 Good afternoon, everyone, and welcome to the Emmy Graduate Seminar Series. Today's lecture is jointly sponsored by the Purdue Engineering Distinguished Lecture Series and the Feddersen Distinguished Lecture in Mechanical Engineering. Now, the Purdue Engineering Distinguished Lecture Series started pretty much in 2018, and this is the first one of the 18-19 academic year, and the series really brings outstanding, renowned experts around the world to Purdue Engineering to engage in meaningful conversations regarding grand challenges in their fields, right? So, without further ado, I would like to introduce Dr. Karthik Ramani, the Donald Feddersen Chair of Mechanical Engineering. Thanks, Erwin. So, it's a pleasure for us to have Neil Gershenfeld, but before that I wanted to say a few words about Don. Don Feddersen was one of the top entrepreneurs from Mechanical Engineering, went on to become a partner in Charles River, but also was the principal behind a very famous computer-aided design software called ProEngineer, later Creo, and now ThingWorks and so on. So, Don was a big fan of merging digital and physical, and to that end we have Neil, and his topic title today is From Bits to Atoms. I wanted to introduce Neil. Neil is the Director of MIT's Center for Bits and Atoms and works at the boundaries between digital and physical worlds. He's also an artist, engineer, physicist, scientist in some ways. His settings that his work has been publicized include the New York Museum of Modern Art. Also, he's been associated with World Economic Forum and also Popular Mechanics 25 top makers, and his voice has been heard around the world in various settings, but more importantly, he's the founder of a growing network of over 1,000 fab labs, and he's going to talk maybe a little bit about it and the Fab Academy. There are a thousand Bechtel design innovation centers that are networked and learning from each other. Dr. Gershinfeld's education includes physics honors at Swatmore College, PhD in applied physics from Cornell, and he has been at MIT for a while, so I'm without much ado, I wanted to give it to Neil. Thanks. And for people in the back, there's a number of seats up front if you're brave. So I direct the Center for Bits and Atoms at MIT that works on the boundary of computer, volume down a little bit this feedback, that works on the boundary of computer science and physical science. And so we did things like, this was the first significant faster than classical quantum computation, or we showed how to do cryptography and materials, or how to do universal logic in microfluidics. We were part of creating a minimal synthetic organism, designing life in a computer. This was part of one of the first things on the idea of an internet of things. Those are kinds of projects we've done, it's this boundary of computer science and physical science. But the people who have done it have gone on to do things like, one of my students Jason built and still runs all the computers at Facebook. He created all of their infrastructure. And curiously one of my students, Rafi, had the exact same job at Twitter. He built all the computers at Twitter. And I'm not a computer scientist, it's not a computer science program, but the way you can think about what made that possible is I taught them to not believe in computer science. I'm happy to take credit with the observation that computer science was one of the worst things ever to happen to either computers or to science. Because it's unphysical. And so one of the things I want to do in this hour is explain why that so, and what happens if you actually accept that computing happens in a physical world. If you're in the back, there's seats up in here. Another thing I want to explain is when I was in high school, I desperately wanted to go to vocational school, where you would learn to weld and fix cars and do cool stuff. And I was told, no, you're smart, you can't. Just didn't make sense. I was being punished, I had to sit in a room instead of welding. And I was at Bell Labs, I had union grievances because the worker said, no, you're smart, you have to tell somebody else what to do. And that's a mistake that dates all the way back to the Renaissance. And I also want to talk about why I wasn't allowed to go to vocational school and what lies behind that. So the original sin in computing you can think of as this. Alan Turing developed the Turing machine, sort of the heart of modern computing. And the Turing machine, which he used to prove basic results about computing, the head is separate from the tape, which sounds obvious, but it's completely unphysical because it means storing information is separate from letting information interact. And it echoes all the way down through modern computer architecture. So to a computer scientist, this is what you get taught. These are models of computation. This is how a computer scientist divides what a computer can do. All of these are fiction. To a physicist, there's only one kind of computation in the universe, which is space, occupies space. It takes time to transit because there's causality. States persist and they interact. That's it. That's the only thing the universe knows how to do. Anything else is a pretend fiction, and eventually, like the matrix, there's a fundamental problem going from the pretend to the real world. The current state of computing is sort of like Metropolis, if you've seen the movie, where in Metropolis, people up top would waltz around in the garden while people down below would move the levers, and when you're in the garden, you don't know what's happening to the levers. And that's sort of where the state of computing is today. And so what I want to talk about is... So in Metropolis, there is a revolution where the people in the basement took over the garden. What happens when we do that? So to start tracing historically, upper left is Vannevar Bush, who at MIT... He was a grand old man of lots of things at MIT. What he's most known for is he created the post-war research establishment. He created the idea of a National Science Foundation. But this is his last big project. He made the last great analog computer. It was a room full of gears and pulleys, and the longer it ran, the worse the answer got. So that was the differential analyzer. It got digitized. That became the whirlwind, which filled the building at MIT, which was the first real-time computer. That got transistorized. This was done at Lincoln Labs. And that's where things like virtual reality were invented, by Ivan Sutherland. And then that got commercialized. That's the PDP spun off from that. And so those of you who are currently looking at your telephone what's in the phone you're looking at came from that picture. So that's Ken Thompson and Dennis Ritchie at Bell Labs, inventing Unix. And that's the operating system running in your phone. And not only that, email, the internet, video games, word processing, all happened at that historical moment when computing became accessible not to a corporation, but on the scale of the work group. Now, one of the students who worked on that room full of gears and pulleys was Claude Shannon. And Claude Shannon was so irritated by the room full of gears and pulleys that he invented digital, literally. So he wrote the best master's thesis ever. If you want inspiration for what a good master's thesis is, read that. It's beautifully written, and in his master's thesis he invented digital logic, as we now understand it. Just enormously consequential. But then he went on to Bell Labs, and at Bell Labs he invented information theory. And it was a profound step. So my wife's Lara's father at Bell Labs helped invent analog amplifiers and feedback and switching and subsea cables, but signals got worse with distance. What Shannon showed is rather than sending an analog wave, you could send a symbol. And it's not ones and zeros. Digital doesn't mean