOS world to the OS/2 world and Presentation Manager, and that's a tremendous confusion factor for people as it is, so thinking beyond that...

Bill Gates: Yeah, it's never clear whether we should be accused of being too pragmatic or too futuristic. I mean certainly in the case of Windows, the graphics interface thing, people accused us of being too futuristic. It's part of *my* job and you know, my personal interest to make sure that we aren't too short-term focused and the things that the development groups are really seeing all this stuff going on that's going to affect us 3 or 4 years from now: the parallel processing, distributed computing, incredible performance that we'll deliver and some more speculative things like the video or speech understanding--that those get factored in.

Marty Taucher: One of the things you might want to spend a little time talking about is just the transition that we've just gone from the character world to the graphics world and what that's going to unlocked for people. People tend to take that for granted nowadays.

Bill Gates: Well, who takes it for granted?

Marty Taucher: Well, I think we do, for one thing.

Bill Gates: Yes, we do. Sometimes internally, we say, "Hey we applied a lot of effort to that and that's going to happen, now let's look at the next thing." And we forget the excitement we had when we first sat down at the Mac and started prototyping our stuff up and seeing what we could use and stuff like that.

Will Zachmann: Well, I like to think that there are a lot of people in the world who recognize that sort of thinking sometimes more broadly is not necessarily inconsistent with running an effective business in the world and my aim was really to try and see if we couldn't talk about some of these more general things since I kind of assume that nearly everybody who wants to talk to you wants to talk about next quarter's earnings and the next series of announcements and so forth.

Bill Gates: Okay.

Will Zachmann: Let me try a still wider circles yet. One very interesting afternoon I had a couple of years ago was at a conference we had done at Silverado where in the afternoon instead of playing golf and tennis the way most everybody else did, Mort Rosenthal from Corporate Software, Alan Hald from Microage and Doug Englebart, who I'd invited to come up and sit in with us, and I went driving around the vineyards and talked about very wide ranging topics, politics and things like that.

But one of the subjects we got on was the question of the nature of and relation among information, knowledge and wisdom and, in particular, the notion of what information has to do with wisdom and what the two of them are. Is that a topic that rings any bells with you or that you have any thoughts on?

Bill Gates: Well...

Will Zachmann: What is or what makes a wise person or is there such a thing and what role does that play in the overall world? And how, maybe, does information technology have a potential to contribute? Are we going to end up more wise as a result of information technology at some point or is it strictly neutral in that regard?

Bill Gates: Well, certainly, having all this information on-line is going to make the rate of advance of knowledge much, much faster. Just look at what's happened over the last 20 to 30 years as all this information has become so widely available.

You know, back 100 years ago science sometimes proceeded randomly as people would write letters to each other or run into each other and things like that. Now, through conferences, libraries and these on-line systems, it's just a lot more efficient and that accelerates a great deal.

I mean, you know, look at...to take the bait a little bit, look at the human...take the visual system as an example. The number of bits that come into the visual system is absolutely awesome and at every stage, as you go through processing, the actual number of bits, the number of neurons, number of signals involved, drops off pretty incredibly, so our self-organizing system has learned to capture significant features and deliver them in a form where even though in your short-term attention you can only deal with a few, very, very few things, the appropriate things are made available to you.

I mean of the millions of bits of information that are being brought into your head, you just get something that says, "Okay, there's nothing important" or "He's getting up to leave." or something like that. So, it's very important to deduce patterns and being able to ignore the noise and not have to dive into the detail.

You know, certainly in my job, there's so much stuff going on in developing code. And this industry, do you know how many periodicals there are and stuff like that? I was wondering, you know, there's a few weeks when i read every single article, you know, and then there's a couple weeks when I say "Hey, I'm only going to read very little. do i do a better job at one time or the other?

Because information, not information, because bits will be so available, data, there'll be an abuse where we all just get so overloaded with data and there's just too much. But then of course, we're not just moving bits fast, we're also processing bits very fast, so the ability to differentiate out information and perhaps even knowledge is there.

The idea of getting the computer to help with this is partly us, you know, giving it rules about what the interesting patterns are, and then eventually, the idea of it being a learning device that can deduce patterns that allow you to make predictions and predictions in the case of humans lead to survival abilities. And it's a very good thing. Well, it's somewhat more artificially in the case of a machine in terms as to what the goals are, but with this incredible processing power, we'll be able to do that.

I happen to be one of these people who believes that personal computers have improved productivity a great deal, even in their current form and that making jobs more interesting really allows people to do a better job, to be more creative. The fact that it's so easy now to make a good presentation or do output like this.

Ah, you know there's some articles about how you diddle around and stuff, well, that's not true. I mean, it's true only in a limited sense. It's just like the people say "Hey graphical user interface is bad because somebody put too many trashcans up on the screen." or "Color output must be bad because I saw some fruitloop display." There's a backlash against everything and you know, Microsoft has been on the industry long enough to be ready that any wave that goes through will have a backlash.

So all this data on-line thing is going to be hit with that because it's a lot easier to apply the technology and deliver the bits than it is to figure out this stuff about how should we summarize this information and put it into meaningful form. You know all these managers who say they don't use computers--all they're saying is that it's too data oriented; it's not information enough oriented when it's delivered across the workstation and the interface they prefer to use is a human who works for them that looks at the data at that level and processes for that particular manager's information like: "Don't show me all the tasks in a project, just tell me is that thing in trouble or can I worry about something else today?"

Humans are extremely good at this taking data and turning it into information. Wisdom is a little vaguer to me. I mean, what's wisdom, abstractly? How do you differentiate? Any of these things are a little hard. In the AI field you can see them really making an effort to do it, but when you talk about...humans are very good at what you call the control problem. Given the idea that you're trying to find the solution to something, if you start to look a certain direction in the search base, you see, "Hey, I'm not making any progress. This is worthless." Or if you start to make some progress, you figure, "Ha, looking for this type of pattern, this is pretty good. Maybe I can play that in a general sense." So those are heuristics. That's what they're termed now. They're techniques that let you spend your processing time on the most important things.

