Science and a New Kind of Prediction: An Interview with Stephen Wolfram

Subject(s):
© STEPHEN FAUST PHOTOGRAPHY
Stephen Wolfram

Interview by Patrick Tucker

Better living through data? When a pioneer of data collection and organization turned his analytical tools on himself, he revealed the complexity of automating human judgment and the difficulty of predicting just what is predictable.

Stephen Wolfram, creator of the Wolfram|Alpha search engine and author of the books Mathematica and A New Kind of Science, is known all over the world for his contributions to our understanding of computation. In 2012, he received a lot of attention for something else: At the SXSW show, he revealed that he had a more than 20-year personal computational log of, basically, the life of Stephen Wolfram. This included everything from every e-mail he had sent, to when he had gone to bed, to how long his phone conversations lasted, and much more. He then released this data on his personal blog.

So, what can one of the world’s foremost mathematical minds learn about life by examining his own computational data? THE FUTURIST called to ask.

THE FUTURIST: In your blog, you write that “in time I’m looking forward to being able to ask Wolfram|Alpha all sorts of things about my life and times—and have it immediately generate reports about them. Not only being able to act as an adjunct to my personal memory, but also to be able to do automatic computational history—explaining how and why things happened—and then making projections and predictions.” What sort of things have you been able to predict based on this data set you released?

Stephen Wolfram: One thing I found out is that I’m much more habitual than I ever imagined. It’s amazing to see oneself turned into full distribution. It got me thinking about lots of different ways that I could improve my life and times with data. What I realized is that one of the more important things is to have quick feedback about what’s going on, so you don’t have to wait for a year to go back and look at what happened. You can just see it quickly.

I was actually embarrassed that I hadn’t had a real-time display of the history of my unread e-mail, as a function of time. We built that after this blog post, and I have found it’s quite amazing. By having this feedback, I’m able to work more efficiently. It’s also telling me things like, Gosh, if I ignore my e-mail for four days, or five days, or ten days, or something, it will get totally out of hand, and it would take me weeks and weeks to recover from that. Those sound like very mundane [insights], but in terms of how one actually spends one’s time, they can be quite significant effects.

THE FUTURIST: Nobel Prize–winning economist Daniel Kahneman tells us that people, even particularly smart people in extremely high-performing situations, will consistently underestimate how much time it takes them to complete a certain task. So now that you’ve been able to rid yourself of subjective bias in terms of how long it takes to complete tasks, it sounds like you’ve actually been able to see efficiency improvements, just based on taking a look at what you can get done, how long it actually takes, versus how long you think and that sort of thing.

Wolfram: I have pretty good metrics now. If I’m going to write out something for some talk I’m going to give, or something like this, I know how long it takes me now to give the talk, or to write it out. I know how long to set aside. I have learned that there’s no point in starting early, because I won’t finish it until just in time anyway. I have to know how long it’s actually going to take to finish so that I can get it done in an efficient way. If I start it too early, it takes me longer. The task expands into the space available, so to speak.

THE FUTURIST: According to your data, you actually go to sleep late and wake up late. And that’s interesting because now I know that about you, and so do millions of other people. When you released this material, you gave the public a really unique window into aspects of your life. The vast majority of the comments on the blog are really supportive. But are there any reactions that really stood out when you released all of this information?

Wolfram: I was waiting for someone to say, Ah, well, you know, based on what I can see in this beaker, that must mean that you have the following terrible condition, or something like that. But nothing like that happened.

I felt much nerdier after I watched the reaction to the blog, because I really thought there were lots of other people who were thinking about it more than I was, collecting lots of stuff. And it doesn’t seem that there were.

THE FUTURIST: Only very recently, a growing number of people are routinely collecting data about themselves all the time. The drivers for this trend seem to be better and lighter computers that make personal record keeping much easier. And things like Excel spreadsheets, and, of course, shareability through the Internet are also pushing this trend toward personal record keeping and sharing of personal data. How do you see these trends evolving in the context of present-day battles over privacy and over access to information technology? And what has to happen in order for the self-quantification trend to become a truly sustainable movement?

Wolfram: Right now, for most people, it’s dealing with this data. There’s all kinds of plumbing that has to be done. Like, how do you actually get your cell phone call record out? It’s going to stay a complex, multi-part, multi-vendor environment, where people have different phones, e-mail systems, computers, and little devices like pedometers. Those devices will come from lots of places. So the main thing that holds things up is just the practical difficulty of all of this plumbing, of getting data from here to there, and so on.

THE FUTURIST: So there’s a lot of data that we create that we don’t really have access to, that’s stored, perhaps, in the servers of supermarkets, of different people that we give money to, of different vendors, and things like that. If more of that could become transparent, then it would become more useful to us. Would that deepen our relationship with those different entities that have that data? Would it basically give us a lot more information about ourselves?

Wolfram: You know, that’s actually something I haven’t done—taking all the online versions of all these financial records and combining them. I don’t even need to collect that data. It’s already been collected.

THE FUTURIST: Your seminal book, A New Kind of Science, is ten years old. You recently wrote a blog post on the anniversary. Can you talk a little bit about the future of science?