ones and zeros. It means symbols. But if I send you a symbol, he showed if the noise is above a threshold, you're doomed. It's just guaranteed to get it wrong. But if the noise is below a threshold, for a linear increase in the resources, how much energy, how much time goes into representing the symbol, the error goes down as an exponential. There are many exponentials in engineering. So the linear increase in representing the symbol versus the exponential error in decoding it means an unreliable device can operate reliably. So the internet works. We can talk to China because of that threshold property. It's called a threshold theorem. So in turn, John von Neumann applied it to computers. So rather than an analog computer that gets worse with time, he showed if you think of computing as communicating, a message goes through the computer. By doing that same trick, an unreliable computer can operate reliably. So those are the ideas that gave us digital computing and communication. It's a scaling property. It's the scaling property of decoding a symbol at the heart of it. And so that's the digital revolution. So then to start tracing forward, Gordon Moore in 1965 made the most important graph in history. So he was one of the founders of Intel. And what he plotted was the number of transistors on a chip. And on a linear graph, it looks like nothing's happening. But if you take a logarithm, which measures doubling, they lined up. And he said, wow, it's doubling, which means it's exponential, not linear. And he said, what if this goes for 10 years? And so another thing I really recommend you read, he wrote this amazing paper cramming more components onto integrated circuits. And in that paper, he basically foresaw everything that happened since. The whole implications of a digital revolution. Now, on a linear scale, which is how we perceive, it looked like nothing was going to happen, and then he was forecasting a revolution in the 70s. But on the log scale, you can see it's a straight line. So here's what actually happened. Around 1970s, the Altair appeared. And so for people like me, it was life-changing. When the Altair first appeared, this was the first personal computer, but it wasn't useful. When it first appeared, all you could do was flip some switches to load a program and then watch the binary lights blink. That was it. That was all it did. But two people started a company called Micro-Soft to write software for it, and eventually they took the dash out. And it came to Palo Alto, and some people had a meeting of what they called the Homebrew Computer Club, and it eventually became Apple. The whole generation started on that. And then what I'm plotting on the left is what actually happened. That's the number of transistors in an Intel chip, and it scaled as a straight line on a log plot for 50 years. Around about two or three years ago, it started to roll off. So Gordon Moore was wrong. It was in 10 years. It was 50 years of exponential scaling of digital communication and computation. He could see all the way back in 1965 that trend happening. So the reason we're here today, and for people in the back, there are some seats up here if you want to be brave and come forward. And maybe move into the middle of rows if you have empty seats to make room for people coming in. What I'm plotting here is a measure of digital fabrication. This is the number of fab labs, and I'll tell you all about what a fab lab is. Fab labs have been doubling for a decade. And so that's now not digital communication or digital computation. It's digital fabrication. And we're 10 years into it. There's more data than Gordon Moore ever had. And so what I'm here to ask is what if that continues for 50 years? What if we have 50 years of scaling performance of digital fabrication, just like Gordon Moore in 1965 saw the scaling of digital communications and computation? So to start tracing that, this is a 1955 article. 1952, there's a seat right here if you really want to be brave. In 1952, MIT invented computerized manufacturing. So I mentioned the whirlwind was the first real-time computer. Right around then, jet aircraft were emerging, and there were parts that were too hard to make by hand by turning cranks. And so there was this bridging between unrelated worlds, this idea that you could take the real-time computer and have it turn the cranks on the machine to make a part for jet aircraft. So that was the birth of computerized manufacturing in 1952. And there was just one of those. Just like there was one whirlwind, there was one NC mill, and there's one planet Earth. The lab I run at MIT, Center for Bits and Atoms, started with an ambitious NSF proposal. We wrote a proposal to say we wanted one of every machine to make anything. And I guess we got NSF on a good day, and that's what they funded. So I run a lab that has nano, meso, micro, macro, input, output, one of just sort of every machine to make anything all smushed together. So that's what I currently run. That's the descendants of the million-dollar NC mill. But I had a problem, which was it would take years to learn to use them. So I started a class aimed at a few research students to learn to use the machines. And every year it looks like this room, hundreds of students want to take the class. And so it was just how to make almost anything. It was just skills to make stuff. And the students did projects. And so I'll show you Kelly. She was a sculptor, no technical background. This was her project. I'm Kelly, and this is my screen body. Do you ever find yourself in a situation where you really have this screen, but you can't because you're at work, or you're in the classroom, or you're watching the children, or you're in any number of situations where it's just a screen body is a portable space for screen. When a user screens into a screen body, their screen is fine. But it's also recorded for later release, where and how a user chooses. That's a web browser for parrots. They have a cognitive ability of a young child to go crazy left home alone. Let's surf the internet. This is an alarm clock you wrestle with and prove you're awake. This is a dress instrumented with sensors and spines to protect your personal space if somebody creepy comes too close. This happened so consistently year after year. I realized the students were answering a question I didn't ask. I was asking how to do digital fabrication. Bits to atoms and atoms to bits, but not why. They were showing the killer app of digital fabrication is personal fabrication. Not mass production, but production for a market of one person. This is Ken Olson. He founded DEC. DEC created the computers that created the internet. Outside Boston, it spawned DEC, WANG, Prime, Data, General. The whole computer industry, making mini computers, was outside Boston. When PCs came, the whole industry failed. Ken famously said, you don't need a computer at home. The whole mini computer industry didn't survive computing become personal. But you don't have a computer at home for inventory and payroll. You have it to listen to music and talk to friends and make all the things you. What we're heading to is the same thing for manufacturing. A lot of manufacturing is like the mini computer industry. It's going to get blown up when manufacturing becomes personal. Thinking about that scaling, the mini computer, thousands were sold. Thousands is the number of cities on earth. On the mini computer, when it became accessible to a work group, not a whole organization, that's when the internet and all the applications of computing emerged. Inspired by that parallel with colleagues, I started setting up fab labs. You can think about my lab as the $10 million research lab of things like nanoscience tools, containing the $1 million lab of manufacturing tools, but within it was the $100,000 lab of just tools to make things. That's what we packaged up and started sending around