This talk was given by videoconference at the American Physical Society March Meeting in Denver, Colorado, on Tuesday, March 6, 2007.

Well, I think I should start off by telling you a little of my story. I started off a physicist. I got seriously interested in physics when I was about 10 years old, around 1970. And within a year or two I’d read the standard college textbooks. And by the time I was 14 I started writing papers about particle physics.

At first, I don’t think they were terribly good. But pretty soon they got better, and by the time I was 16 or 17 I was generating decent particle physics papers at a fairly rapid rate. Which even still get referred to today.

I went to college—at Oxford—briefly. Then went to graduate school—at Caltech.

I got my PhD shortly after my twentieth birthday. And then started being an official faculty physicist.

And perhaps I could have spent the rest of my life just like that. But I didn’t.

I’d always been someone who liked making sure they were using the best tools. I myself wasn’t particularly good at doing things like algebra. But if you looked at my physics papers, they were full of amazing calculations.

And, of course, what I was doing was using a computer to do those. Something that almost nobody in physics at the time knew how to do.

Well, I wasn’t very satisfied with the actual software that I had to use to do my calculations. So in 1979—a few weeks after I got my PhD, actually—I decided the best thing would be for me to build my own software system.

Well, I knew a decent amount of computers and such. But I resolved to learn whatever the state of the art was. And pretty soon I started building a large software system for doing calculations. Within a year or two the first version was finished. And it was clear that—as planned—it was a system that would be useful to more than just me.

So I had to figure out what to do with it. There was a lot less to go on back then. But I realized that to have a serious ongoing development effort, one needed some kind of ecosystem. And it seemed like the best one was a company.

Well, there’s an ugly story about the confused interaction of that idea with the university administration. Wild stuff, complete with real-life corruption and everything. If I have time in the next few years, now that 25 years have passed, perhaps I’ll start posting some of the documents.

But anyway, after some bumps, I ended up starting a company. At the time, I was just a 21-year-old physicist. I didn’t think I knew anything about starting companies.

I remember checking some books on management out of the library. Complete with diagrams about how a manager should sit at a higher desk and things. I thought they were really stupid, and completely ignored them.

But I didn’t think I particularly knew what to do. So I ended up recruiting a CEO. Which, in retrospect, was a pretty stupid idea.

But at the time, he started doing all the official company things. Raising venture capital and so on.

Well, I increasingly thought that the things the company was doing didn’t make any sense. At the time, I didn’t know if I was right. Or if the official management I’d brought in was.

I was just using common sense. And they were supposed to have real business knowledge. Well, in the end, I did turn out to be right. But amazingly, the company didn’t completely implode, and after limping along for years, actually eventually went public.

Well, meanwhile, I’d been back mostly being a scientist. I used to do strategy consulting as a hobby. And indeed I still sometimes do informal CEO counseling. It’s kind of fun… and back in the early 1980s, it was very educational for me. Seeing all the different kinds of bad—and sometimes good—things people tried to do with their companies.

But meanwhile, I was back doing science. I had started studying cellular automata. Started using computers—and software I’d built—to do experiments. And I’d discovered some very, very interesting things.

That seemed to say a lot about how complexity was produced in physics, and in nature in general. It was fun. I was looking at stuff nobody else had ever done anything like. So it was really easy to find all sorts of really fresh, new stuff. Even if people thought the papers I was writing were a bit alien.

But anyway, I gradually realized that the things I was discovering could make a pretty interesting direction in science. And I thought it’d be fun to try to really build that science, and get lots of people involved. So I started promoting what I called “complex systems theory.”

It was a bit frustrating. Almost everybody just wanted to go on doing what they’d been doing. And hoped that nothing would rock the boat.

Well, I started a journal for the field. And then I went to lots of different universities, and got “bids” for where to start an institute for the field. I ended up doing it at Illinois.

I have to say that—unlike many institutions—that university largely lived up to its promises. But I realized that life running an organization at a university wasn’t really for me. And in fact, I realized that my whole scheme of trying to move the science forward by recruiting lots of other people wasn’t what I should do.

Well, now, many years later, the banner of “complexity” that I started pushing in the early 1980s has taken on a huge institutional structure. It’s really cute to see little fragments of the marketing material for the field that I wrote 20+ years ago show up in the latest “isn’t this exciting” manifestos. But really it’s all been rather disappointing. I thought there were some really good scientific directions that could have been taken. But instead the field became dominated by lots of mushy stuff. And—well, perhaps it’s for the best—my contributions to getting the whole thing started have been almost 100% forgotten.

