Open Doors
  Who Is Who  

Amsterdam - 14, 15, 16 November 2002

"This is the revolution for us"

Neil Gershenfeld

Hello! It's a pleasure to be here, it’s been great fun!
I direct a new lab at MIT, called the Center for Bits and Atoms, I want to tell you about that, and some of the implications for design.

As background, I think that the most useful thing to say about the digital revolution is that it's done. We had a digital revolution, we don't need to keep having it; computers are here, but I'm made out of atoms, and I expect to be for the foreseeable future, and all of this computing has no use, unless it can come out here where I live. We have neatly separated hardware from software, computer science from physical science, but it's right at that boundary where many of the hardest and most compelling, most profound problems lie.

And so, the Center for Bits and Atoms is looking at fundamental mechanisms to connect the digital and the physical world. It includes things like showing the world how to make the largest quantum mechanical computers, by programming the nucleative atoms and molecules to compute implementing quantum factoring. We have shown how to make analogue circuits that produce digital results better than digital circuits; bi-directional interfacing between living, biological systems and computers; and we are starting to programme biological systems to compute the assembly of nano-scale structures. Things right at the boundary where you just cannot separate the hardware from the software.

Right away, that presents one design challenge, which is that any engineered system breaks when you get up to about a billion of anything. The power grid, networks, chips - there is a complexity level beyond which people really just don't know how to design. The good news is, there are very compelling demonstrations of how to make emergent systems; the bad news is, these are all toy problems to date, and they don't scale to really serious, useful engineered systems. Fortunately, some people are thinking deeply about theories of scalable engineering emergence. Unfortunately, they really can't make anything work. And so one of the biggest questions (actually it's a huge question!) is, "Can you be both rigorous and relevant in understanding how, to specify how a system comes to function, without actually saying how it works, but do it with rigour?"

Now, I want to look at that in two particular areas - very directly relevant areas for the practice of design. One is (not as research, just in working), we've had to develop little internet nodes; a lot of the things we make we need to connect to the net, and they are not really computer peripherals, they need to be citizens of the net. So this is a few years of evolution in the internet, each of these is a complete website, just done simpler and simpler as we really understood how to do that. So we thought that’s great, I can make a complete website for a dollar and this little thing, we'll put it everywhere, we'll put it in light bulbs, door knobs, we'll fill the world with that. We thought that was a good idea. Until we thought a little bit more, and once again if those things work in any way like this one, it leads to a fairly distopian vision of the future: like if you wake up in the morning, and you are greeted by "Your house has crashed".

Or even worse, you might not even know that that is what it's telling you. This is how a HP printer driver says it can't find the Net: "Sincway Multiple says we have failed"; this is how the CAD tool I use tells you it can't find its license: "Ordinal 294 couldn't be found". Or my favourite one, this is how Netscape tells you it has a problem with the network: "Unable to launch application, reason = 31".

So, this poses the question: how do you bring connectivity to devices, without imposing this kind of clutter. Now, for inspiration one place I turned was physical architecture; like, I started working with some crazy architects in Barcelona, who like me were inspired by Gaudi's buildings; and if you look at these, it's very clear he didn't subcontract the civil engineering. The physical structure and the logical structure is so intimately integrated, they are all part of one expressive form. It’s really depressing to build beautiful buildings and then toss some computing over the fence when you're all done. So the question is: how to make the intelligence in the building as expressive as the physical space.

So, we did a test installation earlier in Barcelona, that's now re-opening in Madrid later, in fact this week. This was a stage with a structural describe, also it was an amusing space, I think this is the world's largest internet explorer window, it was made big enough so that you could browse reality at one to one scale, as part of the architectural space. And then here is a demo of what we did; and on the one hand I think this is utterly trivial, on the other hand I think it's just a bit profound. There are buttons, these buttons can talk to lights, but the buttons and the lights are fully fledged net citizens, it's about a dollar in parts, but it's a website, it's speaking the whole suit of internet protocols; so the buttons talk to the lights over the net; in addition the button exposes its functionality, so that you can see what it's doing over the net, and you can talk to it.