the world. This is $100,000, two tons, fills a room. It includes a 3D printer, but it also includes large format precision machining, surface mount rework, embedded programming, molding, casting, cutting, composites, about 10 or so processes to do digital fabrication. The PDP wasn't a computer, the mini computer. There was a processor unit, there was a communication unit, there was a storage unit. It was about 10 units you plugged together to compute. In the same way, this is 10 different units that you plugged together to make stuff. Again, it includes the 3D printer, but the 3D printer is just one little corner of digital fabrication. If you think about all of that as effectively a machine, with those tools you can do this. All of these aren't made by businesses, they're not made by research students. It's not a startup, it's just in a fab lab making boats, bicycles, furniture, consumer electronics, production tooling. This lab I'm showing you, if you've seen a picture of an island being destroyed by a volcano in Iceland, that's this, it's Vespiner, and they're fine now. In that little island off the coast of Ireland, they have all of these tools. In Vespiner and any other lab, you can make all of this. Then an accident happened, which is we opened one in Boston through a Ghanaian collaboration. We opened one in Sakhandi, Takaradi, through a South African collaboration that led to one in Pretoria, that then led to one in Soshungovi in Apartheidira township. Then through a Norwegian collaboration, it led to one in Ling Sided above the Arctic Circle, so far north that satellite dishes look at the ground, not the sky, because you have to look down to see the satellites. Every time we opened one, somebody else wanted one, and so they've been doubling every year and a half, and they cover Earth, these labs. But all these labs share those same tools, just like the internet, so people in projects are shared across them. This is a lab in Detroit working with at-risk youth. This is one in Bhutan, where Bhutan doesn't measure gross domestic product. They don't measure money. They measure gross national happiness, which doesn't mean they're happy, but it means they measure well-being. Are you healthy? Are you stressed? Indicators like that. This is to make gross national happiness physical, to embody it. This is a mixed community in Holan, in Israel. That's at the Protestant Catholic boundary in Northern Ireland. That's in Arts Colony in Maine. That's with Alaska Natives, where there's fabulous cultural traditions, but terrible alcoholism, suicide, unemployment, and this is mixing traditional crafts with modern digital fabrication. We didn't pick these places. You can't wake up in Cambridge, Massachusetts, and pick this, but each time we open one, somebody pulls one. These are what some of these places look like. Then we had a problem, like the MIT class, that bright inventive kids would show up in these labs, who pretty consistently were considered problems in the schools and bad employees because they were inventive, and so they didn't like following rules. They were kind of refugees from... Everybody wants innovation, but makes a lot of rules to prevent it. They were refugees from all of these places with rules and would come to the labs, but then they'd kind of fall off a cliff. So we started teaching a program called the FAB Academy. You can think of MIT or Purdue as a mainframe. You come here and get processed. You can think about online class as time-sharing. You're like a terminal connected to the mainframe. The way this works is students have peers with mentors and machines and local work groups in the labs. Then we link them globally with video and content sharing. So Guillaume was in a little lab in a forest outside Barcelona, and he wanted food. So at the beginning, all he knew how to do was make a sketch. So this is a sketch of his aquaponics system. Then he's learning CAD tools, so here's a model of it. Then he's learning laser cutting and he's making prototyping a small version. Then he's learning large format machining. He's making a big version. Then he's doing plumbing and instrumentation for it. Around here, he's learning to design circuit boards and write microcode. So he's working on the control system for it. And so here's his Aquaduino, if you know what an Arduino is, to run the aquaponics system. And then he's putting all the parts together. And then I love this picture. Around week 16 in the class, he's eating lettuce made on it. And so he's not an engineering student. He's not a team. This isn't a startup. The tools are now so accessible. He can do all of these different skills in one project to make food. And now this has started a whole urban agricultural product. You can grow food much more efficiently than digging the dirt by providing precise agricultural inputs. And this has started an urban agriculture aquaponics initiative. So projects like that have led to things like this. In the middle is Barcelona's mayor. Barcelona has a fabulous design history, but over 50% youth unemployment. A whole generation can't leave home as we understand it and work. My counterpart there, who started the Fab Lab, became the city architect. And what he started doing is filling the city with digital fabrication facilities, these Fab Labs, as urban infrastructure. So if you live in the city in the same way it provides electricity and clean water, it now provides the means to make as urban infrastructure. And currently they think of Barcelona as a product to trash conversion device. Products go in one side and trash trucks go out the other side. What they want to do is they want to have things stay, physical things stay in the city, but data comes and goes. They want data to go in and out, but atoms to stay. And so this is Barcelona's mayor and he started a 40-year countdown to urban self-sufficiency. A few percent a year of growing food, producing energy, doing all the things I described through digital fabrication. Not as a step change, but just as this ramp of a few percent a year progressing. This is a lab we ran at the White House and the Obama Administration and currently there's legislation in the U.S. House and Senate to do the following. I'm United States Congressman Bill Foster and I'm one of the few members of the United States House of Representatives who was a scientist before entering politics. So I often tell people that I represent about one-third of the Strategic Reserve of Physicists in Congress. But when I came into work each day in physics, my first stop often wasn't to my office computer or some meeting, but to the laboratory machine shop to check on the progress of some parts that I designed for an experiment or part of an accelerator. So I can think that, I believe I can safely say that I'm the only member of the United States Congress that knows how to program numerically controlled machine tools. I'm proud to announce that I recently introduced legislation in the United States House of Representatives which supports the goals and mission of the National Fab Lab Network as in the best interests of our people and the best interests of promoting the goals of greater science and technical education, greater access to research and production tools, and empowerment of individuals to understand and use technology to improve their lives. You can think of the MFLN as a new kind of national lab in the United States that's a cloud laboratory, a national network of connected local labs. I've been lucky to have a chance to visit NATO and see the project. So the idea of this bill is it's a charter in the national interest to create a public-private partnership for universal access to digital fabrication. In the same way Barcelona is doing it, it's now to do it on a national scale in the U.S. and right now it's in the House and Senate but it's already attracted a number of commitments to do this. You don't need to