In his keynote presentation to SXSW 2014 last week, computing pioneer Stephen Wolfram demonstrated the capabilities of his new Wolfram Language. But he ended his talk with a radical vision of the future of humanity.
In keeping with the title of his talk, which was “Injecting Computation Everywhere,” Wolfram described a future where computers are so tiny and so cheap that they are embedded in everything, and everything becomes programmable.
“Languages will become the stuff that everything is made of,” Wolfram said. And, if you can program the stuff around you, “you can build any sort of universe for yourself.”
In the end, Wolfram said, “everything is reduced to a box of a trillion souls, and they’re all running whatever computations they want to.”
Read part 1 of this interview: How Stephen Wolfram plans to reinvent data science & make wearables useful
We spent an hour with Wolfram shortly after his keynote to explore these ideas further. In part one of our interview, Wolfram described the potential for the Wolfram Language to revolutionize computer science and data science and talked about how it could make wearable devices far more useful.
Here, in part two, we quizzed Wolfram on his vision for an entirely computable future. What he said kind of blew our minds, but he’s quite serious about injecting computation literally everywhere.
VentureBeat: At the end of your talk today, you were talking about how computation becomes embedded in everything. This is the thing that is commonly referred to as the “Internet of things.” Your vision makes that term almost too modest because you’re talking about computation being in everything. Tell me how we get there, how that starts to happen.
Wolfram: The first thing that happens is sensors show up everywhere.
It’s not worth having anything other than a general purpose computer in any of these little, tiny devices. You might have seen me do an announcement with Intel a little while ago about their Edison thing. There’s more coming along various lines in those kinds of directions.
In the past it was, “We just want to put a little bit of electronics into our thing, not a real computer.” In the future, it’ll be, “Forget that. Just put a damn computer in it.” Then, it’s just software [that’s needed] to figure out what it does.
The next level is what’s the computer really made of, how big does the computer have to be. This is a place where there’s, in a sense, a bigger jump.
The fact that you end up getting computers the size of SD cards, we know that’s coming. They’ll get to be the size of microSD cards, and then they’ll get even smaller than that. They’ll be a whole computer, and we can run our whole technology stack inside that very tiny computer.
Then there’s the “Will we run out of Moore’s law or some analog of Moore’s law for size of computers?” Maybe.
Then the thing to understand is there’s a lot you can make computers out of. The approach that we’ve been taking right now is we build these really complicated CPUs, and they have lots of complicated circuitry in them. One of the discoveries from the science I did is that you can make computers out of almost anything, even these little tiny, simple cellular automaton things with little rules. Those can be universal computers.
Then, when you think, “How do I make the computer out of molecules?” for example, one approach is to say, “Let me take the computer that I know works, and let me use photolithography to make the picture of the computer really, really small.”
The different approach is to say, “Let me start off with the molecules, and let me see how to, essentially, search the computational universe for a way to arrange these molecules so that they will act as a computer.”
It turns out that that’s possible. That means it’ll be possible to make vastly smaller nanocomputers, so to speak, that will allow us to take the stack of technology that we have, with all our software, languages, and all those kinds of things, and compile it down to these things for these weird nanocomputers.
VentureBeat: That is really weird. [laughs]
Wolfram: You can compile to almost anything. Compilation is a very powerful idea. Compiling to nanocomputer code will work just fine.
How to build a very small computer
VentureBeat: Explain this for me, and let me know if I understand this right: What you’re saying is the computer engineers of the future will be finding small processes, like a chemical reaction or a molecule that does a certain thing when stimulated in a certain way.
Then they’ll figure out ways of assembling these tiny, very simple switches and find ways to compile them into computing machines that can run anything, including the Wolfram language.
Wolfram: For example, yes.
To explain that process: Technology is always about that. We go out into the physical world. We go find things that we can somehow harness for our particular purposes. We find wood that we can use to build things out of. We find liquid crystals that we can make displays out of, stuff like that.
The big thing that we’ve learned is you can just go out into this computational universe of possible algorithms. You can just go mine things directly from that. For example, in Wolfram Alpha, in Mathematica, and in the Wolfram Language, a whole bunch of algorithms that we have in there, we didn’t invent them as humans. They were things where we just sent some search system off [to find]. It searched a trillion possibilities, and it found one that was really good. We found the optimal one out of a trillion possibilities, and that’s the one we use.
When you, as a human, go look at it, you say, “What the hell is this? I don’t understand how this works.” Sometimes, you can go in and prove that it works, but it’s something where you say, “I never would have guessed that this would work.”
What’s interesting about that, to me, is that’s how nature is set up. There are lots of things that do what they do in nature, but it’s hard for us to understand how they work. That’s the same thing that we find in the computational universe.
Our own technology is, in a sense, still very primitive. One way we can see that is we can recognize when something was built, as an artifact. In other words, it’s got a lot of history in it. It’s got gears, levers, and little pieces of CPU chips that have a particular form that we can recognize as being very regular structures, and so on.
When you go out into the computational universe and search for possible technology, in effect, a lot of what you find is not stuff that is as readily recognizable to us. Just like when we look in nature, we see lots of things which we don’t readily recognize what this is, how does it work. We’ve had to spend centuries of science to find out how it works.
VentureBeat: In the way that a virus is a computing machine or a replicating machine, you’re talking about other processes in nature potentially being computational machines.
Wolfram: Take the weather for example. People always say, “The weather has a mind of its own.” They’re right. It’s effectively computing things just like any of these other systems are computing things. I think you could say, “Well, okay, let’s bottle that up. Let’s make some fluidic computer that’s based on some fluid dynamics process.” Maybe that’ll happen; maybe it won’t.