Still, back in 1986, what I decided was that instead of trying to get lots of other people to push the science forward, I’d just try to do it myself. I’d build myself the tools and the infrastructure to be as efficient as possible. And then I’d just do the science. Which would be good for me, because doing science is something I really enjoy.

So, actually, that was when I started building Mathematica. Twenty and a half years ago. I wanted to make a computational system that—for the rest of my life—I’d be able to use to do the science and other things I was interested in.

There were lots of intellectually difficult issues in really making a unified system that was as broad as that. But the good news was that I knew enough about fundamental issues in computation that I figured out some sort of foundational ideas—that I think I can fairly say have worked out well.

Well, in addition to the intellectual side of things, I knew I had to build some real, business infrastructure. And this time around I decided that I would just do it myself. Without any investors or professional managers. With me as CEO.

Well, twenty years later, I think I can say that that’s worked out really well. I’ve had a good time. We’ve built some wonderful stuff. We’ve had a company that’s been stable and consistently profitable for nearly twenty years.

Actually, right now I’m tremendously excited, because some things we’ve been building for about the last decade are finally about to come to fruition. And they’re really exciting.

But anyway, back to my story. So, late in 1986 I started building Mathematica, and started Wolfram Research. Well, after Mathematica Version 2 came out in 1991, I decided I should do what I’d planned, and start actually using Mathematica to do some science.

I wanted to continue what I’d started doing with cellular automata. I wanted to go on—as I call it, “exploring the computational universe.”

I wanted to do, sort of, the basic physics, the basic natural science, of the universe of all possible universes. Just enumerating simple programs, and seeing what they do.

It was an amazing experience. Having Mathematica felt like I imagine having the first telescope must have done. One could just point it somewhere in the computational universe. And immediately start seeing all this incredible stuff.

Well, the only problem was that I started discovering an outrageous amount of stuff. About the basic science of the computational universe. About physics. Biology. Computer science. Mathematics. Lots of things that had been pretty mysterious before. I could suddenly see how they worked.

Well, that was very exciting. But it was definitely not trivial to go on being a company CEO while spending 10+ hours a day doing basic science. And actually, the result of that was that I kind of limited the growth of our company during that time. I wanted to make sure the company stayed stable. Which in some ways was good, because during those years we built some extremely solid systems.

But anyway, I kept on thinking I’d get to the end of what I could easily discover in my science. But it kept on not happening. And finally, after ten and a half years… I had sort of filled out my original table of contents. And I wrapped up all the science I’d done, in that big black-and-yellow book that I hope you’ve all seen, called A New Kind of Science.

The science in A New Kind of Science—or NKS, as people seem to call it—is pretty interesting. The core is—as the title of the book says—a “new kind of science.” That one day will be like a physics, or a chemistry, or a mathematics. But concerned with studying what’s out there in the computational universe.

And from this basic science come all sorts of applications. Modeling all kinds of systems in nature—in physics, perhaps even including our whole universe. But perhaps ultimately even more importantly, come the ideas of “mining” the computational universe: finding simple programs out there that can be used for technology. It’s like the early days of chemistry, or materials science. Finding simple programs out there that turn out to be useful for particular purposes. Nowadays we use this idea all the time for finding new algorithms in Mathematica. There’s a little application that some of you will have seen in our WolframTones site. But there’s a vast amount more to come. And I’ve no doubt that one day it’ll be a trillion-dollar industry.

It’s fascinating to watch it all get going. I’m always amused at our NKS Summer Schools, realizing that the alumni will one day be the leaders of this big, institutionalized field. I was charmed seeing the actual response—five years ago now—when my book was published. I’m a big student of the history of science. And it was really quite exciting to see a paradigm shift, on the ground.

Not that I was in much doubt about the things I’d discovered. But it was kind of nice—and, if history is a guide, a very good sign for the future—to see all the characteristic “paradigm shifty” behavior. “It’s wrong.” “It’s been done before.” “It’s wrong and it’s been done before.” “It just isn’t science.”

I was a little sad to see some otherwise-reputable physicists making asses of themselves. But I suppose at least it’ll be fodder to encourage a future generation.

After the fact, it always seems like successful innovations were obvious, and came in completely smoothly. But on the ground, there’s lots of turbulence, often for years. Still, afterwards, it becomes the big surprise when people study the history: “Can you believe it? When NKS first came out, Professor So-and-So said blah, blah, blah.”