wait for the bill to pass to start doing it and so if you're interested in this talk at all this is one of the interesting ways to follow up to start spreading this locally. And so at this point here's a number of the cities participating in this FabCity initiative. Detroit is one a couple of days ago I was in Oakland and while Silicon Valley has among the worst genie income inequality coefficients and the price of living is going through the roof and the workers who work for the people whose income is going through the roof in Silicon Valley live in these miles long broken down lineups of mobile homes because they can't even afford to find a place to live, Oakland doesn't want to develop that way. And so a few days ago I was in a meeting in Oakland of the city leadership plus the FabLabs and Makerspaces plus local businesses about implementing this FabCity initiative. So not just a smart city but a city that can produce housing and produce food and all the other things you need as part of the urban infrastructure. And so it's a really interesting laboratory. And so if you think about among the biggest battles today over trade tariffs, diverging income, inequality, if you can go into a lab and make something, it's kind of an end run around all of that rather than having a global supply chain, rather than having a job to get money to produce a product from a global supply chain, if you produce it locally it just dodges around a lot of those battles. Now it's not utopia and it's not a false dichotomy, it's not everything is mass manufactured or nothing is mass manufactured. But if you think about software or music, they now, you can do it yourself, you can do it for friends, you can do it on a small scale, you can do it, you can share it, you can do it on a big scale. In the same sense it makes possible manufacturing for one or ten or a hundred or a thousand or a million. You can still mass manufacture but that's the boring thing for everybody needs the same thing. You can locally produce the things that differ. So in Barcelona, Blair who runs that lab has a model where a third of the time in the lab is for money, you get paid to make stuff, a third of the time is not for money, it's for barter or exchange, it's for a kind of a post-money economy in the community, and a third is not for income, it's for you or the well-being of your community and it's really sort of creating a very different economic model. The whole assumption that you get trained to become an engineer to design products that get manufactured to get sold to computers bakes in a whole bunch of assumptions that change when consumers can actually become creators. So let's keep scaling. Today there are about a thousand fab labs one per city. The hobbyist computer, there were millions of those, not billions but millions. And so the analog there is it doesn't work to keep on spending $100,000 and getting two tons. The next step in this evolution is to use a fab lab to make a fab lab. And so here's a bunch of machines made with machines, like this was a fun version. This was a student, Nadia and Alon, and this is a TSA carry-on briefcase that's a whole little lab in a briefcase that has interchangeable huds that can, mill and cut and print, merging all of them into one. And that was an open source design but nobody else made it. It didn't work well. What it led to is we spun off a bunch of companies like Form Labs that makes high-resolution 3D printers or Shaper Tools make this really interesting router with augmented reality and a vision system that lets a little tool make a big thing. We spun off companies to make those but we realized we were, and we worked with companies like Autodesk or SolidWorks on the software tools for them but we realized we were making a mistake which is software used to be something like Fortran. You just write a program. Nobody does that anymore. You write software in modules you can recompose. And the internet beat the bitnet because with the bitnet the terminal couldn't change the mainframe but with the internet what the internet does is define by what you connect to the network not by how you built the network. So those are really well understood for networks and software but machines are like bitnet or Fortran. When you get a 3D printer, a laser cutter, it can't change. It only does what's built into it centrally. So we started looking at how you build machines more like modular software like the internet and so Nadia who you saw that did beautiful work picked up by a number of people and then this is a current student Jake and so here's a parametric machine generator. Do you want linear axes or rotary axes? Are they long or short or stiff? Here's where you compose them into how many degrees of freedom in the machine. Then this produces cut files to be able to assemble the machine. Then what's going on here is subtle. It's really interesting. The machine doesn't know what it is. This is a data flow network and that's a router and so the machine has no idea what it is but then you add software modules that overlay on the hardware modules to make a graph of the logic of the machine. So this was at an arts colony where we have a lab and they wanted to print clay and so Jake threw together a workflow for a virtual machine on a physical machine to make a clay plotter. And so the stateless machine is a data flow graph passing the software graph that tells it what to do and so again in this case it's extruding clay. But then the same parts this is a three axis cutting tool. This is a drawing tool. This is a thing that plays a piece of music by Steve Reich called Music for Blacks of Wood. That's usually a room full of people but Jake made a machine to do it. And these are all the very same parts but just composed in different ways to do rapid prototyping of rapid prototyping to make the machine as easily as a project on the machine. And so that's what it looks like to get to a million. The equivalent of the hobbyist computers is you don't go to a fab lab to use it and you don't buy 3D printers and laser cutters. You go to a fab lab to make the next machines which are these modular composable machines. So computing got to a billion which is roughly the number of people on earth and there we have a problem. The thing you can't read at the top is the inventory in a fab lab. Fab labs can do everything I said but you need to buy electronic components. You need to buy resins for casting. You need to buy bearings. There's a whole bunch of inputs you need to go into the lab to make everything. At the bottom is the inventory from Digi-Key, the electronics vendor, to make Jake's motor controllers, just the electronics components. And so you can make anything but you need this global supply chain for the inventory in. And so here our inspiration is you, is us. Life is based on an inventory of 20 parts, the amino acids. The amino acids are just like Lego. I'm showing them the upper right. And what's interesting is they're not interesting. They're hydrophobic or hydrophilic. They're basic or acidic. They just have 20 or so properties. But by composing these parts together you build, this is how you smell, that's how you move, that's how you think, that's how you reproduce. All of that machinery is made out of just 20 parts by composing them. And so inspired by that we started thinking about the idea of digital materials. If you go back to digital communication or computation, manufacturing today is analog. You continuously deposit or remove material. But biology is digital. You code the construction of these building blocks. And it lets you detect and correct errors. It lets you determine global geometry from local constraints. It lets you unbuild as well as build all these properties of code but in materials. So we started scaling Nano-Lego down. So if we zoom in on Digi-Key, Digi-Key stocks not 500,000 resistors, but 500,000 types of resistors, and 500,000 