Digital is one of the most widely misunderstood concepts. In computing there's a notion of a sign bit error, where you calculate something and you get one bit wrong, so the sign is the opposite of what it should be, which means everything you calculate is the opposite of what it's supposed to be. There's a sense in which that's happening right now in maybe three different areas.
Claude Shannon wrote the best master's thesis ever when he was at MIT, inventing digital. He went on to Bell Labs and did two core things. The one that's most interesting for me is he proved the first threshold theorem. What that means is I could send my voice to you today as a wave, or I could send it to you as a symbol. What he showed is if I send it to you as a symbol, for a linear increase in the resource used to represent the symbol, there is an exponential reduction in the error of you getting the symbol correctly as long as the noise is below a threshold. If the noise is above the threshold, you're doomed. If it's below a threshold, a linear increase in the symbol gives you an exponential reduction in error. There are very few exponentials in engineering. That's the big one
What he showed is you can communicate reliably even though the communication medium is unreliable; that's what digital means. That's the essence of digital. It wasn't obvious, Claude Shannon got that. When I was at Bell Labs, Bob Lucky was still around there and could tell me stories. Claude Shannon had this idea that we should communicate digitally. There was a real battle between analog communication and digital communication.
The sobering lesson from Bob Lucky is the resolution of the battle was death. The analog managers died and a new generation of digital managers took over. Then we had digital communication, and now the Internet. But the meaning of digital is this threshold property, this exponential scaling.
Around that time, computers were analog. Vannevar Bush at MIT made a differential analyzer, which was a roomful of gears and pulleys, and the answer got worse with time. What John von Neumann did was show you could compute reliably with an unreliable computing device by computing with a symbol. It was exactly applying Shannon to viewing a computation as a communication through the channel of the computer. He showed that you can compute reliably with unreliable devices. The heart of it isn't ones and zeroes, it's the threshold property—the exponential scaling, the exponential reduction in error. Those are the digital revolutions in communication and computation. We'll come back to von Neumann because what he did in computation afterwards is, I'd say, relatively completely misunderstood in computer architecture. We'll come back to that.
We have digital communication; we have digital computation. The thing I've been most recently been involved in is digital fabrication. This is the first of the sign bit errors I want to talk about. At the same time Shannon was digitizing communication and von Neumann was digitizing computing, at MIT in 1952 the first numerically-controlled milling machine was made. You could argue that's digitizing fabrication. In Norbert Wiener's servomechanism laboratory, the Whirlwind computer from Project SAGE, which was an early air defense computer, was one of the first computers you could do anything with in real time rather than batch. There was this idea that you could connect the computer to a machine to turn the cranks on a milling machine and make aircraft parts. At the time, this was a huge leap. It was connecting two alien realms: this new computer thing and a milling machine. What it let you do was make aircraft parts you couldn't make any other way.
Digital fabrication, in that sense, dates back to 1952. Now there's lots and lots of attention about the Maker movement and digital fabrication, but it misses the key point. In the 80s, Chuck Hull invented 3D printing. Now if you run a shop like mine at CBA where we have one of every kind of computer-controlled manufacturing machine, there's maybe twenty processes that can control a computer to make something: cut with lasers, supersonic jets of water, EDM with wires, machining, plasmas, fusing, bonding. For all the attention to 3D printing, we might use 3D printers 20 percent of the time. The other 80 percent of the time other machines are faster, make higher performance parts, all of that. Maybe 20 percent of the time 3D printing is best. 3D printing isn't a revolution; it's decades old. It's a little bit like microwave ovens in the fifties. The kitchen of the future was going to have one appliance and you push a button and make all your food. Of course, we still have a stove. You wouldn't get rid of your stove even though you have the microwave oven.
What all of that misses is, it's analog. The design is digital, but the process is smooshing material. You might cut it or you might squirt it but it's smooshing material. The real invention is 4 billion years old, that's the evolutionary age of the ribosome. To understand the ribosome, think about a child playing with Lego bricks and compare it to a state-of-the-art 3D printer. The child and the ribosome do much the same thing.
When the child assembles Lego bricks, the first attribute is metrology that comes from the parts. When you snap the bricks together, you don't need a ruler to play Lego; the geometry comes from the parts. What it means is that a child can make a Lego structure bigger than themself. The same way in the ribosome—the Lego bricks are amino acids, and the ribosome assembles amino acids to elongate a protein. You can make an elephant one amino acid at a time because the geometry comes from the parts. In a 3D printer today, what you can make is limited by the size of the machine. The geometry is external.