Well, we're not at that level. We don't even have those concepts in computers yet, but we will very soon. Now people think that just because computers are fifty times faster, they can tackle these problems but if you try to use brute force solutions, you'll find that it's exponentially difficult to take a lot of data and refine it into patterns or information, and so you need a lot more, you need a lot of I guess you'd say wisdom about what types of patterns to look for and how to spend your time. Anyway...

Will Zachmann: When I was about five years old, that'd be 1947, I was given a book that had just been published called "Wonders of Modern Science" and it had two pictures in it that especially struck me. One of them on the left hand facing page was called "The Human being as a Multistoried Factory," and it tried to show a human person where it was all pre-World War II technology images basically, so that the base of the brain was a telephone switchboard with the old kind of plugs in it and the upper portion of the brain was the executive offices. The arm was a crane and so forth. On the facing page there was a picture entitled, "What Happens When You Touch a Hot Stove." It showed a woman in the kitchen, with a child playing on the floor, who had just touched a stove and showed the nerve impulses going to the brain.

Those pictures have impressed me a lot over the years in various ways, but in particular in thinking about information technology in organizations. Two things occurred to me. One is the idea that you could literally describe an organization like Microsoft or IBM or AT&T as a living creature, a biological creature, that has its life birth, death components and so forth on more than an analogy. But that second, if you do that, you can kind of turn the imagery of those pictures inside out and think about the information system as a whole of that organization as playing a very similar role to the nervous system of the body.

You used the example earlier of the production of the visual information from the eye in terms of what's relevant, say somebody's coming in or leaving. If you made those analogies, then two things would be suggested. One of them would be that the output of the information system isn't really information at all, but as in the case of the nervous system that the relevant output really is action and the enabling ability of the nervous system of the person or the information system of the organization to enable that creature to first survive and then hopefully thrive in whatever environment it works in.

But the second suggests that kind of transition we're going through now away from the mainframe, minicomputer or multi-user system model to what I've always thought of as the distributed resources system model of workstations and multiple servers, is that you really have to characterize that as a major evolutionary step, by analog at least if not a little more, in the nervous system of the organization. Does that seem like a useful way of thinking about things? can you take it a step further?

Bill Gates: Sure, but it's a dangerous analogy because what you're taking is something that we understand very poorly, namely, how information flows inside companies and how you can think of those as a...,or how the changes take place due to outside forces and comparing them to the human nervous system, which we understand very, very poorly, and so the parallels are a little dangerous.

I mean once we've decoded, you know, the brain and it's going to take us some time, then we could make this a lot more concrete. It's certainly true what you say about the best way to analyze the thing is to look at its outputs, although it almost sounds like behaviorism. There's always these waves in various sciences that touch on cognitive science where they'll all say for a while that you can't talk about concepts, you can't talk about inner mechanisms because we can't prove them, so let's just look at the outputs and that, probably because our ability to do analysis that we'd do normally, we didn't have enough science to do those things. I mean it was okay but it really has held things back, and it's still a problem that whenever somebody starts to talk about those concepts, somebody will say, "Hey, that's not worthwhile. You can't talk about that stuff.

Will Zachmann: A case of nothing-but-ism.

Bill Gates: Yeah, and it's not... I mean the... It's just a very hard way to describe things. I mean we talk about IBM, we talk about companies as being very energetic or having a lot of smart people. I think that's a valuable way to characterize organizations. It's very hard to characterize them by all the things we do. I mean what do we do at Microsoft? We write code. There's a lot of people who write code. You really want to dive down into our code and see what's different about it or is it easier to come meet with me to...sort of like cracking open the skull and, you know, try to understand what our strategy is.

It's a lot easier if you can get inside the thing, which we only, in the case of the human brain, have extremely limited ways of doing that today. We can look at a single neuron or we can do what are called lesion studies, and basically we're going to need better techniques to proceed on the decoding problem.

In any case, the point about micros being a major change in how information is available to people in a company and it really changes how that company works hopefully, I think that is a very key point. You get into organizations, you get a high degree of specialization, but a lot of that leads to just not letting people see the global picture so they can't contribute to it.

I mean, why are top executive so valuable? You fly them around in jets, pay them a lot of money, and stuff like that. Well, they have access to all this information which right now it's because you don't want to have too many people in meetings and primarily it's people sitting and talking to each other, so there's this one guy who just, you should, he's like the queen bee or something because he's had the chance to meet with all these people who hand him all this information, and his decisions have all the leverage.

But to the degree that you can have more summarized information available to a broader set of people, then it's in electronic form. It's not quite as bothersome as having 200 people come to all your meetings. You just have them sit out there on the network and be available. Maybe you can get more people thinking strategically, understanding the global considerations the company goes through, not just being told, "Hey, you've got your own specific thing."

You can certainly pull down enough data to play with what if? scenarios at the product level or look at pricing decisions. You know we get, we have, I would say, so much more bottom up creativity about strategy in an organization like this where the information is around than in most companies. Now, that may be also that it's more necessary in our business because it's a much more dynamic business. Nobody has really decided how things should be delivered, or priced, or what various roles various participants play in our industry, and so it's kind of dynamic and you can't imagine the top level managers being the only people who'd come up with those things in a company that has shown a lot of survivability in our business.

I mean, you know, hey, twelve years in this business is a long time. We've met at least that test. Steve Jobs used to talk about how this thing of getting computers down to the individual level really changed this thing about just a few people at the center really knowing what's going on and therefore, they're being leveraged to giving them greater authority, increasing authority, because they've got this great tool, as opposed to everybody having the tool. And as we've seen the convergence of the amount of processing power on the desktop with what you can have centrally, we're just going further and further to avoid the individual worker not being able to have that same information and that same analysis.