types of capacitors, and 500,000 types of connectors. But conceptually they're all made out of only three properties, conducting, insulating, and resistive. So we started making Nano-Lego. And then here's a design tool to design, in this case electronics, by how you assemble Nano-Lego and then model the physics. And then this design tool exports the design to an assembler that's like the kid playing with Lego but it places them. And then this is a first generation assembler that takes a feedstock of the micro-Lego and assembles it into functional three-dimensional volumes. The idea of digitizing the materials then led to this project which is today when you make a jumbo jet you need tooling the size of the jumbo jet and we looked at how you take the big structure but decompose it into, again, these discrete little parts you link together and we showed if you do that with carbon fiber you can set the world record for the highest performance ultralight material and that led us to make these, I think, adorable robots that locomote on the structure they're building to build the structure and so little robots can collaborate. We're doing this to make aero structures with NASA to build space structures to have little robots building discrete cellular structures on big scales and then we use that to do things like the first fully morphing aircraft. This was the first one in MIT's wind tunnel that can change shape by making it out of these parts. We've since done it in one that the size of one of NASA's biggest wind tunnels. So, to keep scaling, computing then reach the Internet of Things stage and it's approaching a trillion. The way to understand a trillion is a person has roughly a thousand things and so if you have a billion people and a thousand things you have a trillion Internet of Things and if you take the Nest thermostat, it does what the PDP does. The Nest thermostat has the capacity of the PDP but now it's in a thermostat and so computing not as a metaphor literally went from one to a thousand to a million to a billion to a trillion. And so at the trillion stage there's a problem. If you want to build not by printing and cutting but by assembling and disassembling as I showed you at the speed of inkjet printing and at the resolution of 3D printing it would take about a day to make something. But if you wanted to make say my cell phone down to the smallest transistor but up to the whole cell phone that way, it would take a million years which is a long time to wait for output. So the solution biology does is really interesting. The molecular assembler that builds you is called the ribosome and it's really slow, it runs once per second. So once a second it attaches a molecule. But you can make an elephant one molecule a second because ribosomes can make ribosomes and so you get the exponential again. So a cell can have a million ribosomes, you can have a trillion cells so while you're listening to me you're placing 10 to the 18 parts per second. And so that thought led us to look at could we do that. So you could do it in biology, this is part of the collaboration I mentioned on making a fully synthetic organism you design in a computer but there you're limited to biological materials. And so in biology there's a primary coding sequence, what you want to make but unlike CAD for manufacturing, the coding sequence doesn't simply describe the thing the coding sequence becomes the thing. The code actually becomes the object, the object is the code. That's primary structure. Then it folds into shapes, those shapes become functional units and then those become molecular machines and so that's called primary secondary tertiary quaternary structure. So it led us to ask can you do primary secondary tertiary quaternary structure for the rest of engineering to make an assembler that can assemble assemblers out of the parts that it's assembling, a self-assembler. And so we started looking at just like the hierarchy in biology can you take roughly 20 material properties and compose 20 properties to build all of modern technology. And so to give you a sense of where that is, here are five part types that you can decompose and distill smaller part types and with those five part types that's enough to assemble a motor. So here's a linear motor, then here's a rotary motor but because it's made out of these discrete part types there's many ways you can organize it. So here it's being driven out of phase then it's being driven in phase and this is made out of just assembling these five part types to make actuation and that's part of this larger project that here's the assembler I showed you before but now we've decomposed it so it's made out of its own parts so the assembler can make more assemblers and even if it's really slow you get exponential ring up in assembling assemblers. And then to design that we had to write a really interesting design tool because you have to design a program that communicates its own construction. So you need a design tool that merges design of computation, communication and construction all in one. So this is designing the self-reproducing assembler. You need to completely merge digital and physical in a single design workflow. In turn to program that you have to forget what you're taught in computer science because here now a program is a thing and so we built a programming environment where you turn computation into construction as geometry. For the experts these are the BLAS underlying high-performance computing but it lets you fly out to the whole program and so like the BLAS one of them is a dot product which contains a multiplier which contains an adder tree which contains adders and you can fly all the way into the physics but all the way out to the computation and we're explicitly representing space and time and construction in computation so we can overlay the program in the thing we're building. So at the very beginning I complained about computer science. I've shown how you can merge much more deeply bits and atoms not as metaphor but literally but the founders of computer science understood it. The computer science classes might not teach it but one of the pioneers was John Von Neumann and one of the last things he worked on was self-reproducing automata. This was at heart trying to get to the essence of life but he was asking how do you design a communication, a computation that communicates its own construction. To do that he invented cellular automata, an interesting thing but he was completely merging communication, computation and fabrication. That was a theoretical study and I've now shown where to point where you can really literally not yet make but we're on the path to making the Von Neumann assembler. Alan Turing who's really the founder of our modern understanding of computing the last thing he studied was morphogenesis and morphogenesis is an amazing problem. Your genome doesn't store your construction. Nothing in your genome says you have five fingers. The genome stores a program and it's an unusual part of the genome. It's actually read out as a program and it's one of the oldest parts of the genome. What are called morphogenes are genes that program genes and they control what are called morphogenes and by these morphogenes running a program with the morphogenes they specify things like make gradients, grow until you hit a limit, symmetry break, those sort of operations and then by following those steps you end up with you. So there's two reasons biology does it. One is compression. A billion bases can place a trillion cells but the deeper one is if you randomly perturb the genome almost anything you do would either be inconsequential or fatal but morphogenesis gives you a really interesting design space for search and this is very similar to the heart of machine learning. The heart of what machine learning does is find representations and this is a beautiful representation for design