The second difference—now we come to the Shannon part—is the Lego tower is more accurate than the child because the constraint of assembling the bricks lets you detect and correct errors. The tower is more accurate than the motor control of the child.
In a lab when you mix chemicals the yield is maybe a part per 100. In the ribosome, making proteins, the error rate is a part in 104, and when you replicate DNA there's an extra step of error correction and the error rate is 1 in 108. That 1 in 108 is the exponential. That's the exponential scaling for working reliably with unreliable parts. Because the parts have a discrete state, it means in joining them you can detect and correct errors. That threshold property may sound like a technicality but it's exactly the difference between an analog telephone and the Internet, or a differential analyzer and a PC. The second difference is you can detect and correct state to correct errors to get an exponential reduction in error, which gives you an exponential increase in complexity.
The next one is you can join Lego bricks made out of dissimilar materials. In the ribosome there's twenty amino acids that represent the basic properties of life. It's very hard to 3D-print a conductor and an insulator and a semiconductor through the same process.
The last one is when you're done with Lego you don't put it in the trash; you take it apart and reuse it because there's state in the materials. In a forest there's no trash; you die and your parts get disassembled and you're made into new stuff. When you make a 3D print or laser cut, when you're done there's recycling attempts but there's no real notion of reusing the parts.
The metrology coming from the parts, detecting and correcting errors, joining dissimilar materials, disconnecting, reusing the components—those are all the things Shannon and von Neumann taught us. They're digital fabrication. But the crucial distinction is that the code isn't in the computer, it's in the materials themselves. It's digitizing physical reality. There's an exact historical alignment between going from analog to digital in communication and analog to digital in computation, and now analog to digital in fabrication. That's the research revolution: digitizing fabrication, coding construction.
Shannon and von Neumann were very aware of the physical context in doing this. You can trace what Shannon did back to, roughly, Maxwell's demon, the one molecule demon that seems to violate the second law of thermodynamics. Leó Szilárd's analysis of it—reducing it to a single molecule—and Rolf Landauer's explanation of the erasure in the mind of the demon, Shannon was familiar with the history of all of that. Late in life, Turing and von Neumann both started to think about geometry and physics in computing. Turing, von Neumann, and Shannon were all very aware of this physical context in what they did in communication and computation. I've been attributed, and I'm happy to take claim for saying computer science is one of the worst things to happen to computers or to science because, unlike physics, it has arbitrarily segregated the notion that computing happens in an alien world.
A year ago for the White House Office of Science and Technology Policy, I ran a meeting because every federal agency pretty much wanted to talk to me about their 3D printing initiative. I was yelling at them that it's sort of shuffling deckchairs. That's not the new opportunity. I got together all of these agencies, and then I got together the people at the frontiers of this emerging field of the deep sense of digital fabrication as coding construction. What's emerging from that is in a whole bunch of areas we're discovering we can do things that were just not considered remotely possible before.
On the very smallest scale, the most exciting work on digital fabrication is the creation of life from scratch. The cell does everything we're talking about. We've had a great collaboration with the Venter Institute on microfluidic machinery to load designer genomes into cells. One step up from that we're developing tabletop chip fab instead of a billion dollar fab, using discrete assembly of blocks of electronic materials to build things like integrated circuits in a tabletop process.
A step up from that, we had a paper in Science last year showing we can make the world's highest performance ultralight material for things like airplanes by digitizing composites into little linked loops of carbon fiber instead of making giant pieces. Now we're working with the aerospace industry on making printers of jumbo jets. But the printers are really assemblers.
Bigger scale, we're working with Homeland Security on geoprinting. Extreme events like Katrina or Sandy do tens or hundreds of billions of dollars of damage. National technical means to defend against them are bags of wet sand. We're now developing machines that are like robotic ribosomes that link discrete parts to build geological scale features to make landscape. We're working with NASA on doing this in space, leading up to the idea of how you bootstrap a civilization. There's a series of books by David Gingery on how to make a machine shop starting with charcoal and iron ore. You make a furnace and you melt it, and then you make hand tools, then slowly you bootstrap up to make a machine shop. When people think about a notion like colonizing space and bootstrapping a civilization, that's what they're thinking of implicitly.
Now to come back to the ribosome again. There are twenty amino acids. With those twenty amino acids you make the motors in the molecular muscles in my arm, you make the light sensors in my eye, you make my neural synapses. The way that works is the twenty amino acids don't encode light sensors, or motors. They’re very basic properties like hydrophobic or hydrophilic. With those twenty properties you can make you. In the same sense, digitizing fabrication in the deep sense means that with about twenty building blocks—conducting, insulating, semiconducting, magnetic, dielectric—you can assemble them to create modern technology.