Will Zachmann: One of the things about the picture on the left-hand page of my "Wonders of Modern Science" book that struck me as being correct in a way, you know, even though it's quite a few years old now, is the notion that the upper part of the brain is not the computer room, although there wasn't one at the time the picture was drawn, but the executive suite. That is to say that the information system of the organism which is the organization is not strictly defined by the electronic components, that if you try to characterize...if you use the analogy, try to characterize the information system of the organism which is the organization, that in fact it's a hybrid electronic/organic system where in a way its most critical and most expensive parts are the organic nodes, if you will: the managers and professionals and executives, scientists and engineers, and the clerks and so forth that provide the organic information processing components.

I mean because, in a sense, isn't decision making an information processing process? When we talk about decision makers, aren't we really talking something...if information flows through the organization, it flows through the minds and the perception of the individuals as much as it does through the wires and transistors.

Bill Gates: Yeah. Andy Groves book does a good job of trying to take a pretty cut and dried, you know, factory output-type view of the manager's role. And what is it if you say one manager is doing quote better unquote than another manager, what does that mean? He gets down and talks about that, talks about therefore how does delegation, too much or too little delegation, either improve this metric or reduce this metric? Pretty good thing and I think that sort of cut and dried analysis on managers makes sense. It's the type of cut and dried analysis we've applied to software, our software factory and it's been pretty useful.

You're talking Gardiner's books, "The Minds' New Science," where he summarizes various fields of science, how they've contributed to cognitive science. He says at any point in history our model of the brain has always been whatever the latest technical marvel is. At the time when they had these fountains that were really cool, they thought, "Hey, it's just like these, you know if we just got enough of these little wheels and levers and then we could understand the human mind." And of course, now it's the digital computer.

And there's good reasons. I mean, it's a better analogy than there have been in the past. The managers of an organization, yeah, they're there to make decisions and they *are* the information system and the degree to which this electronic information system can get in there and help them, I mean, right now it's super limited.

Ask somebody how they made a business decision and how key a role the computer played in it. Oh sure, the computer keeps track of forecasts and budgets and prints those things out, but even those things, you know, somebody sort of anticipated the results by setting a fairly high forecast because they thought, "Hey, it's a good idea to do a product." The computer's not that involved in the decision making process, although it will increasingly be, and I think organizations like this one are a little bit further along in that than most.

Will Zachmann: Let me try a little different line of asking about things. Just in terms of your own way of looking at things, doing things, who are people who you would point to as having had a major influence on the way that you look at things? Individuals directly that you've encountered or intellectual influences? What are the most important few, if you can distinguish any, sources of the way that Bill Gates looks at the world and sees things?

Bill Gates: Well, there's the stuff outside the industry and then there's the stuff sort of inside the industry. Outside the industry, I mean, when I was fairly young, I did a lot of mathematics and this notion that things that are true, you know, that you can prove them, and things are very cut and dried, there's no sort of...just a very scientific view that everything can be explained without resorting to mystical influences. Intuition says to you that...that...it sort of leads you away from thinking that there's incredibly clever solutions but then in mathematics, you realize, "Hey, there's these things where you can think about them for years and then there's a one paragraph result that comes out of all that thinking, and that it's almost an information theoretic thing that there are very simple...often very simple solutions to things, and that if you're energetic enough about them and stare at the thing long enough, you could continually improve on those.

You know, I started playing with computers a lot very early on and having people look at my code and show me the same thing, that is, that writing a piece of code that you think is super good, then give it to somebody, they can look at it and just make it a lot better and a lot smaller than that, and so there's almost no limit...well, there is a limit...but there's less of a limit to these things than you'd normally think there is.

Then I started to do things in a company sense. There's a lot of people--people I've worked with super closely, my co-founder, Paul Allen. He and I really collaborated on all the original vision and plans and stuff and we'd just sit and talk about it, and read and read about it, and kind of argue about it. Collaboration. Guys like, oh, Steve Jobs, who sort of shows that leadership inside a small group can create a lot of energy and a lot of dedication.

I've met a lot people who quoted me as that but I'd say he's certainly one of the best at that, a phenomena. There's guys like, I don't know if you've ever met this Kazuhiko Nishi, he was are guy, he's my guy in Japan. I'm not working with him at this stage, but he is a real energetic visionary. I mean other than Paul--I've worked most closely with him.

And now, it's a lot broader, the company's a lot bigger. In terms of my business experience, it's really John Shirley, and Ballmer. Ballmer's a guy I've sort of grown up with you could say. We were in college together.

Will Zachmann: Is Paul [Allen] actively involved with Microsoft at this point?

Bill Gates: He doesn't work here. He actually has his own small company called Asymetrix that's doing some AI stuff that's completely unannounced. He's still very involved in talking with me about strategy and what's going on in the industry. He lives a few miles from here. He's very, very involved in the industry. In fact, he just went on to the Egghead [Software] board and he has an investment in a company in Boston called layered software. He still owns 20% of this company, [laughing] not an insignificant investment!

Will Zachmann: Looking back say, from the perspective of a hundred years from now and somebody looks back and wants to characterize you and Microsoft, at what role you played in what at that point I guess it gets to be human history, what's your...how would you like to have that come out? How would you like to be characterized? How would somebody describe what you've done and what Microsoft has done from that kind of time perspective? A hundred years out?

Bill Gates: Well, it's probably a pretty minor thing at that point. You look out a hundred years from now and what you say is some guys discover transistors and the ability to miniaturize those transistors and that led to this ability to do computing, very, very inexpensive computing.

[brief break to other side of tape]

Bill Gates: [continuing] ...company's contribution. You've got to say there's a certain naturalness to these things that are happening and an individual can make them happen a little faster and in a little bit better way but these are natural trends. The magic of the semiconductor industry is still...and the incredible general nature of software, and the fact that we can improve it so much. Those are what are really pushing this stuff forward.