One has to choose. One can be involved in things that are well established. Where one operates within the existing institutions. With all the political skills that needs. But in a sense needing little intellectual entrepreneurism.

Or one can be involved in things that are new. And then one needs to be an entrepreneur. Doing things just because one believes in them oneself. Not because there’s a whole institutional structure that tells one one should be doing it.

Some people find that really scary. And some people definitely fail at it. But, perhaps because I’m a much better entrepreneur than I am a politician, I’ve always found it vastly preferable.

Well, so, back to companies. They’re a bit the same kind of thing.

One can choose to be a professor. Where at age 30 one in a sense has one’s whole life laid out for one. And where, so long as one can handle the politics of the university, one can just keep doing what one’s doing for one’s whole life.

And for some people, that’s a really outstanding deal. But, I have to say, if one wants to do new things, and really make things happen, it’s not such a good deal. In the 20 years or so since I completely checked out of academia, I’m afraid many aspects of it have become yuckier and yuckier. I think professor salaries at top schools have gone up.

But my impression is that the academic scurrilousness rate has gone up. And the pressure to toe the line—and not to innovate, or even think different thoughts—has gone up. The management of universities tends to be kind of weird too. With a certain lack of clarity about who or what the university is being run for.

You know, it’s always interesting. When people from academia join our company, there’s one thing that often surprises them. That the company is upset if, for example, their computer is broken. That the company is actually motivated to make them productive. Rather than it just being a question of whether some administrator feels like helping them.

But the thing I like most about companies—at least entrepreneurial ones—is that they’re such great vehicles for turning ideas into reality. I like ideas. I even have quite a few of them. And the great thing about a company is that one can take ideas and get a group of people to turn those ideas into something real—and something that actually has significance in the world.

People sometimes imagine that companies—”the private sector”—is all about money. Sometimes it’s useful that one can fall back on money as being the driving force in making a decision one way rather than another. But at least for someone like me, most of the time it’s about doing the right thing, the useful thing, the interesting thing.

If I look at companies, some of the ones that have made the most money are ones that—at least as far as I’m concerned—do the most boring things. Very repetitive. Well-oiled machines. In a particular track.

Of course, on the other side, there are companies with wonderful-sounding ideas, but which never manage to make them practical enough to make any money at all.

I’ve taken some trouble to try to architect a niche that’s somewhere in between. To build a company that does things that I find intellectually really interesting. But that’s practical enough that it’s consistently profitable. So I can concentrate on doing interesting things, rather than being concerned about money and staying afloat all the time.

In many ways our company isn’t very typical. We’re really at the leading edge of a lot of things. Yet we’ve been around, and been stable, for nearly twenty years now.

And my job probably isn’t very typical, either. Over the years, I’ve built up a really excellent team. A lot of very, very talented people, in lots of areas. Who are also practical, and very civilized. Who like doing what they’re doing. And whom I like working with.

I’ve tried to keep our company fairly small. In many ways it’s more fun like that: one can actually know the name of the person who’ll know about X, rather than just thinking that some faceless department will handle X. And we’re still—at last count—a little under 400 people. Though actually right at the moment we’re growing at the rate of perhaps two people per week.

And one feature of our company is that people who manage things all actually do things, too. They write code. They write documents. They figure stuff out.

My own main activity right now, and for the last couple of years, has been product design. Figuring out just what Mathematica should do. Not so much how it should do it inside. But what it should do.

What functions should we have? How should they work?

I’m a remote CEO—I spend most of my time here in Boston, and the largest part of the company is in Illinois but with people scattered randomly around the world. So my typical day consists of a long sequence of webconference meetings. Where I work with groups of people, trying to figure out what Mathematica should do. Just before this I was doing linear algebra. Right after, if I read my calendar right, I’ll be doing computational chemistry. Then something to do with how people search our vast documentation system. Then… then something I can’t talk about yet.

Mostly what I try to do is to get concrete results out. We’ll—in real time—write a page of documentation about a function, defining what it should do. Or write the words that explain to people how to use some web service. Then it’ll get implemented.

In building a company, one of the most important things is people. I happen to find people—and the paths they follow—very interesting. Some people might call me a “people collector.” And perhaps that’s why I’ve built up such an interesting team at our company.