Digi-Key—the electronic parts vendor—sells 500,000 different kinds of resistors but at heart there's only three attributes: conducting, resistive, insulating. That's what we're doing. By discretizing those three parts we can make all those 500,000 resistors, and with a few more parts everything else.
That's the revolution. It intellectually exactly aligns with digitizing communication and computation, but now for fabrication. In turn, the alignment is even closer with the history of computing. Where I realized this alignment was so close was, to do this research CBA got a big NSF grant to buy machines. We wrote an ambitious proposal to get one of anything to make anything, and that's luckily what we got funded, which is an interesting story. But the problem we ran into was that it would take too long to teach people to use all of those machines. I started a class called How to Make Almost Anything and that wasn't meant to be provocative. It was just aimed at ten or so research students to use the machines to do that research. Something strange happened, which is hundreds of students showed up to take a class for ten people, and they would say things like, "This is too useful. Can you teach it at MIT?" Every year hundreds of students try to take this class. Then in turn, the next surprise was they weren't there for research, they weren't there for theses, they wanted to make stuff. I taught additive, subtractive, 2D, 3D, form, function, circuits, programming, all of these skills, not to do the research but just using the existing machines today.
Kelly Dobson, who’s run Digital Media at RISD, made a device that saves up screams and plays it back later when it's convenient. And Meejin Yoon, who runs Architecture now at MIT, when she took the class she made a dress instrumented with sensors and spines to defend your personal space. That happened year after year until, finally, I realized that the students were answering what I hadn't asked, which is: what is this good for? I was asking: can you do digital fabrication? It didn't even occur to me to ask why. It was obviously just such an interesting question. What they were answering was the killer app for digital fabrication is personal fabrication, meaning, not making what you can buy at Walmart, it’s making what you can't buy in Walmart, making things for a market of one person.
Let's go back to the history. From the Project SAGE Whirlwind, MIT developed the first transistorized computers, the TX Series. The TX Series then got commercialized as Digital Equipment’s PDPs. The PDPs gave us the Internet. The Internet, word processing, videogames, just about everything you do on a computer today was first done in that era. This was the time of mini-computers. At that time, Route 128 in Boston had Wang, Prime, Data General, DEC, the whole computer industry. Every single one of them failed. The organizational lesson is it didn't matter how good you were at organizational change, they were just doomed.
Ken Olson, the head of DEC, famously said, "Nobody needs a computer in the home." Personal computers are a toy. They don't scale. Play with your toys, we'll make the real machines. Obviously you have PCs in the home, and DEC is twice over bankrupt. DEC was bought by Compaq, Compaq was bought by HP.
To play that forward, they were the minicomputers. The minicomputer industry completely misread PCs. The transitional stage between minicomputers and PCs were hobbyist computers. This was the era of the Altair, and the Altair was life changing for people like me. It was the first computer you could own as an individual. But it was almost useless. The killer app was you could flip switches on a panel, load in a binary program, start it running, and the lights would blink. This was late 70s when that was beginning to happen. Minicomputer, hobbyist computer, then PC. To understand why a 3D printer isn't analogous to the PC: When I was at Bell Labs we used PDP 11/73s and there was a rack, and in the rack there was a unit that's a processor, there's a unit that's storage, there's a unit that's communication, there's a unit that's power, there's a unit that's I/O, there's a unit that's graphics. There's all these systems and you have to plug them all into each other. It was hard to use but it brought the cost from a million dollars to 100,000 and the size from a warehouse down to a room. What that meant is a workgroup could have one. When a workgroup can have one it meant Ken Thompson and Dennis Ritchie at Bell Labs could invent UNIX—which all modern operating systems descend from—because they didn't have to get permission from a whole corporation to do it.
If you follow that history, the minicomputer died, but the reality is every year the machines got faster, better, smaller, better integrated. At the PC stage what happened is graphics, storage, processing, IO, all of the subsystems got put in a box. The ten subsystems of the PDP that were separate units all fit in one box.
To line that up with fabrication, MIT's 1952 NC Mill is similar to the million-dollar machines in my lab today. These are the mainframes of fab. You need a big organization to have them. The fab labs I'll tell you about are exactly analogous to the cost and complexity of minicomputers. The machines that make machines I'll tell you about are exactly analogous to the cost and complexity of the hobbyist computers. The research we're doing, which is leading up to the Star Trek Replicator, is what leads to the personal fabricator, which is the integrated unit that makes everything.