Now, it takes two or three kilobytes and a dollar’s worth of parts to implement the whole internet protocol stack. If you look at the way the internet is embodied in here, it's an engineering representation of human bureaucracy. There are different groups that do standards at each level, and each one gets a different piece of software with message passing. That’s if you actually do, not what the specs say they do, but what they really do (it's only a tiny fraction of what they say they do, because of all of the subtraction). So what we did was to implement the functionality of what it really does, not what it says it does. Now, that leads to amusing things, like this button that costs a dollar that is a complete internet site, its functionality is associated with the button, not where the button is wired in, so if I put it somewhere else it does the same thing.

The next question is how you programme these, because we don't want to have a light switch and a button, but then force you to go off to a computer somewhere else to configure it. And that's enormously important, because there is a lot of discussion about smart homes, where we are quietly ignoring the killer app, which is the construction industry. In the US, the construction industry is worth a trillion dollars a year, of that 80 per cent of the cost is labour, only 20 per cent is materials; they're desperate for soft-stage, just the notion that associations between buttons and switches can be done on the fly; the economics don't work if you need a sys-op with the button, if you look at the implied costs to connect the computer to a network.

So, the next thing that we did was to take the algorithms and the data structures that live in servers and we chopped them up, so that each little internet site owns its piece of the algorithm and the data. So here is a new button, it doesn't know anything, it has no life experience, I'll take out a sophisticated programming implement, my pair of tweezers, ... now introduce the button to the light; what I'm really doing now is I'm sending a broadcast message over the internet, they are hearing each other's broadcast messages, they are sending IP addresses (for the experts, they are doing a two-faced commit pointer assignment). I’ve now made a button, a mapping of buttons and lights, now I do a one-to-two mapping and connect to the doorbell. On the one hand I just turned on a light with a switch, on the other hand I’m plugging these in. I'm building a network, I'm also building a data structure, and I'm building a computer to process that data structure, and it all scales because it's physical space.

In a world like that, these computers aren't prescriptive; they don't run anything. They are descriptive; they can add value, but they're not actually in charge. So, in fact, in doing that project I was talking to somebody who was part of the internet 2 project, you know gigabits to your refrigerator; and we were talking about this and why we were doing this, and he says, "That's fine, but what's the bit rate?" Then I said, "Well the light bulb doesn't really need video on demand," and I explained all the reasons we're doing it, and he said, "OK, that's fine, but what's the bit rate?"; and I said, "No, no, no! You don't get it. It's not internet 2, it's internet zero."

And the joke stuck, this is now becoming an industry-standard process (ask me more later if you're interested); it's interoperable IP, so it's the internet, but everything else is different: the physical transport, the way you distribute it so that it's computable with the physical world without invoking all the clutter of computing; and this is becoming an industry-standard process to bring the internet out into the physical world, so that you really build networks and computers just by the process of assembly.

Now, back at MIT, one of the people in our lab was inspired by that, and started looking further. He was a former chip designer, and it struck him that the way people are making chips is becoming increasingly dumb; they are making bigger and bigger wafers, with more and more transistors, which are more and more expensive; and it occurred to him that what you should really do is to make the tiniest chip you can get away with, and just make a lot of them. So what he's trying to make is literally a paintable computer, a viscous medium with tiny silicon fragments that makes a pour-out computer, and if it's not fast enough or doesn't store enough, you put another few pounds or paint out another few square inches of computing.

These dots represent the devices. What's happening here is they are communicating locally; each silicon particle is passing a message to its neighbours; from that local message-passing, they are then wiring themselves up, in this case to form a network, and so now they’ve made an ad hoc network, then it'll go back to a quiescent state. You've probably seen ad hoc networks before. What's different about this one though, is that it's really like this piece of wood here; it's a fungible raw material, and then a computational front travels across and leaves it, in this case, as network; it could just as well be a data storage, or a sensor, net; it becomes a programmable raw material. So this is the device modeller, and right now we are working on the silicon for that.

Now, a step after that is making conventional chips and pouring them out; painting them: we have realised we can paint the computer itself. We've developed a range of printing technologies; so this, for example, is an electronic ink you can print, it has the contrast mechanism of ink on paper, but you can change it after you put it down. This is a printed semi-conductor, that lets you print; this is a printed mechanical structure; this is a printed piece of paper that can move another piece of paper, your desk, and clean itself up; it was made out of this.