Will Zachmann: Let me ask you to go a little further in terms of how do you see that trend? I mean, what is the broader thing that is going on and of which this a part in terms of new technology and how it affects life on the planet? I mean, is this an incidental byway or is this...Marshall McLuhan made, I think, a very compelling case to suggest that things like the invention of the alphabet or the adoption of the phonetic alphabet by the Greeks or the development of the printing press have had far ranging ramifications that go way beyond the immediate and obvious impact of that technology. It would seem to me that if Marshall Mcluhan were writing today, he could probably make a pretty good case that personal computers, well information technology, but specifically personal computers and their widespread availability will have potentially equally widespread ramifications. Does that seem plausible and if so, where does the rest of it go or where might it go?

Bill Gates: Well, okay, we'll that this from the farthest out. The best predictor of the future, the best book I've read on that stuff is a thing called "Engines of Creation" by Drexler. It came out about four months ago. Incredible book! Incredible book! It talks about how the information revolution and biotechnology in combination are just part of a trend and he extrapolates that trend somewhat in a very plausible way over the next, oh, fifty to a hundred years. And it's pretty amazing.

There's this guy, Dawkins [sp.?] , who's written a lot about evolution. he wrote this thing called "The Selfish Gene: Blind Watchman." He talks about, okay, evolution are these self-replicating systems. Well now...and genes were the method of passing this along. Now he talks about menes, the idea of ideas that the human mind is sort of the host for those ideas and there's a competition for which ideas work well, and are effective, and so those ideas contain within them the seeds that allow them to be passed along to other humans and therefore survive and be somewhat modified and things like that.

I mean this is the very broadest view and electronic storage lets those menes, which now in some cases are actually like programs, and eventually self-learning and even self-organizing programs. It gives them a host in which to replicate and, you know, eventually, you can get a system that is pretty impressive and perhaps is moving at a pace far greater than gene-based improvement for replication. Anyway, that's the take of it at the highest level which, I think, is interesting and even somewhat scary.

To move down, you know, one level from that... Originally, we were designed to learn from our, just direct experience and that was all that we had that you could learn from. And then you get books and you can learn from a lot of other people's experiences but it's still kind of slow. The information gets written up a lot later and you can only read a certain amount of that stuff and there's no device to sort of summarize it to a level beyond that. It's sort of random like books you happen to read. Hopefully you know, people who are good choosers or something.

Now with all this information on-line, the ability to react to situations as a corporation or educationally, I mean it's just a whole new thing, and i like to think, you know, corporately I like to think that. sort of, the clock ticks faster here than it does at other places.

I mean, I really hope so. 'Cause we're all just programmers and, y'know, writing the same way, and eventually people will ship products and their users will tell them they're crummy and they'll get it right and it's just only by sort of accelerating that process, better tools, more efficient feedback, little bit better vision, that sort of intuition.

Will Zachmann: It's almost like in some very broad sense becoming a kind of conscious agent of evolution in a very broadly defined way. I mean, evolution in the sense of whatever the overall flow or direction of what is, is. I don't know, it sounds kind of mushy when I put it that way, but...

Bill Gates: You've got to think of this stuff as anything that's a surviving sub-replicator is what, you know, is what'll get pushed forward and it has nothing...I mean that's it! I mean there's no purpose. As long as you're in an environment where there's enough resources for you to go on doing that replication, great. I mean, there'll be a tendency for other things to come along that take advantage of you, as you get to be more numerous.

I've always got this temptation in this conversations to go back and talk about the whole evolution thing. But its, I mean, there are no very direct parallels. There's this guy who works with Scott Oki at Micros [?] who's writing this book about how, you know, business competition should be thought of as evolutionary struggle but that's not a very popular metaphor, nor do i think it's a very...a metaphor that helps you all that much in terms of specific decisions because, it's another one of these things, we don't know all that much about evolution either. I mean, we know a little bit about it but...

Will Zachmann: Actually, that's a metaphor that I personally find very appealing. One of the things that I've thought about a lot is the...an analogy I've used a lot really...because I tend to present things in terms of the idea that what's going on right now is a transition to what's really only the second generation of information systems. Where the first was that of the mainframe, multi-user/mainframe minicomputer systems where the paradigm has been the same up until now and where now we have a completely different paradigm beginning to establish itself.

I've often thought and used in talks and speeches the analogy of the meteorite hitting the earth or whatever it was 65 million years ago that changed our climate and killed off the dinosaurs because, in a very literal sense, it seems to me that the traditional multi-user mainframe and minicomputer systems are in effect dinosaurs compared to the kind of fertile mammals of PC workstations, networked with a multiplicity of servers, which looks to me like a more adaptable technology life form, if you will.

Bill Gates: Well, eventually you have to throw out the old and start with the new, and it's not clear when you decide to do that, and you know, micros as a cheap form of computing, very distributed, is an opportunity for some people. Even if you ask some of these MIS guys, you know, do you want to convert everything today, they'll say, "No, never, never!" But the truth is they *are* starting to move their stuff down onto micros. They *are* starting to distribute it. It's just that they don't want to do it all at once. They're being relatively pragmatic about this stuff.

I think a lot of people are going to restructure. The notion of applications development or how you get information out to other people inside the organization. It *is* happening. PC's are selling quite well and it's not just running word processing.

Stuff like this united airline thing--it really strikes home when banks and airlines and other people who do transaction are starting to say, "Hey, out with the terminals and in with PC's." And they're becoming totally dependent on these things. That's big stuff.

I mean, we've worked with most of these airlines in terms of their software, like the United one is Covia, who's doing that stuff. Man, that's important stuff to them, making their schedule and having the stuff easy to use and they view that as the center of their competition with American Airlines and lower profit margins. This will let them offer more information and just the help we've given them in terms of working with Windows. You know we have a good relationship with those guys.

When you deal with this mission-critical stuff, people are very anxious to get first-class answers. You know, understand the systems software, stuff like that. Definitely it's an opportunity to get rid of some old mistakes that were dead ends.