Now let's do that in powers of ten. My research lab—think of it as ten one million dollar machines. So, 10 million dollars to make molecular nano assemblers. Within that there's a workshop—think of that as a million dollars in ten $100,000 machines, which is things like high-speed mills, water-jet cutters, and very powerful lasers. But within that there's a core set of tools that we used in things like the How To Make class, and what happened was, starting about ten years ago, I had a big NSF grant and a law was passed in Congress that government programs need to measure social impact, called GPRA. NSF had no idea how to do it, so they turned to us and said we had to measure social impact, and we had no idea how to do it. But we had good NSF program managers. We that thought rather than tell people what we're doing, we would give them the tools. We set up the first fab lab, and it was about a 100K investment. Think of it as ten ten-thousand dollar machines. It's the basic set of tools today to do digital fabrication. Now, there's a casual sense, which means a computer controls something to make something, and then there's the deep sense, which is coding the materials. Intellectually, that difference is everything but now I'm going to explain why it doesn't matter.
The fab lab is 2 tons, a $100,000 investment. It fills a few thousand square feet, 3D scanning and printing, precision machining, you can make circuit boards, molding and casting tooling, computer controlled cutting with a knife, with a laser, large format machining, composite layup, surface mount rework, sensors, actuators, embedded programming— technology to make technology.
Think of it as a town library scale. You wouldn't ask your town library, "Do you want to skip literature or history?" There's a basic set of things for knowledge. The fab lab is sort of like that for turning data to things, things to data. We set up one of those for NSF and then they accidentally went viral. They've been doubling every year and a half. There's 400 now, there's about 400 coming. They're above the Arctic Circle. They're at the bottom tip of Africa, they're in rural shantytowns, they're in big cities. We didn't plan that. We only set up one but they just started doubling.
Now comes the historical alignment, which is the Internet didn't come after the iPhone. The Internet was invented in the minicomputer era, and every year the computing got faster, better, cheaper and better integrated. But you didn't have to wait twenty years for minicomputers to start using computers. You could use it then. It's an exponential. Socially it looked like a revolution but it was going from one to two to four to eight. On a log-log plot it's just a straight line.
In that same sense we're ten years into the doubling of fab labs. Ten years you can just plot this doubling. Today, you can send a design to a fab lab and you need ten different machines to turn the data into something. Twenty years from now, all of that will be in one machine that fits in your pocket. This is the sense in which it doesn't matter. You can do it today. How it works today isn't how it's going to work in the future but you don't need to wait twenty years for it. Anybody can make almost anything almost anywhere.
Let's play that forward with some of the implications and then relate it to computing and communication. Around this point I began to realize that I was a victim of and was fixing a mistake from the Renaissance, which is, in high school I desperately wanted to go to the vocational school where you could weld and fix cars and do cool stuff like that. I was told, "No, you're smart. You're not allowed to." I had to sit in a room; it seemed punitive. At Bell Labs I had union grievances because I would try to make things in a workshop and they'd say, "No, you're smart, you have to tell somebody else what to do." And it just didn't make sense.
Finally, when I could own all these machines I got that the Renaissance was when the liberal arts emerged—liberal for liberation, humanism, the trivium and the quadrivium—and those were a path to liberation, they were the means of expression. That's the moment when art diverged from artisans. And there were the illiberal arts that were for commercial gain.
We've been living with this notion that making stuff is an illiberal art for commercial gain and it's not part of the means of expression. But, in fact, today, 3D printing, micromachining, and microcontroller programming are as expressive as painting paintings or writing sonnets but they're not means of expression from the Renaissance. We can finally fix that boundary between art and artisans.
Now, technically, the roadmap we're going down is very clear. If you take this alignment between mainframes, minicomputers, hobbyist computers, PCs, the research tools we're using are like the mainframes, the fab labs are the minicomputers. They're being used to do the equivalent of invent the Internet. The next step is we're doing a lot of work on machines that make machines. You don't go to a fab lab to get access to the machine; you go to the fab lab to make the machine. To do that we've had to rip up CAD-CAM, machine control, motion control, all the ways you make stuff, to make machines that make machines. That's the next step. Over the next maybe five years we'll be transitioning from buying machines to using machines to make machines. Self-reproducing machines. But they still have consumables like the motors, and they still cut or squirt. Then the interesting transition comes when we go from cutting or printing to assembling and disassembling, to moving to discretely assembled materials. And that's when you do tabletop chip fab or make airplanes. That's when technical trash goes away because you can disassemble.
An early version of that is Google's Project Ara. Ara was one of my students. That's based on modular reconfigurable cell phones intentionally as the first step down this roadmap. Instead of buying and throwing out a cell phone, it's made out of building blocks you can reconfigure. The research will replace this bit by bit. We'll reconfigure the blocks in the building blocks and then the blocks in the blocks in the building blocks. That's maybe the twenty-year roadmap technically from where we are today.