Wolfram: The main idea of A New Kind of Science was to introduce a new way to model things in the world. Three hundred years ago, there was this big transformation in science when it was realized that one could use math, and the formal structure of math, to talk about the natural world. Using math, one could actually compute what should happen in the world—how planets should move, how comets should move, and all those kinds of things.

That has been the dominant paradigm for the last 300 years for the exact sciences. Essentially it says, Let’s find a math equation that represents what we’re talking about, and let’s use that math equation to predict what a system will do. That paradigm has also been the basis for most of our engineering: Let’s figure out how this bridge should work using calculus equations, and so on. Or, Let’s work out this electric circuit using some other kind of differential equation, or algebraic equation or whatever.

That approach has been pretty successful for lots of things. It’s led to a certain choice of subject matter for science, because the science has tended to choose subject matter where it can be successful.

The same is true with engineering. We’ve pursued the particular directions of engineering because we know how to make them work. My goal was to look at the things that science has not traditionally had so much to say about—typically, systems that are more complex in their behavior, and so on—and to ask what we can do with these.

It’s a great approach, but it’s limited. The question is, what’s the space with all possible models that you can imagine using?

A good way to describe that space is to think about computer programs. Any program is [a set of] defined rules for how a system works. For example, when we look at nature, we would ask what kinds of programs nature is using to do what it does, to grow the biological organisms it grows, how fluids flow the way they do—all those kinds of things.

I’ve discovered that very simple programs can serve as remarkably accurate models for lots of things that happen in nature. In natural science, that gives us a vastly better pool of possible models to use than we had from just math. We then see that these may be good models for how nature works. They tell us something about how nature is so easily able to make all this complicated stuff that would be very hard for us to make if we just imagined that nature worked according to math.

Now we realize that there’s a whole different kind of engineering that we can do, and we can look at all of these possible simple programs and use those to create our engineering systems.

This is different from the traditional approach, where I would say, I know these things that work. I know about levers. I know about pulleys. I know about this. I know about that. Let me incrementally build the system where I, as an engineer, know every step of how the thing is going to work as I construct it.

THE FUTURIST: One of the key themes of A New Kind of Science, and also a key theme in your TED talk, is this notion of irreducibility. There are certain things that can’t really be predicted, no matter what. You can’t model them in advance. They have to be experienced. And I wonder, given the future of digitized knowledge, the exponential growth in structured and unstructured data that we can look forward to over the coming decades, is it possible that the space of irreducible knowledge, of unpredictable knowledge—while it will still always exist—is shrinking? Would this mean that the space of predictable knowledge is in fact growing?

Wolfram: Interesting question. Once we know enough, will we just be able to predict everything? In Wolfram|Alpha, for example, we know how to compute lots of things that you might have imagined weren’t predictable. You have a tree in your backyard. It’s such and such a size right now. How big will it be in 10 years? It’s now more or less predictable.

As we accumulate more data, there will certainly be patterns that can be seen, and things that one can readily see that are predictable. You can expect to have a dashboard—with certain constraints—showing how things are likely to evolve for you. You then get to make decisions: Should I do this? Should I do that?

But some part of the world is never going to be predictable. It just has this kind of computational irreducibility. We just have to watch it unfold, so to speak. There’s no way we can outrun it. I suspect that, in lots of practical situations, things will become a lot more predictable. That’s a big part of what we’re trying to address with Wolfram|Alpha. Take the corpus of knowledge that our civilization has accumulated and set it up so that you can automatically make use of it.

There are three reasons why one can’t predict the things that can’t be predicted. The first reason is not enough underlying data. The second is computational irreducibility—it’s just hard to predict. The third is simply not knowing enough to be able to predict something. You, as an individual, don’t happen to know enough about that particular area to be able to do it. I’m trying to solve that problem.

We’re seeing a transition happening right now, and more and more things can be figured out in an automatic way. We’re seeing computation that is finally impinging on our lives in a very direct way. There are lots of things that used to be up to us to estimate, but now they’re just being computed for us: a camera that auto focuses, for example, or that picks out faces and figures what to do, or automatically clicks the shutter when it sees a smile—those kinds of things. Those are all very human judgment activities, and now they’re automated.

I think this is the trend of technology. It’s the one thing, I suppose, in human history that has actually had a progression: There’s more technology; there are more layers of automation about what we do.

THE FUTURIST: You talk about the twenty-first century as being the era where personal analytics really took off. But for you, it was actually the twentieth century. You’ve got the data set that would allow you to answer this question with the largest degree of evidence behind it: Where do you see yourself in 10 years? And what do you see yourself having accomplished 10 years from now?

Wolfram: Well, that’s an interesting question. My gosh. That’s the kind of question one’s supposed to ask at a job interview. I never ask those, because I always figure that they’re silly questions. I’m hoping I’ll do a few new things. We’ll see. For the last decade, Mathematica and Wolfram|Alpha were my main activities. I felt pretty good about those. I hope for a few more in the decade to come.

About the Interviewee

Stephen Wolfram is the creator of the Mathematica computational platform, the author of A New Kind of Science, the creator of Wolfram|Alpha, and the founder and CEO of Wolfram Research, www.wolfram.com.

This interview was conducted by Patrick Tucker, deputy editor of THE FUTURIST, and is a preview of his book, The Naked Future: What Happens in a World That Anticipates Your Every Move? (forthcoming from Current Books, a Penguin imprint, March 2014).