Will Zachmann: One of the really intriguing things to me about OS/2 in particular is the way that I think it is already beginning to radically the change the perception of personal computers on the part of information systems professionals. Of course, there's the interesting history on how exactly OS/2 got to be called OS/2, but within in the world of IBM software and IBM-related software it's really quite intriguing because I, from the very beginning of looking at OS/2, had this incredible sense of deja vu when I thought to the first times that I'd opened the 360 Principles of Operation in 1966 and the first time I looked through the Supervisor and I/O macro Instructions Manuals for OS in '67.

One thing that I'm finding in talking to the information systems professionals, the computer people, is that they're all of a sudden in much more familiar territory. In fact, it's in some ways it's more familiar to them than it is to the people who started out using personal computers and started out using DOS. The concept of a multitasking operating system and of memory allocation and all those kinds of services that are provided by the API in OS/2 are the kinds of things that somebody who's been in IBM systems, not just a minicomputer Unix people, but even more the mainframe people, right away you feel like you're...

In fact, there's a column that I just wrote yesterday that will be in Computerworld, I think next week, I compared it to being like in your own neighborhood, if you go back to the neighborhood you grew up in. You, all of a sudden, feel, I mean, you realize it's different, it's not exactly the same cause no things are exactly the same after 20 or 25 years interval, but it's a very familiar world.

I've always felt that personal computers would eventually not only not stimulate the demand for larger systems but in fact, tend to replace a lot of that demand. You really can see that now, both in terms of the way people are starting to put serious applications out on microprocessor platforms but also in the way that some of the more traditional data processing people who are a little wary of it are now starting to feel at home in it and starting to feel like it's a world that they can deal with and they can build serious applications and do serious things with.

Bill Gates: I think there is some similarities in the evolution of what's taking place inside the individual's computer, to what took place inside the organization's computer. I mean they went from running one application at a time... They actually went through quite a few things, where they went to fixed partitions and variable partitions.

Will Zachmann: That was the original OS/MFT and MVT. Interestingly enough, they were strictly defined in terms of multiprogramming, no mention of multi-users since that was a batch system. But MFT and MVT were exactly that: multiprogramming with a fixed number of tasks and multiprogramming with a variable number of tasks.

Bill Gates: So we're skipping... Going from MS-DOS to OS/2 is like going from, like the original 360 operating system sort of up to VM level. So, I mean we do have the benefit of a lot of experience or you could say cheaper memory or whatever. You know, it's not exactly the same. You know, there's no analogy to the graphical user interface and trying to preserve the simplicity.

But it is true that the people who are jumping on OS/2 earliest are people who are either at the mini or mainframe level, are used to an operating system that really is in control of resources, as opposed to MS-DOS which, yeah generally people are nice enough to call us through the file I/O, but we're sort of a subroutine library and their application decides what should go on in the system.

That's an unstable situation, if you ever want to have multiple things in there working together and unfortunately, the original Intel architecture doesn't give you any way to protect between two things. And in any case, we eventually decided that the 640k limited or one megabyte, whichever you say, architecture: it's not that worthwhile to try to make multiple things fit in there anyway.

So the move away from that architecture and move to really having the true operating systems come at the same time. There's certainly this confusion created by the fact that there's this 286 level and this 386 level and there's still people who don't believe in graphic user interface and it's confusing because we're not shipping that out at the same time that we're shipping other stuff.

But in any case, I do have a tough time with people...I do...the fact that people are debating the details now sort of means a lot of people are in agreement on the general things, which is that a true operating system *will* make sense at the individual level so you can run multiple things and a lot of that user interface stuff is shared. Even, you know, I'm sure there'll be articles that praise OS/2 and articles that are negative about it, but it is moving us in the direction that things should go.

Will Zachmann: I've suggested in various contacts lately that perhaps the most important aspect of the April 2 announcements from IBM and from Microsoft isn't really OS/2 directly at all and isn't PS/2 at all. I've suggested that the most important part of the April 2nd announcement actually was first delivered in 1985 in that it seems to me that the part that's going to have the most impact on what users do directly is the effect of IBM, in fact, making Windows part of their systems application architecture interface and defining that as a major commitment for them for the interface for the future; where the multitasking in OS/2 gives a nice way of enriching that.

But it seems to me, at any rate, that the most significant step is this incredible broadening of the bandwidth of the user interface with Windows and with IBM's commitment to it and to including it in SAA. It seems to me that...as well as the maturation of the base of software and so forth starting to get serious...it seems to me that that's probably the single thing that's going to have the biggest impact over the next few years. Is that...?

Bill Gates: I totally agree with that. Their...OS/2 over time has a huge impact but that was fairly predicted...that they would do an operating system that would tap the 16 megabytes and all that. The *big* move is the move towards this new user interface and if you look at what they specifically did in the PS/2, what did they do? They gave you higher resolution and they cut off the machine that was too slow to do this type of user interface and they gave you a mouse port. Well, that's pretty radical for IBM to be tuning their hardware to the graphic user interface, whether it's Windows on DOS today or Presentation Manager as part of OS/2, that hardware is great graphics hardware.

Also the idea... You know, Microsoft has always had this view we'd get everybody to jump on board consistent user interface, but that's a little tough for us just to do as a software company. Now having IBM behind that, not just in their PC line but as part of their grand strategy, is going to make it come true relatively quickly. And then there's two parts to play here; one is that it's sort of officially blessed. The other thing is you sit down and play with an application and you know, then you just get used to it and then when you get the next one, you feel bad if it's not the same.

That's certainly had its impact on the Mac. You know, the marketplace has soundly rejected those applications that have come out that've not been true to the interface. I think it's this user interface thing that going to have the biggest impact and in fact, some of it is even independent of OS/2 because a lot of it will be on top of DOS.