Now, the biggest surprise for me in this is I thought the research was hard. It's leading to how to make the Star Trek Replicator. The insight now is that's an exercise in embodied computation—computation in materials, programming their construction. Lots of work to come, but we know what to do. The thing that's been most surprising for me is the consequences of this. The equivalent of inventing the Internet. As the fab labs have been spreading, we've been working with heads of state, NGOs, and tribal chiefs, and community activists, and generals—this amazing range, because if anybody can make anything anywhere, it challenges everything.
Start with education. I love the maker movement and I also get irritated by the maker movement for the failure in mentoring. At something like a Maker Faire, there's hall after hall of repeated reinventions of bad 3D printers and there isn't an easy process to take people from easy to hard. In the fab lab network we had this problem that kids would come in all sorts of places all over the world, learn amazing skills and then fall off a cliff. There'd be nowhere for them to go educationally.
We started a project out of desperation because we kept failing to succeed in working with existing schools, called the Fab Academy. Now, to understand how that works, MIT is based on scarcity. You assume books are scarce, so you have to go there for the library; you assume tools are scarce, so you have to go there for the machines; you assume people are scarce, so you have to go there to see them; and geography is scarce. It adds up to we can fit a few thousand people at a time. For those few thousand people it works really well. But the planet is a few billion people. We're off by six orders of magnitude.
In computing terms, MIT is a mainframe. You go there and get processed. I don't like MOOCs—massive open online classes—that trumpet: "Our class has a million people in it." It's just not education as I understand it—a person sitting at a screen—it's like time sharing. There's still a mainframe and you're a terminal plugged into the mainframe. The way the Fab Academy works, in computing terms, it's like the Internet. Students have peers in workgroups, with mentors, surrounded by machines in labs locally. Then we connect them globally by video and content sharing and all of that. It's an educational network. There are these critical masses of groups locally and then we connect them globally.
We started teaching the same digital fabrication class I teach at MIT, "How to Make Almost Anything", but now instead of just teaching at MIT we're teaching it using the whole planet as the campus. Amusingly, I went to my friends at Educause about accrediting the Fab Academy and they said, "We love it. Where are you located?" And I said, "Yes" and they said, "No." Meaning, "We're all over the earth." And they said, "We have no mechanism. We're not allowed to do that. There's no notion of global accreditation." Then they said something really helpful: "Pretend." What they meant was self-accredit with a skills-based portfolio where people document skills and at some future date we'll figure out how to accredit you. And that's been working great. Of all the questions we get, the one question we never get is, "Who accredited us?" because the content speaks for itself.
The interesting thing in turn is we're now teaching this global version of a digital fabrication class using the whole fab lab network. A lot of what we've done isn't tied specifically to digital fabrication. Digital communication means we can talk to each other at a distance. I run a giant video bridge where we have 100-site video conferences where everybody talks to everybody, where each of those sites is not just a person, it's a workgroup. Digital computing means you can store the knowledge of the world and get access to that. The profound piece of digital fabrication means you can bring the campus to the student, not the student to the campus. Once you have a basic set of tools, you can make all the rest of the tools.
Next year we're starting a new class with George Church that we've called "How to Grow Almost Anything", which is using fab labs to make bio labs and then teach biotech in it. What we're doing is we're making a new global kind of university.
We're inventing this new kind of global university and part of what I like about MIT is John Reed, who was chairman of MIT's Corporation, came to see what this was all about. And instead of in any way being threatened he was just delighted to see it. His comment was, "This is how you change MIT. Change the world, MIT will catch up to it." There's a core set of skills a place like MIT can do but it alone doesn't scale to a billion people. This is taking the social engineering—the character of MIT—but now doing it on this global scale.
This isn't my core competence. I know how to invent the machines, but I can describe what's been happening as they spread. To understand the economic and social implications, look at software and look at music to understand what's happening now for fabrication. Software at one time was Microsoft or IBM. A few exceptional people could write it for themselves but for everyone else it was Microsoft or IBM. Open Source came along. There's a brief spike of: "Yippee, it's free. Nobody ever pays anything for anybody ever. You still have Microsoft or IBM now but, with all respect to colleagues there, arguably that's the least interesting part of software. If you think about app development, you can write little scripts for yourself, you can write an app for ten people, or 100, or 1,000, or a million. In terms of writing software, there's powers of ten. Mainframes didn't go away but what opened up is all these tiers of software development that weren't economically viable.