We began to realise that, if you just take these four boxes and put them in one machine on the desktop, you have a machine that can make any machine. There is a tremendous historical parallel here with mainframes (that were big industrial machines for limited operations), which, when they became PCs, made the digital revolution happen. If we are done with the digital revolution, what comes next is personalising the physical world, and the embodiment of that as personal fabrication. At MIT that takes the form of a class I teach, modestly called how to make almost anything, where students with no technical background, designers, learn to do all this stuff.

The star the first year I taught it was Kelly; this was her project:
[video: "Hi, I'm Kelly, and this is my scream-body. Do you ever find yourself in a situation where you really have to scream, but you can't? Because you're at work, or you're in a classroom, or you're watching your children, or you are in any number of situations where it's just not permitted? Well, scream-body is a portable space for screaming. When the user screams into the scream-body their scream is silenced...but is also recorded, for later release where, when and how the user chooses." (screaming noises)]

Now you laugh, which is great, but think about it: she designed a circuit board, she designed in a micro-controller, she programmed the micro-controller, she made the board herself, she developed PA-resistive squeeze sensors to control it, advanced engineering phones. You know, in a normal company there would be a team of engineers (a normal company wouldn't even do this anyway), and for her it's not product development, it's really kind of post-digital literacy, it's just mastering the means of expression. Design increasingly really can, and does, mean designers are crossing all these levels of description.

And so the thing I want to leave you with is this:
In between the billions of dollars of equipment at MIT that lets us make stuff like this and the basic research, we've realised more recently that, for about $10,000 in a space about this big, you can approximate this class. You can make circuit boards, you can do 3D precision machining, you can do 3D metrology, you can programme micro-controllers, you can do basic analytical measurements. And so what we've started doing is taking little field fabrication labs, which is an approximation of the personal fabricator to be, just taking them out into the world, to see who would do what with them. The amazing response we're getting shows that when technology moves out into the world, it becomes relevant to the rest of the world - everybody but the technologists and the executives and the press-pundits. It’s the little kids and old people in developing countries; they're the ones who say, “Thank Heavens! This is the revolution for us.” Everybody in this room, their needs are all already met, more or less. It's the people with needs not yet met by technology who are desperate for personal fabrication, so we are setting this up in places like rural India and the north of Norway (ask me and I can tell you about all kinds of amazing things they're making).

What that finally led to, earlier this year, was a visit I made to the Himalayas with the Indian general who's in charge of Jammu and Kashmir, the on the Pakistani and Chinese borders. We were working on one of these field prototyping labs, where the interest was in low-cost incremental deployment of mesh wireless networks. The general is in charge of the world's current nuclear battlefield. He got there, his job was border security, and he came to the conclusion that the best way to provide border security is through human security, and the best way to provide human security is through human development, and the best way to provide human development is through information, and the best way to provide information is through the network, therefore Indian army soldiers should bring internet connections to Muslim girls!

So, it's this amazing project where his soldiers are going in and bringing these net connections to little villages. And, in particular, the Muslim girls and Buddhist girls in these breeding-grounds of insurgency, who used to run when outsiders came, now they come running to these places. We were helping them with things like low-cost antennae and embedded controllers, so you can make incremental hubs of networks without any central control of the infrastructure. And what's amazing is the extent to which it flipped a community, from being a breeding-ground of insurgency, to having a tremendous sense of connection, a tremendous sense of belonging, transformed by these low-cost, distributed, locally developed technologies. And so, in a very real sense, I believe the deepest consequence of all of this stuff is not just making it easier to win wars, but preventing the need to fight wars in the first place.

The challenge I give to all of us is who to have that conversation with. I’ve made very strange trips to Washington, where just a couple of days ago I went from the World Bank to the National Science Foundation to a Pentagon meeting with a bunch of army generals about the notion that investing in local design and development of appropriate technology is the best way to stabilize the planet. They all had exactly the same conversation, but it is not clear whose job that is. And so gears are grinding at that interface, but it really is design. It is about taking not just technology to the masses, but technological design to the masses. I think this may be the most potent thing happening on the planet for its future.

updated Monday 31 March 2003
Address: Wibauthuis, Wibautstraat 3 • 1091 GH, Amsterdam
The Netherlands • T +31 20 596 3220 • F +31 20 596 3202
Doors of Perception 2002. We are happy for this text to be copied and distributed, as long as you include this credit: "From Doors of Perception: www.doorsofperception.com".
Want to send us your comments? Email desk@doorsofperception.com