Certainly, our focus of development, what you hear from [Jeff] Raikes is towards getting these things out initially on real mode Windows on top of DOS and creating some real excitement. What happened on April 2nd enables us to do that. It gets rid of this thing where whenever you talk about windows, people say, "What about IBM?" They can still say, "What about the 8088?" but it's a less damning problem than it used to be. They've enabled us to really carry forth the message about the user interface. They themselves, of course, will do all their new applications in that form, but, I mean, I'd be surprised if they did that fairly soon. I mean as far as in print goes, I'd say I'd be surprised but I'm pretty sure...I'm quite certain it'll take them longer to get all the pieces put together. So, the truth is that we're at the vanguard of pushing out this new user interface stuff. In fact, you know like Quick C and Quick Basic 4 and Works and Excel will be the first applications that are really following that new user interface.

Will Zachmann: Just drawing back here to the question of information, something that I've gotten to think about a lot...we're kind of constrained by time and by the context of both of us being aware of the tape recorder, but sometimes, if we ever have a chance to talk about it, in fact I'll send you a copy, in fact I meant to send a copy to Gordon Letwin as well...something I've gotten into a lot is to look at the nature of information and to run that back against the concept of causality in Aristotle and, well actually, in Plato because there's a segment in, I think it's in Theatetus, where he talks about the nature of causes and he tries to characterize causality in terms a four-fold concept of causality, which in medieval scholasticism is conventionally discussed in terms of formal, final, material and efficient causes and a kind of modern reductionism that is there is modern behaviorist psychology or actually characteristic of modern pre-quantum physics which still dominates the way the people think about it, is an attempt to reduce everything into basically the Newtonian notion of what is, is matter in motion and what's to be explained is why it's moving and it's basically moving because something bumped into it. It's kind of that elementary level of things.

The way that Aristotle, actually I guess it was Plato originally characterized it, he talks about silversmith making a goblet out of silver for the purpose of being used for, say, drinking water or wine or something like that and in terms of the causes, the material cause is the matter itself--the silver that's worked upon. The efficient cause is the silversmith who shapes it. The final cause is the purpose for which it will be used: to drink water or wine. And the formal cause is the concept or the idea, the pattern, the structural concept or whatever it is.

When you're talking about "concept" there's an overlay of 2,000 years of philosophical thinking that distorts it. What I think basically happened is through the filtering process of history, we got into the notion that final cause is abolished. There are people like Tolman who ran rats down mazes to try and prove that rats have purposes in a reaction to the kind of Watsonian behaviorism, so that got X'ed out. The formal cause has long ago been abolished so that we get basically into the efficient cause and the material cause, where the efficient cause is energy and the material cause is matter.

And that, I think, actually is a pretty good characterization of the kind of dominant pop ontology, if you will, that characterizes our thinking today. One of the reasons why I'm very intrigued by the nature of information...See, it seems to me that there is a kind of phenomenology of information that has to do with structures, the concept of a mene...there's some fashion in which structures remain the same across completely different physical embodiments.

I have a... My proof of the independence of information from matter and energy is that if we decide to have lunch next week, I can send you the message, the information, that yes I'll meet you for lunch next week with distortions of sound waves that impact on your ears, I can make marks on paper. If we agree to the appropriate conventions of coding and decoding--the heuristic, the interpretation of it--and i had the resources to do it, I could send nuclear tipped missiles to the moon that would explode in a pattern that you could look at through a telescope that would say, "Oh, yeah, Will's going to be meeting me for lunch next week."

I would maintain that in just an ordinary language sense, the information is THE SAME in all cases. It's embodiment in matter and energy is completely different, but in each case there the notion of some sort of communication, of some sort of encoding and interpretation. Heuristics, very broadly, interpretation...when you talk about the heuristics of a program means also that process of interpretation which is an essential part of this nature of information.

Now I think that, in a very real... Partly, I'm influenced by a lot of philosophy--Heidegger, in particular--but lots of others as well. I think that in a sense we've kind of lost track of the foundation of what is, the basic connection with what is, and one of the forms in which we've lost track of is losing track of the final and formal causes.

What particularly intrigues me about information is that information seems to be very closely related to the notion of the formal cause, to the notion of somehow a structure or pattern that exists independently or at least that in practical terms you can talk about existing independently of its embodiment in matter or in energy. This same information can be represented as a series of energy transformations, it can be represented as a static condition of matter or of different static conditions of matter. What you need is the appropriate interpretive mechanism.

The growth of a fertilized embryo is basically a matter of that condensed information that's in the genetic code having the appropriate interpretation mechanism, which in that case is the amniotic fluid and the availability of certain nutrients and so forth. But it seems to me like a very intriguing...it's not the kind of question that has an answer but the kind of question you keep thinking about and find more and more aspects.

And it seems to me that at least there's a possibility that there's something much more interesting, you know, having a much more broad implication for the way that we even conceive of who we are and what we are and what we're doing here, that potentially could fall out from the stimulus to the thought about information that modern information technology provides.

I guess that's getting about as far afield as you can get, [laughing] but it seems to me to be a very intriguing area for thinking about in one's spare time anyway.

Bill Gates: Well, interaction of all this idea of are there concepts, what type concepts in the human mind, how does the human deal with information and then this whole thing where people say it's illegitimate to talk about concepts, which I think is total nonsense.

Will Zachmann: I couldn't agree more.

Bill Gates: I mean when we agree to meet for lunch, I don't have some reaction arc, you know, to the idea of lunch. You could get that concept in my head so many different ways and we can discuss it. Anyway, the parallel...the benefits that we may get in terms of processing information, seeing patterns in information, making it easy for people to interact with a machine from all this study that going on--cognitive science and the whole thing--there's going to be some very interesting interplay there.

I'm not, I don't follow the philosophical side of the thing at all, because I just, I don't know, I couldn't state my philosophy but it's a very sort of scientific one that says even if those other ones happen to be self-consistent and they work and stuff like that, I haven't finished working out the scientific--taking all the evidence and just coming to good scientific deductions. And those other ones, there's less formal procedures for deciding whether something is right or wrong or valid or stuff like that, so I happen to like this one a lot better and I don't want to read about those other ones. They might confuse me.