If you look at your phone and look at the diversity of the apps you use and how you use them, how they're being developed and sold, some are given away, some you pay a few dollars, some you pay more, but they're being developed and sold into markets that just weren't viable on the scale of Microsoft or IBM. A string of data becomes an algorithm, becomes a program.
Now look at music. Music was the labels or you'd play your piano. Napster comes along: "Yippee, it's free. Nobody ever pays anybody for anything." In software, copy protection failed—easily circumvented by dishonest people, irritating to honest people. Copy protection doesn't work anymore. In music there was digital rights management—easily circumvented by dishonest people, irritating for honest people. Amazon now sells you tracks without protecting them, but they make it easy to buy and sell. The labels fought it tooth and nail, now it's beginning to finally turn around. If you look at music development, the most interesting stuff in music isn't the big labels, it's all the tiers of music that weren't viable before.
You can make music for yourself, for one, ten, 100, 1,000, a million. If you look at the tracks on your device, music is now in tiers that weren't economically viable before. In that example it's a string of data and it becomes a sound. Now in digital fab, it's a string of data and it becomes a thing. It doesn't replace mass manufacturing but mass manufacturing becomes the least interesting stuff where everybody needs the same thing. Instead, what you open up is all these tiers that weren't viable before.
Now, in turn, what is it good for? The answer to that in some ways is almost the opposite of what you think, which is you can make all kinds of stuff but the real value we're seeing in digital fabrication is one step removed—it's the benefits of having made it. To understand that, remember Google doesn't sell search; they give away search, and they sell the benefits of having searched, which is advertising. Facebook doesn't sell talking to your friends; it gives away talking to your friends; it sells the benefits of having talked to your friends. It took about ten years for the dot com industry to realize pretty much across the board you don't directly sell the thing. You sell the benefits of the thing.
To understand what that means for digital fab, the most obvious thing you can do is invent a widget and sell it. You invent the widget and then you go to China and mass-manufacture it. Now, it happens we're working very closely with Shenzhen. There's an annual meeting of all the fab lab network. It's in Boston in 2015. 2016 it's in Shenzhen because they're pivoting from mass manufacturing to enabling personal fabrication. We've set Shenzhen as the goal in 2016 for Fab Lab 2.0, which is fab labs making fab labs.
To rewind now, you can send something to Shenzhen and mass manufacture it. There's a more interesting thing you can do, which is you go to market by shipping data and you produce it on demand locally, and so you produce it all around the world. There's a parallel with HP and inkjet printing. HP's inkjet division is in Corvallis, Oregon because they had to hide from Palo Alto because they were told that inkjet printing would never scale, it would never be fast enough. But their point was a lot of printers producing beautiful pages slowly scales if all the pages are different. In the same sense it scales to fabricate globally by doing it locally, not by shipping the products but shipping the data.
What is work? For the average person—not the people who write for Edge, but just an average person working—you leave home to go to a place you'd rather not be, doing a repetitive operation you'd rather not do, making something designed by somebody you don't know for somebody you'll never see, to get money to then go home and buy something. But what if you could skip that and just make the thing?
Vicente Guallart was a colleague who started the first fab lab in Barcelona. He's now the city architect, the planner of the future of Barcelona. He's putting fab labs in every district in the city as part of the urban infrastructure. There, they consider IKEA the enemy because IKEA defines your taste. Far away they make furniture and flat pack it and send it to a big box store. Great design sense in Barcelona, but 50 percent youth unemployment. A whole generation can't work. Limited jobs. But ships come in from the harbor, you buy stuff in a big box store. And then after a while, trucks go off to a trash dump. They describe it as products in, trash out. Ships come in with products, trash goes out.
What they want to do is what they call DIDO: data in, data out. The bits come and go, globally connected for knowledge, but the atoms stay in the city. The idea is you have fab labs in every district in the city, then when you want furniture or consumer goods or all of that, instead of working to get money to buy products made somewhere else, you can make them locally. You might pay somebody else to make it, or you might do it, but it all stays there. The cities provide electricity and light and sewers. Now it's this new notion of infrastructure if they provide the means to make stuff as part of the infrastructure of the city.
In Barcelona's case, the attraction is whether or not you make anything any different from what you're buying today, it means you can make many of the things you consume directly rather than this very odd remote economic loop.