and so the same Hox genes based on the order you read them out can give you a fly or a baby and it lets you explore the space in between flies and babies and beyond them. It's very very different from how any engineering is done and so to be able to ultimately design systems with billions of little building blocks what we've been implementing is morphogenesis as an engineering workflow where you don't design a thing the program designs the thing and then you design the program that designs the program that designs the thing implementing biological morphogenesis as an engineering workflow and so what this ultimately adds up to is for exploring space what's called in ISRU in situ resource utilization is how you build a civilization in space and the model for that has been essentially you need the 500,000 resistors from Digi-Key so Gingery wrote a beautiful series of books book one is Make a Charcoal Furnace book two is Make Hand Tools and by the time you get to book seven you have a machine shop it's sort of how you can personally do the industrial revolution and the assumption has been that's what you need to do to create technology I helped start an office to do science in Hollywood and this is a program with Bill Nye and some other very interesting people that's bonus material for the movie The Martian and this is about how to go to Mars without luggage what are the minimum building blocks to bootstrap a civilization and at this point in the talk you should recognize what's behind me what I'm explaining is you don't go to Mars and make the 500,000 resistors and make the machine shop you can essentially create life in non-living materials by decomposing it to just these 20 properties you can then hierarchically compose to create all the rest of technology so one earth, a thousand cities a million towns, a billion people, a trillion things one mainframe, thousands of mini computers millions of hobbyist computers billions of personal computers trillions of internet of things all of that's happened there was one 1952 NC Mill at MIT today we have about a bit over a thousand fab labs that's where we are now I've shown you technology to get to a million personal fabrication tools machines making machines to get to a billion assemblers and then to get to a trillion self-assemblers and so like Gordon Moore in 1965 you can see the decades of scaling this ranges from today to decades in the future but you can see the roadmap for all of them is in place and the thing to take away from this is in the internet video chat word processing, video games they weren't invented after the iPhone they all came from the mini computer era and so in the same sense that's where we are today we're right at the moment the invention of the internet may sound like old people history but that's where we are right now and what we found is if anybody can make anything anywhere the way we segregate you go to school to learn you go to work to get a job you do entertainment for entertainment you do aid for aid the way we segregate them kind of all gets turned on the side and many fundamental assumptions about consumption versus creation aid, economy, infrastructure sort of how society works fundamentally get turned on their side when anybody can make almost anything and I thought the research was hard but that's actually humming along what's been really hard is we've had to spin off all these organizations like a fab foundation and a fab academy and all of these programs because it just doesn't fit incumbent organizations and what's much harder is the social invention not just the technical invention of if anybody can make anything anywhere how do we live, learn, work and play so I recently wrote a book which was a horrible experience because I did it with my brothers my younger brother Alan developed and ran the world's biggest video game studio at Activision and my older brother Joel led the national labor relations organization and what the book is about is they trust me to get the technology right but not the impact in society and they pick up what I can't do which is to really trace out when with the digital fabrication revolution it took us decades to catch up to things like diverging income inequality and spam and fake news and both the opportunity and the problems this time around we don't need to wait 50 years to confront it it's happening today but it really requires the challenge and the opportunity to grapple with this is changing very fundamental assumptions about how society works if anybody can make anything so that's the technology that's some of the impact and with that I'll thank you and take questions yeah thank you very much for the talk it was amazing you said I think there are a thousand fat labs now in the world could you explain how they were financed yeah here's what's interesting about that some were top down from governments some were philanthropic some were companies doing outreach some were bottom up by just local initiative by just community initiative it turns out that it wasn't that hard to raise the money for them but there's two lessons about financing them that very much apply if you want to do it here the first is it's much easier to create a network than one of them if you want one of them you're going to a donor and saying I need some money for my lab but if you have the organizational capacity to say we're going to create this regionally then you talk a level up to people with more signature authority and in many ways it's easier so you know I was with your dean of engineering earlier today who's going to talk to your president later and about exactly this but a conversation at that level can raise much more money by dealing with this network not as individuals so the isolated fails the power comes in the network but the bigger thing we've learned is however you fund it to start it the bigger issue is the steady state cost to keep it running and there it turns out what works is an all of the above model in that you could fund it as community infrastructure with free public access you could fund it with a membership model you could fund it from working with businesses you could fund it by spinning off businesses you could fund it by creating infrastructure and what we found is what works is all of them because any one of them unbalances it but by blending them it's much more successful and in doing that there's a common mistake which is to assume the product of the lab is the thing you make but remember Google learned after years it gives away search but it sells the benefits of searching in the same way if you look at all the labs I showed you they don't try to make money by selling products from the lab they sell the benefits for the community of making the products and it turns out that's much more significant it has real economic value but you sell the benefits of making much more importantly than just the thing you made in the lab back there the examples you've given are really cool but they have also huge infrastructures that go with them and the example is the integrated circuit found has billion-dollar fabs now and of course cell phones have satellites and billion-dollar infrastructure and fiber and that sort of thing what can you envision is going to be a structure that's going to be needed to do this at the self-assembly level you get down there so there's a couple steps into that so one step into that the boundary is surprising Sam Ziloff was a enterprising high school student who now has started as a freshman at CMU and he made a CMOS line in his garage literally he made a chip fab in his garage and he's producing CMOS chips and so they're not Pentium but he actually wanted to ask can I produce CMOS in my garage and he did and if you read about it it's a beautiful project that's not a universal answer but that's one step in but then a second step in Google did a really interesting project with us called ARA and