That’s not a particularly fertile direction, but it’s not obviously silly. What’s nontrivial about it is that it is possible to do sophisticated computations just with some piece of fluid or something.
VentureBeat: Well, pretty much anything that can be programmed will be turned into a computer somehow by nerds somewhere. In Minecraft, you have people making computers, Turing-complete machines, out of blocks and torches and pistons, right?
Wolfram: This is one of the big things from my science project. We did a lot of searching the computational universe, like 256 of the simplest possible cellular automata. We already know that four of them are universal computers, so four of the first 256, we know are universal computers.
Then for Turing machines, we actually know the simplest universal Turing machine. About five years ago now, we put up a $25,000 prize for somebody to prove or disprove that this particular Turing machine is universal. I thought this might be one of these kind of “Fermat’s Last Theorem”-type problems that was open for hundreds of years. But actually, in about four months, this young guy in England came up and said, “Okay, I have a proof.” It’s this long, complicated, semi-automated, 50-page thing; and indeed it is a proof that this Turing machine is universal.
We know that this tiny Turing machine — it has two states, and three colors, so it’s a really simple thing — is the simplest universal Turing machine. It’s amazing how simple a system can be while still being able to compute; and yes, Minecraft is a great modern example of that.
VentureBeat: The upshot is, at some point in the future, we are surrounded by computing. Computing is the fabric of our lives.
Wolfram: Everything we make is.
VentureBeat: Everything is a computer?
Wolfram: No, but everything will be made of computers. That’s the weird thing that I realized recently.
In other words, you take some fabric or something, right now it’s just made of molecules that have long stringy molecules and things. But in the future, you might as well just make this out of molecules with computers. At that point, everything becomes programmable.
In a sense, the languages are in charge at that point. Then what matters is the software layer that you’ve built above that hardware thing.
Now, there’s another piece, which I’m eliding here, which is what happens to the future of robotics. Robotics has been this weird area, because for the 40 years or something that I’ve followed it, it kind of seems almost the same. Because we’ve got a robot; it’s got these motors; it’s got these things that move in very specific [ways]. Now, the details of gotten a lot better, but the ideas are still the same.
My own guess is that there will come a moment when we get a version of modular robotics, where it’s possible to turn the problem of mechanical robotics into software by essentially saying, “We’ll just make this robot out of lots of identical pieces, and then the actual operation of the robot will consist of lots of these identical pieces moving around in some determined way to effect mechanical actions.”
I actually know how to do it, but it’s stuck in a technology loop because of details about how you have things move, but not take too much power. It’s one of these things where it’s obvious that it’s valuable, it’s just we’re not quite there yet, or at least I can’t figure out how to get us quite there yet.
A box of a trillion souls
VentureBeat: At the end of your talk you painted this picture where we end up with a “vast box of a trillion souls.” It sounds very Matrix-like, and it made me deeply uncomfortable.
Wolfram: It makes me uncomfortable too, actually.
VentureBeat: Are the trillions souls people, or are those people and the programs that they’ve created?
Wolfram: They’re the end results of people. One of the things that’s interesting is once the whole thing is virtualized, you get to have any universe you want, and people get to explore the computational universe, the possible universe, and that’s a weird thing in its own right.
But the real question is, we’re forced very much into a, “What’s the point?” situation, because we’ve got this box, and it’s computing things, and we’re looking at it. It’s got all the stuff going on inside. And then we go next door, and we’re looking at this bucket of water that has all kinds of flow going on inside it, and it’s not obvious that there’s a fundamental difference between one thing and the other. One is the end result of our civilization, the other is just some simple physics thing.
The question is, “What was achieved by this whole trend in our civilization?”
The answer is, I think, not as depressing as it might be, because in a sense, sophisticated computation is generically possible for almost everything in the universe.
If we say to ourselves, “That’s what’s special about us: We do computation,” that’s not the right answer. The same reason that somebody might say, “What’s special about us or special about biology, is that we do self-reproduction,” or something. But we clearly know that that’s not actually special about us; we can easily have systems that do that.
What we realize in the end is that it’s an inevitable sort of thing. What’s special about us is precisely our whole detailed history, and there’s nothing abstractly special about us. It’s just that our actual history is special, so to speak.
In other words, [is] there’s something where we can say, “Humans achieved this. Humans made this thing that is just totally qualitatively different from anything else that could exist in the universe”?
I don’t think so.
One of the things, if you look at technology and history and so on, is that there’s been this interesting interplay between people having a certain sense of purpose, and then they create technology to achieve that purpose. That technology grows to a certain point. Then they realize that there are new purposes to be had.
We’ve seen that cycle go on for 10,000 years, or something, [across] human history.
I think we know the endpoint of technology, and the endpoint is that everything’s possible. The endpoint sounds very great, but then the issue is, what does that mean for the endpoint of human purpose? Because in the past, human purpose has often been defined by what is not available, the fact that human life spans are finite, the fact that it’s hard to get energy, those kinds of things.
When everything is possible, it lays bare more questions about what role human purpose actually is. This is one of my current hobby-obsessions, trying to unravel this.
At times it seems that for me personally, with my personal [bias toward people as more than collective computers], it just seems like, “Hmm, that’s a shame.”
But I actually am more optimistic than that. I can’t say that I’ve figured out what the picture really is, but I think it’s going to be richer than one expects, and not as Matrix-like.
VentureBeat: [laughs] I hope so.
Wolfram: Yeah, so do I.