But all the other sciences are making very interesting contributions. The truth is that a lot of this stuff happens slower than we ever think it does. I think that it's a real interplay between those two fields. It is probably ten years in front of us. It really hasn't had that much impact today. I mean, you know, the way we program and what it means to use a spreadsheet or word processor and data base or operating system really has not been much influenced by the work in psychology or even AI type stuff.

Will Zachmann: I'm hoping to accelerate that timetable a little bit, because I'm thinking very actively about how to use the kind of software platforms that you're creating now in the timeframe of the early 1990s to start creating more efficient learning experiences for people aimed at the kinds of things that people might really want to or benefit from learning. That is to say, not simply taking a textbook course and trying to figure out how to put it onto a computer, but thinking about what are other kinds of things that people might learn about: things like how people work and think and do things or even things like even biofeedback.

I mean it's sort of an off-the-wall thing in many respects, but one of the things that really intrigues me and a product that I'd love to see developed by somebody and if somebody else doesn't do it, I'm going to figure out how to do it myself, is an inexpensive biofeedback monitoring device that will monitor brain waves--do a basic analysis of the harmonic components of brain waves--and provide visual graphical feedback on the screen of the current state of your brain wave patterns from the left and right hemispheres which is actually a fairly straightforward digital signal processing application that would have a fairly straightforward interface with some software which you could then use to drive application software that would give you the visual input that in fact maybe could be used to stimulate the brain wave patterns which maybe is only a curiosity, but then electricity was only a curiosity in Ben Franklin's time, I guess.

Bill Gates: Gross electrical phenomenon from the brain, we don't know much about it. We do know that if you get a little more localized, things like alpha waves, you can of course teach people to do those things. You can teach epileptics and make them learn to cause seizures by looking at the electrical signals that lead up to it, learning somehow to reinforce it and stuff like that. It's kind of scary stuff.

Will Zachmann: Well here's a lot evidence--a big body of academic evidence now--that shows that there are what we normally want to call states of consciousness very distinct associated with brain waves patterns that have a lot of very distinct physiological concomitants in terms of blood pressure, blood lactose levels, all kinds of things like that that go with it. But right now the devices for teaching and learning that kind of biofeedback...basic galvanic skin response is pretty easy...but stuff that really gets into doing an analysis of the harmonic components of the brain waves from the hemispheres and providing feedback on that is still pretty much an expensive laboratory--it's in the position today that music synthesizers were 15 or 20 years ago, still in the domain of specialists.

Bill Gates: I think you're right. I think there'll be some pop science that'll relate to that stuff, because the truth is we don't know what those... I mean, hey, that Scientology stuff--they still use galvanic skin responses for pop science. So fine. They should move up here and those gathering that information--explain that information--and then that information will get real cheap and then hopefully there'll be some legitimate science around too, where we really do figure out what do these gross electrical measurements--what do they represent? And why does it seem like in some cases you *can* correlate those things to real stimuli.

Will Zachmann: Well, certainly, one possibility that seems highly probable, I would suggest, is that the human brain is modal, I mean, in other words, it has distinctly different operating modes. Sleeping and waking modes are clearly different. I mean, there's no doubt about that, that in some way the operation of the brain is different. And a lot of this biofeedback stuff on meditation of one sort or another strongly suggests, at least, there are very distinct modes of lots of physiological concomitance that can be entered into and that have certain characteristics.

Bill Gates: Yeah. When you're trying to decode a program, it's nice to think that it's not all local states, that there is some global state. It gives you a handle on starting to decode that damn thing and fortunately, this one does have some global state and you can look at how does that affect the various areas. Why should there be this notion of the brain needs sleep? What sort of clean up process is going on? And stuff like that. I mean this is super interesting stuff.

It you're really interesting in this learning thing--this idea of lifelong learning and using it, I really encourage you to go see John Seely Brown's stuff. I think that they are really at the vanguard of that stuff right now and they get a lot of money from Xerox, and fortunately, they can do that stuff. They are actually looking at very practical things. They are not trying to model in the computer. they're not trying to make computer alone, they're trying to have the computer model enough of the imperfections and what's interesting to keep people wanting to try to learn, to make it a valuable tool in the near term.

Also, if you do get a chance to read that book, "Engines of Creation," it's the best!

Will Zachmann: I intend to. Another book that people have been recommending to me a lot lately is a book by Flores and...

Bill Gates: Oh yeah, Flores and Winograd? It's a complete horseshit book.

Will Zachmann: Oh yeah?

Bill Gates: ...but it has to be read because it's fascinating that Winograd, who when he was young, was a great contributor to AI and certainly believed what I continue to believe...that he ran into this crazy Chilean guy and got convinced that hey, it's all too complicated; there's too much context; we'll never be able to simulate the context, all the dead ends, all that nonsense is in that book. But the book, the book is bizarre because then the book turns around and starts to promote this ah, well, this thing called the Coordinator, which Action Industries which is Flores' company has, which is this e-mail package that doesn't just send messages. It tries to interpret everything as a commitment or a completed commitment or... It's really weird. They say, okay, we've now proven that computers can't really deal with real concepts because it's very complex but what shall we have them do? Well, let's have them run Flores' little package and help us communicate because humans are cool and if we can help them communicate, maybe well have less wars and stuff like that. It's in my view a little bit of a fruitloop book, but it's actually a pretty short book and absolutely worth reading.

Will Zachmann: That's what I've been told, that it's short enough that you can read it in an hour and a half or so. In fact, the folks who recommended it to me, I spent the day with the information systems people at Aetna about two weeks ago, and one of the people in the centralized information systems planning group was very keen on this. I also had the impression that it might very well be a fruitloop book because one of the things they were talking about was that according to Flores version of things that the CIA had engineered the overthrow of Allende's government in order to prevent Flores' experiments from going on or something like that, which struck me as a rather intriguing interpretation.

Bill Gates: He's a very strange guy--very energetic, smart. Most of the stuff I'm reading now is that the data gathering level. Real information about what's known about evolution, you know, learning theory stuff; going back over Piaget's stuff. Then, actually the best book recently, but this gets kind of hard core, on how computers can learn is something called "Induction" that a guy named Holland up at the University of Michigan did.