Let me give you more examples. To talk about what you can make, again, today it requires ten different machines in a fab lab. In twenty years it's all integrated in one machine. But a good index is what people do make in fab labs or the How to Make class and it's awfully close to the range of things today. One good example is furniture. Anything IKEA makes you can make in a fab lab. The biggest tool is a ShotBot 4'x8'x1' NC mill, and you can make beautiful furniture with it. That's what furniture shops use. You can plot out custom furniture. Another example has to do with mobility. People make bicycle frames. There are serious projects making DIY cars. One step before that is super go-carts, and there are some very serious project making cars. Boats are made in fab labs. Consumer electronics—you can make antennas, radios. There's a couple of surprisingly successful DIY phone projects, and the most interesting part of the DIY phone projects is if you're making a do-it-yourself phone, you can also start to make the things that the phones talk to. You can start to build your own telco providers where the users provide the network rather than spending lots of money on AT&T or whoever. To a surprising extent almost any of the things you buy today you can make. There's consumables, but you can build with them using the tools in the fab lab.
Let's keep playing through the benefits of doing it. One is this economic one. We helped the White House plan a White House Maker Faire and we set up a mobile fab lab literally outside the Oval Office. This is one of the most sensitive places at the White House. Even if you have a White House badge, you cannot stand outside the window of the Oval Office because it's such a sensitive place. The White House guards were going crazy because we had all our big lasers and machines there. President Obama loved it. What was going on was the administration couldn't directly say to American manufacturing, "You're Wang, and Prime, and Data General," but they could demonstrate it. We had a fab lab at the World Economic Forum last year for heads of state and CEOs and it's the same thing. Traditional manufacturing is exactly replaying the script of the computer companies saying, "That's a toy," and it's shining a light to say this creates entirely new economic activity. The new jobs don't come back to the old factories. The ability to make stuff on demand is creating entirely new jobs.
There's one Danish fab lab that's been focused on incubating businesses and they counted in ten years the community lab made a thousand jobs and 300 million euros in turnover. Multiply that by all of these labs. The new jobs just aren't coming back to the old factories.
To keep playing that forward, when I was in Barcelona for the meeting of all these labs hosted by the city architect and the city, the mayor, Xavier Trias, pushed a button that started a forty-year countdown to self-sufficiency. Not protectionism. Globally connected for knowledge but the notion is Barcelona produces what it consumes. Shenzhen is pivoting to help provide the technology for it. And that's what the White House Maker Faire was about.
Why I am helping Shenzhen as an American is two levels deep. My ability to do everything I'm describing rests on a global supply chain that crucially passes through places like Shenzhen. I need high-torque efficient motors with integrated lead screws at low cost, custom-produced on demand. All sorts of the building blocks that let us do what I'm doing currently rest on a global supply chain including China's manufacturing agility. The short-term answer is you can't get rid of them because we need them in the supply chain. But the long-term answer is Shenzhen sees the future isn't mass producing for everybody. That's a transitional stage to producing locally.
To be clear, we're not telling people they should become part of this. Each of these doublings is people opting to join. But they’re rich, poor, north, south, east, west, rural, urban, and it's all the same person basically. This leads to the social engineering. My description of MIT's core competence is it's a safe place for strange people. These anomalously inventive people that wouldn't function in normal society fit in a place like that.
The real thing ultimately that's driving the fab labs ... the vacuum we filled is a technical one. The means to make stuff. Nobody was providing that. But in turn, the spaces become magnets. Everybody talks about innovation or knowledge economy, but then most things that label that strangle it. The labs become vehicles for bright inventive people who don't fit locally. You can think about the culture of MIT but on this global scale.
My allegiance isn't to our border versus anybody else's border, it's to the brainpower of the planet. I don't know how far this goes. Bright, inventive people whose lives are being transformed by this—call it one in a hundred—billions of people on the planet means tens or hundreds of millions of these bright, inventive people that are exactly the kind that keeps me happily based at MIT. We find them in Arctic villages and African shantytowns. My allegiance isn't to any one border, it's to the brainpower of the planet and this is building the infrastructure to scale to that brainpower.
First let me pause to relate digital fabrication to digital communication and digital computing technically and then play out some of the implications. Our modern computer architecture dates to von Neumann, and you can, in a sense, trace from von Neumann back to Turing. Both of those were accidents. Turing's machine was never meant to be an architecture, it was a theoretical construct for a proof. Von Neumann wrote things he considered profound. He never really wrote about his architecture. The most he did was he wrote a report on how to program the EDVAC. Turing's machine and von Neumann's architecture are completely unphysical. The best way to say it is the head of a Turing machine is distinct from a tape, and the reason that's so important for being unphysical is a patch of reality in nature takes time to transit, stores state, admits interaction and occupies volume. All those resources are coupled.