the ARA phone was the first version of a phone the phone being a monolithic thing you buy it was decomposed into a small set of functional building blocks like radios and antennas and batteries and then you could compose those building blocks to make a phone but every phone could be different and no phone was ever obsolete because you could keep rearranging the building blocks so that's a second step in but to go back to your question and that's the heart of the slide is the fab lab today at the machine's making stage you don't source the machines you source the parts of the machines and there a week ago I was at the World Intelligent Manufacturing Summit in China and there's a really interesting pivot of the manufacturers in China that make the products that we compete with today are busy pivoting many of them to recognize that in the world of personal fabrication you don't mass produce products you mass produce the things that enable local production so things like not just normal integrated chips but chips with integrated multiple functionality with unusual packaging or precision machine components and so one step down the supply chain isn't the machine it's the machine parts but to your question the last 15 minutes were about your question in that to get rid of all of the supply chain the assembler I showed you came from a DOD project I'm running right now where somewhat fancifully I proposed to reduce the entire DOD supply chain to 20 parts and it was based on this idea of primary secondary tertiary quaternary structure starting from just these 20 material properties like semiconducting conducting insulating and we're not doing it from the bottom to the top we're doing it from all the layers at the same time but we're making good progress to show if you take the cell phone at the near the resolution of the transistors but we're already at a point where you can make a lot of what's in here from just a small set of building blocks and so the heart of the answer to your question is reducing the global supply so life doesn't need a global supply chain it just needs 20 properties in the same sense mass producing these building blocks with this small set of properties is how we're trying to eliminate the whole global supply chain and really get down to eliminating the billion dollar fab and making the chips and everything else that's one of the core research projects upstream from it that's dear to me that's central to this research roadmap but to a surprising extent it doesn't matter and what I mean by that is the mini computer wasn't like the cell phone the mini computer filled the room and it had all these racks and you had to do a lot of things to make it compute but that was enough to invent the internet in the same sense in the coming decades we'll have assemblers and self assemblers but you don't need to wait decades to live in this world today you can do it in the fab lab bit by bit we'll reduce the inputs to the lab but the outputs are already here today thank you sir see where are you? over here on the right in the front at the start of your talk you mentioned about the struggle between creativity and bureaucracy yeah I would be very curious to know what would your teachings for the younger generation for the what? for the younger generation yeah so um this is a dear question for me the um there's a surprising number that um uh these people my lab has spun off billions of dollars in businesses and MIT has spun off trillions the somebody added up the economic output from MIT businesses falls between India and Russia the world's 10th economy and that's a crazy number it's not that these people are more clever than anyone it's that they're an environment where they can create and I consider the core competence in my lab as not ready aim fire but ready fire aim if you do ready aim fire if you hit what you're aiming at we get ready by doing work then we sort of don't look we shoot in a random direction roughly but then we see what we hit um and so MIT is an environment or another way to say it I consider MIT's core competence is it's a safe place for strange people and so um those people are everywhere there's probably many people in this room um but institutionally there's lots of constraints that throttle them that prevent that and so the productivity of a place like my lab or MIT or these fab labs is not trying to legislate all the steps of what somebody does but creating these creative environments where they can be productive which isn't freedom from constraints but it involves ready fire aim and lots of traffic and lots of interactions so given that note in arts and artic villages or african shantytowns we find exactly these people and again I mentioned they're considered problems in the schools or businesses and so part of what we've done is create room for them but I wouldn't underestimate your own ability to make your own space so like the fab lab in social Gov um is with the bright youth council and so what's the bright youth council the local this was an apartheid or a township school was really an instrument of apartheid or oppression so students there were still being taught to lay bricks or fold sheets and make beds and so kids from the township said we don't want to do that they invented an organization the bright youth council and just pretended that it was real and we met them and they were wonderful and we gave them the lab and it's flourished ever since I wouldn't underestimate you know if you pretend enough at something you can make it real rather than being frustrated by organizational constraints you don't fit to invent your own organizations where you do fit that you know I'd go back to that number of MIT spinning off a trillion dollars almost none of that business was based on here's a problem here's a solution, here's a business plan most of it was based on a vision of change in the world or as often as not a culture of how people wanted to spend their time their days working more power than you think to do it but with the note that you need to do it in networks you can't do it by yourself plugging into networks is the way to do that so you know the thousand labs I showed was really driven bottom up by people opting in to say I want to be part of inventing that new world I'll repeat it if you say it I'll repeat it yep oh now it's working okay I'll repeat your question if it's not working okay it seems to me that you can't decompose the assembly there's going to be limits to what can self assemble or what modules can be used to make things like your little assembler modules had copper windings and motors and everything and that's something that doesn't seem like it's going to be easily self assembled and something like gorilla glass on a phone that's a kind of exotic material and there are things that might require particular trace elements or something that you can't just kind of grab out of 20 components and then just another question should there be a footnote for K. Rick Drexler in all of this okay so two notes on that on the first I don't agree with your question but it'll take me decades to prove you wrong and what I mean by that is we made a mistake initially by trying to start here and going up hierarchically and what I was showing on this slide was the what I was showing here is sorry let's go back to here this is nanometer it's like lego but this is nanometer lego and when the building blocks get that small a number of interesting things happen so initially we started at the bottom now we're making all of those so the easy part I disagree with is the coils and magnets are at the level of modules made out of functional elements but at the very same time we're developing nanoscale building blocks just out of the conductors insulators and magnets exactly to make the structure you saw one level up so you were looking at like secondary or tertiary structure in parallel we're developing the primary