It's the best explanation of how humans try and categorize things and given contradictions to rules that they, predictions that you develop about how things are going to happen, how you react to those contradictions, how many takes, how you partition that off, how you try to differentiate those cases from these other cases. It's super well written.

Will Zachmann: "Induction" by Holland?

Bill Gates: "Induction." That just came out like a month ago. I'm really in...there's a couple areas where I know the publishers pretty well. Like wantons original book on biology, when it first came out...

Will Zachmann: "Double Helix"?

Bill Gates: No. What is the name of this book? [going to bookshelf] "The Chemical Biology of the Gene." This is Volume 1 of this thing. I have this here because I have to give it to Gordon [Letwin]. But anyway, this was first written just by Watson and he wrote editions one, two and three himself. And edition three came out about 3 years ago. This one is over twice as big and he had to bring in four collaborators to write it with him. They had to break it into two volumes. Volume two isn't coming out for like three weeks or something, because it really talks, this is super up-to-date stuff, it talks about the immune system and protein folding and all these things they figured out about that. But it is such a good illustration of how complex the stuff is becoming. This book is so much larger. The third edition was real simple. Unfortunately, the third edition simplified things that aren't true at all. Anyway, so they keep delaying this volume two thing because new things are being discovered. It was supposed to be out like three months ago but it'll be out in less than a month.

Will Zachmann: It's a very exciting time these days. I can't recall any time...the late sixties were equally exciting, but in a very different way. It was sort of externally exciting. It's more internally exciting right now. Just look at all the kinds of things that are going on.

I very strongly believe that we're in a key edge situation right now, a key transitional period. I mean, I think that's quite evident in the use of information systems in business organizations, which is my primary business concern and yours. But I think that it really does have a lot of overflow effects that probably ultimately are going to be a lot bigger than *just* the evolution of the use of information systems in the business organization. Which I suspect means that organizations are going to be dramatically changed within the course of the nineties as well, because I don't think that again, using heuristic analogy, I don't think that you can evolve the nervous system of an organization dramatically without evolving the organism as a whole.

Bill Gates: It's hard to see what that's going to be. I haven't read any really good stuff about that. I mean people talk about some of the small symptoms like, you know, work at home, or...they don't really talk about what it means...it's unclear whether it means bigger companies are better; it's unclear whether it means having all this information available, whether that means that decision making can be more centralized because you get a few quality people at the center or it means it should be more distributed because you've got information out on the leaves. It's not clear what sort of impact it ends up having.

Our key thing right now is just taking the organizational structure as a given, which over any two- or three-year period is the way you've got to look at these things and then seeing how we can make these things as easy to use as possible. This whole thing about the same sort of relationship people have with IBM, really knowing that over time they're going to solve problems very well and behave super well, even if individual products aren't ideal or something like that. Can Microsoft create that same relationship with a lot of large corporations as far as how we're evolving software, productivity software, user interface software? Can they see the role we've played over the last twelve years and view us as a partner in this? And the whole thing about what that means about a family of products, the networking software. All these guys in the field are trying to explain this complex stuff for us. That's a big business challenge.

Will Zachmann: That seems to be a big change over the last year or so. It seems that Microsoft is getting much more involved, conceptually at least in terms of idea exchange and so forth with end users, I guess, really, than ever before.

Bill Gates: Oh yeah! We always knew that would be sort of a valuable thing but until we really got super serious about networking, until we had OS/2 to talk about, until we had the family of applications that represented this graphics vision. I mean there really was no reason to build what is a fairly time consuming and costly infrastructure and really go out there. It's better to do it when you've really got something to say and now we've got something to say.

I'm sure that a lot of these companies that sell to these large corporations, when they say, "Hey, come out and do a headquarters visit." you know, you can't get the top guys to come out and it's hard to get them to come out. In our case it's almost the opposite. We have to pick very carefully. Like today, I'm meeting with Kodak. You know, we try to make sure that we're talking to the right ones to help our educational process.

Will Zachmann: Is there a possibility of Microsoft spawning a new business unit to be a systems integrator, sort of the EDS of the micros in large organizations' revolution here?

Bill Gates: Well, in a formal sense, no. We'll never take a cut of hardware revenues where we're going in with that server box, but in the sense that what is the key profit element or complex element in systems integration of software, well the Microsoft SE is going in there and saying, "use OS/2 like this, use Lan Manager like this, Excel can pass data up to your DB2 data bases like this and then saying, "We'll sell you (we don't do this at this stage but) we'll do some consulting or development for you is probably a natural evolution as well as sell you those products. Then it's like being a systems integrator and so in the same way, do you really call IBM a systems integrator? No. they're a special case for everything [laughing]. We'll be a special case for software.

Will Zachmann: Well, if I'm a Fortune 200 company, WFZ Inc. and if were to come to you today and say I've got a big project I want to use personal computer workstations and local area networks and servers, develop this and build a serious application, are you prepared to respond to me and say you'll help me?

Bill Gates: We can't do that. We'll brainstorm with you, yes, but today, we're not in a position--we won't be for probably two to three years to really say, "Hey, we'll put together a complete plan for you to do that." Today, we talk about how you should OS/2 vs. DOS, how you should move data back and forth between the mainframe. People come to us and say they want to do corporate-wide electronic mail and we have a clear view of how this can work--how you can find other users and printers, have all that stuff converted. So, we're doing a good job helping people with that one right now. but anyway, our goal would be to be able to take an even stronger role there, largely driven by the opportunity to sell our software products. IBM's driven by the opportunity to sell their hardware and software products.

Marty Taucher: Well we've got to wrap up.

Bill Gates: Okay.

Will Zachmann: Well, thanks very much for taking the time.

Bill Gates: It was real interesting.

Marty Taucher: A lot different from what I expected.

[tape ends here]