In computer science there's a fiction that they're unrelated. Computing happens in a pretend world that we then try to make work in a real world. A lot of what's hard now in computing—programming multicore computing, cache concurrency, back-plane bandwidth—is like the Matrix, going from the pretend world to the physical world. There's a completely different parallel history of computing where you make hardware look like software.
If you zoom from transistors to microcode to object code to a program, they don't look like each other. But if we take this room and go from city, state, country, it's hierarchical but you preserve geometry. Computation violates geometry unlike most anything else we do. There's an independent history of computing where you make hardware look like software, and so computer science scales like physics because it's based on physics. It turns out in many ways that's easier, not harder to do. The reason that's so important for the digital fabrication piece is once we build molecular assemblers to build arbitrary systems, you don't want to then paste a few lines of code in it. You need to overlay computation with geometry. It's leading to this complete do-over of computer science.
There's a lot of hype now about the Internet of Things and that's a strange one. I did early work with a number of the Internet architects on what became called Internet of Things. The core architectural principle is the Internet succeeded over the Bitnet because what it does is defined by what you connect to the network—the state is at the edges of the network. A lot of what's called Internet of Things today is actually Bitnet of things, meaning it's dumb devices connected to central sites you can't control. What makes the Internet work is the state is pushed to the edges so you don't need central control to invent new applications.
If you take digital fab, plus the real sense of Internet of Things—not the garbled sense—plus the real future of computing aligning hardware and software, it all adds up to this ability to program reality. We're going to bring the programmability of the digital world into the physical world, and it's going to be much bigger than the earlier digital revolutions because it's out here where we live.
Go back to Wang, and Prime, and Data General, there's a whole bunch of incumbent entities across how we live, learn, work, play, how we divide all of those. The next Silicon Valley isn't a valley. There's this race for what is the next place going to be? When you connect digital communication, computation, and fabrication, what you do is you create networks. There are collaborating networks where you can see people, you can interact with people, you can share content and crucially, bits become atoms, atoms become bits. I can do something, I can put it into the computer, it can come out on your side and become a thing again. I run a giant video infrastructure and I have collaborators all over the world that I see more than many of my colleagues at MIT because we're all too busy on campus. The next Silicon Valley is a network, it's not a place. Invention happens in these networks.
In Silicon Valley, one of my students, Jason Taylor, who did a thesis with me on molecular quantum control, is in charge of Facebook's infrastructure. He's building their data centers and all the scaling of it. Another one of my students, Raffi Krikorian, who did early work with a dear colleague, Danny Cohen, on the beginning of Internet of Things, ran Twitter's infrastructure. A number of former students are in these unexpected places in the tech revolution. It's not just a historical accident. It has to do with being grounded in reality and thinking deeply. I was at Bell Labs before deregulation, which was one of the most wonderful research environments I was ever in, and it was merciless. People would come and get mowed down and be challenged, but then they would push back. You weren't meant to be there forever. It would forever turnover and there was endless energy in it. MIT, if you add up businesses from it, it's the world's eleventh economy. It's trillions of dollars of volume. Fewer single billion-dollar companies but lots of $100 million companies.
You earn your way at MIT from what you do each day. Your pedigree doesn't matter, and there's all kinds of turnover and all kinds of energy. When Edwin Land was kicked out of Polaroid, he made the Rowland Institute, which was making an ideal research institute with the best facilities and the best people and they could do whatever they want. But almost nothing came from it because there was no turnover of the gene pool, there was no evolutionary pressure.
John Bell's theorem was published in the Journal of Physics, which was designed only for the smartest people, ordinary people couldn't publish there. It expired because there was no turnover and no evolutionary pressure. The way it's related to this conversation is a lot of tech industry is recreating a failed history of the wrong way to do research, which is to believe there's a privileged set of people that know more than anybody else and to create a barrier that inhibits communication from the inside to the outside rather than recognizing the attributes: you need evolutionary pressure, you need traffic, you need to be forced to deal with people you don't think you need to encounter, and you need to recognize that to be disruptive it helps to know what people know. You do your homework, you know what people know, then you can turn around and blow it up, but against a background of having done your homework.
Right now I think a lot of the tech industry is in the process of getting that wrong. But the resolution isn't going to be a better billion-dollar company. Business is going to move into distributed networks and education is moving into distributed networks. One person in one fab lab in one village can be a node in a network doing economic activity and doing research and getting education. It turns on its side all of our organizations.
For me the hardest thing isn't the research. That's humming along nicely. It's that we're finding we have to build a completely new kind of social order and that social entrepreneurship—figuring out how you live, learn, work, play—is hard and there's a very small set of people who can do that kind of organizational creation.