But it’s always a very interesting problem: given a person, what should they do? Where do they fit? Where will they do something they’ll really be good at, and really enjoy. Most people at any given time in their lives have various parameters. Like how large a project they’re optimized for. And that ranges from a few minutes, or a few hours… to days, months, or years. And in a company like ours, there are important jobs at all those timescales. From technical support—a common training ground for people—where you can answer a question in minutes, and then be onto something else. To software quality assurance, where you might spend a day writing tests for some particular function. To the various development groups, where sometimes people go off for two years to just develop something, and come back with it finished.

Sometimes there are projects where people just sit and code all day. Sometimes there are projects where people have to make a bunch of judgment calls. Sometimes where they have to interact with people. Sometimes where they have to have funky ideas. And so on.

One of the most interesting things about management, as far as I’m concerned, is matching up people with projects. Sometimes people self-identify correctly what they’re good at. Sometimes they need help. Sometimes they need a lot of pushing.

Often people have some kind of self-image—from their education, their family background, whatever—about what kinds of things they should be doing. And sometimes that’s fine. But sometimes there’s a perfect fit for them, that their background would never make them consider.

Sometimes you can tell from peoples’ hobbies. There’s something they really like doing, but they’ve never thought they can actually do it for a living. Sometimes you have to try people out on different things. You—and they—just can’t tell until they try.

It’s rather nice right now at our company that we happen to have just opened up a bunch of new types of projects that fit different types of people. And that’s part of the reason we’re hiring a lot right now—because we have some new niches.

We hire quite a lot of physics PhDs, by the way. What I like about them tends to be this: That they’re analytical, and fairly precise. But yet they’re prepared to jump into new things, and just start doing them. They expect to be able to learn new things from books. Rather than having to be spoon-fed them. And of course, the better ones essentially all know Mathematica.

Well, I think this is supposed to be a session about entrepreneurism. I have to say that I think you don’t have to be a company founder to be an entrepreneur, at least in the right kind of company.

If you can pick a company at the right stage, and you’ve got the right kind of can-do attitude and aptitude, you can go in and just start getting interesting projects done. In a very entrepreneurial kind of way.

Actually starting a whole company is not for everyone. It’s a lot of work. It’s a lot of responsibility. You’re really the bottom line… the person who actually has to figure out what to do. There’s no committee to fall back on. No advisor to ask.

Myself, I’ve always found it very liberating. I’m a really bad committee person. I find all that stuff incredibly frustrating. When I was an academic, I had a few amusing run-ins with that sort of thing—which I think taught me why I couldn’t be an academic long term. But as a CEO, it’s so much better. For me, being the person responsible is just fine. I make decisions. Usually they’re right. Sometimes they’re wrong. But I don’t get terribly stressed out about it.

And I think that’s a very desirable trait if you’re going to start a company.

I should say, by the way, that if you’re going to start a company, you actually have to do it. This idea that somehow you can bring in business professionals who can make it all work is nonsense. The person with the idea, the vision, the original know-how, has to be the one actually running the company. If you can find a terrific partner—well, that’s very hard—but if you can, that’s great. But otherwise, just try to hire employees. And, at least if my own experience is a guide, a good principle to follow is this:

If there’s something you’re told that you don’t understand, try harder to understand it. And if you really can’t understand it, just ignore it in whatever you’re doing.

Don’t get led to doing things because some “expert” told you so.

Well, so how can you tell if you’ll be able to start a successful company? It’s funny, I’ve been involved in high-tech and so on for a quarter of a century. And I think I’m pretty knowledgeable about people. And one can tell within a certain band what’s going to happen. One can say: yes, this person will do OK. They’ll be able to start a company that won’t be completely hopeless. Or: forget it, this person will never be able to start anything that works. They’re too much of a dreamer. They don’t understand enough. They’re not prepared to work hard enough. Whatever. But somewhere on the margins, it’s hard to predict. It’s always fun to see people who’ve come through the company, or whom I’ve otherwise interacted with, do really really well. There’ve been a lot of those now. And they’re hard to predict. Sometimes the people themselves have a certain dogged tenacity that means they kind of have to go somewhere. Sometimes there’s an element of luck involved. Of being in the right place at the right time. Though in my experience, luck is overrated. Usually, the people chose to be there. They chose to work on NKS rather than string theory. Or whatever. And that means they have a certain key element of judgment.

Which is why they’re the ones who go further. Well, I think I should wrap up here. There’s lots to say about all this. And I’d be happy to try to answer whatever questions there may be.