structure to build out of that that's an easy one Gorilla Glass is a good example Gorilla Glass works by implanting to have surface tension there's a mismatch between the lattice with the ion implants and the surface one level down that prevents cracks from spreading and in the same way we made the first really successful morphing arrow structures because people have spent years trying to put mechanisms in them but we showed with just a rigid and flexural part we could turn it into a mechanism in the same sense once you get down to these nanoscale parts I suspect we can make something like Gorilla Glass by having lattices with incommensurate pitches that does that same thing that limits the crack propagation likewise the way we're working on nanoscale logic isn't semi-conducting we're looking on nano mechanical logic based on these parts which is having a renaissance for things like TXRX switching or H-bridges things like that so those are steps towards in fact yes I think we can do all the things you're describing out of the small set of building blocks but there's a lot of hard work to come to get there but there's no fundamental limitation against it we're working on that project for example with the materials genome initiative at NIST which today is based on roughly kitchen chemistry you mix stuff together and this question of breaking it down to the building blocks now for Eric Drexler who's one of the visionaries of nanotechnology and he's a friend as this work started emerging he gave me what I consider a nice compliment by saying I cheated which is his vision of atomically precise manufacturing the closest that's come to it is things like DNA origami but it doesn't really exist I'm doing what he promised but I did it by redefining the atom and by redefining the atom I moved the goalpost but it's very much doing what he was forecasting much sooner and the important part of that is in any one of these applications there's a dynamic range between the minimum feature size you need control over and the scale of the system and so we're not trying to go uniformly from the very smallest feature size to the very largest system size we're trying to span the dynamic range between what you do need control over and what you don't need control over there and then there's one here you talked you talked about your ready fire rain policy philosophy so could you give us an example of a product that arose from that policy from the philosophy and how it came through these are fun I'll give you a couple my lab spun off a hundred million dollar a year business making the most common sensor in the auto industry to control airbags in a crash that had the problem they were killing infants in rear-facing child seats and the sequence was I was interested in the physics of musical instruments so I visited at the time the media lab at MIT and collaborated on a project to put sensors on Yo-Yo Ma's cello we noticed his hand was interacting with our sensor fields and so it led me to do it became a student at a PhD thesis on tomography to see with electric fields then we were playing with that and that led to a collaboration with the magicians in teller to do a Las Vegas show where they had a spirit contact to contact and then an NEC executive came through the lab and saw that and said that looks like something you could use to control an airbag in a car so we took the magic trick from the Las Vegas show we packaged it up took it to the auto trade show and the car company said when can we buy it and became a hundred million dollar safety business and so I would have never picked auto safety control they would have never picked any from the Las Vegas show what was missing was not constraining here's the problem find the solution but a vehicle for us to find each other or here's another example companies approached us about being able to like drive out walk out of a store and have it scan all your products so they wanted fancy shop lifting tags to read your products and we couldn't figure out how to do it a penny's worth of material in the product to tell the store what was the product so we started looking at ways materials could send messages and we realized we had found a way to make a quantum computer and it was by programming nuclear spin dynamics to do spin-spin exchange and that led to years of work that led to the first significant faster than classical quantum computation implementing factoring in database searches very fundamental advance in the foundations of quantum computing the students who did that had to make agile radios that could do very complex pulse sequences to modulate the nuclear spin dynamics they started a company called Thing Magic to build intelligence into everyday objects turns out nobody was particularly interested in that then but the spin programmers they made for quantum computing could read silicon tags and they created the reference platform for what became the whole RFID industry which is now this giant industry for supply chain and so if you follow those histories nobody can plan them the connecting thing is they do involve the ready part there's a lot of homework they do involve the fire part you had to do something significant but the aiming part then involves not the problem finding the solution but letting us cross paths to figure out when somebody has solved something and to a surprising extent figuring out when somebody has solved a problem happens much less than trying to figure out how to solve the problem and that's a connecting thread through these and I can keep going with those I love those stories he was waiting one last question so the fab lab seems to be able to manufacture a wide range of products in terms of their functionality but how could you evaluate or ensure their quality when you have manufacturing it's an interesting question to leave on because the whole system by which we handle that now assumes command and control you assume you have a standard that can evaluate quality you have a safety agency that can evaluate safety all of that but in this fab lab world nobody asks but one observation is when I do research at MIT conventionally I'll file patents but in the fab labs patents are useless because there's no barrier to infringement in the same sense as fab labs they're starting to make cars and houses and things that could be really dangerous but you can't just slate quality or safety but it means you can't do it but what you have to do is create incentives to opt in so a good example is in software open source software has progressed to generally being much more reliable and safe than proprietary software because of the way it's not free play it's not everybody contributes but there's a structured process by which anybody can contribute and then the community can evaluate in the same sense for example in the fab academy we found it rests on a tree of relationships like in a standard agency like NIST they have the reference for the meter or for time and then you can calibrate against it for a secondary standard and then you can calibrate that for a tertiary standard in the same sense this network has grown up with a set of relationships where there's a core group coordinating and then they know a larger group and then they know a larger group and then they know the labs but you can't legislate quality or safety but you can build incentives to participate in networks to provide oversight but it's soft power it's not hard authority it's soft power for alignment there are a number of examples where it works and now it's hard but we're pushing that into new places where it's never been stretched so with that I'll thank you and I'll note if you want to know much more we wrote this book and in particular if Joel and here were now they'd be groaning about everything wrong with what I said so if you want to know what's wrong with what I said and the second two-thirds of this it's in there thank you