Nature demonstrates that productive nanosystems can work cleanly and inexpensively, converting common materials into billions of tons per year of intricate, atomically precise structures. Progress in molecular and nanoscale technologies has laid the groundwork for engineering simple productive nanosystems. These will enable the development of more intricate and complex productive systems, creating a feedback loop that drives accelerating change. At the 2006 Singularity Summit at Stanford, K. Eric Drexler spoke on how advanced productive nanosystems will deliver unprecedented productivity.
The following transcript of K. Eric Drexler’s Singularity Summit at Stanford presentation entitled “Productive Nanosystems: Toward a Super-Exponential Threshold in Physical Technology” has not been approved by the author. Video and audio are available on the Singularity Institute website.
I’m here this morning to talk about manufacturing, a subject that the United States these days regards as rather antiquated. I will be talking about an information-driven revolution in manufacturing. Information driven in developing the foundational technologies, information driven in running those technologies, and producing a physical technology that is itself remarkably much more like software; one that can be thought of in many respects as treating atoms like bits.
The topic here is productive nanosystems, by which I mean small machines that build things with atomic precision under digital control, and how this will lead us toward an exponential threshold in physical technology; a path where we can see already clear mechanisms for acceleration, for succeeding generations moving more rapidly. My affiliation, Nanorex, is a company developing open source, freely available software for computer-aided design of molecules and molecular machine systems. We’ll be GPL‘d, if you track such things.
Outline of topics. First, as I’ve said, this is a technology that will treat atoms like bits. By that I mean, as discrete entities that can be handles rapidly, where you are forming new patterns under digital control and doing so with extreme reliability. The same physical principles that allow one to manipulate bits in microscopic, increasingly nanoscale circuitry, and do so reliably, apply here. The two principles are noise margins, which can push error rates down exponentially as the margins get larger, and error correction. The second major point will be that enabling research is advancing rapidly. The foundations are being laid. The path for this will I think be a super-exponential path. Toward the end of this path we will see a far-reaching transformation in physical technology, but nonetheless there will of course be physical limits.
Looking back to 1990, a gentleman nearby here at IBM Almaden took some hours to make a pattern of 35 atoms with precise digital control. 16 years later we find these structures; they are a series of designs at the top two rows and microscopic images in the bottom two rows. Each of these structures has roughly a million atoms. You make 50 billion of them at a time, and the paper, which was in Nature earlier this year, is unusual in being written in the first-person. There is a single author. I met him. Nice guy. He is able to design and make these things with a turn-around time of about one week. They are DNA structures. DNA can be used as a structural material. Here is another example from last year’s publication; again, a group that can crank out things like this quickly and routinely of DNA being used as a 3D structure.
On the nanoscale there are a lot of building blocks today that people make and are increasingly learning how to put together into larger systems. In the upper left-hand corner is a protein. We have heard it remarked that protein folding is a very difficult thing to simulate, taking gazillions of cycles per second and so on. Protein design, not simulation of the behavior of the molecule, but designing one that physics will then turn into what you want, takes hours on a personal computer. Protein design. Taking nanometer scale objects with all the atoms in the right place and a lot of control over the structure and function is now becoming routine, as is, as I mentioned, DNA engineering. The two fit together.
There are a range of building blocks: organic molecules, nanotubes, various kinds of cluster structures, all of which are atomically precise. They can be made with suitable design to undergo self-assembly in solution to make larger atomically precise structures. That’s a path to making integrated functional nanosystems with bits of metal, ceramic, semiconductor and polymeric materials.
We have an example of a productive nanosystem in the world today, which set some kind of a relatively early target for moving down one of several development paths. This is an object mapped in atomic detail of a productive nanosystem, a ribosome. It takes in digital information from the genetic system and uses it to direct the assembly of small molecular building blocks into proteins. In fact, the light blue parts there are proteins. The sort of spiral-looking tan parts are nucleic acid. If we look at this in comparison to some molecules that I showed earlier being engineered, it’s pretty clear that we are making things of the right kind and the right scale to be able to engineer things on this scale. It’s a reasonable objective. People aren’t there yet, but I think it is where we are going to be going in the not-too-distant future.
Looking a step beyond that, protein is thought of as icky soft things, mostly because people think of meat, instead of one of these. Meat is mostly water, this is inedible, deadly, comes from another part of the beast. It’s a cow horn. It’s a nice piece of plastic. That’s protein, not meat. Nonetheless, it’s only as stiff as typical polymers. This is something a lot stiffer, more regular. It’s easier to imagine building with this. This is a sheet of graphite. What is special about it is that it was made by solution chemistry without anyone being able to put the pieces where they wanted, and fairly recently. The fact that you can make things like that without controlling where the pieces go in a direct digital fashion I think indicates that a little ways down this pathway, we are going to be in a position to design things with materials like this that are intricate and can make structures something like this.
These are examples of the kind of molecular machinery that one should be able to build when one can build things just one step beyond what I showed a moment ago. They are atomically precise structures, they do familiar operations in the mechanical world on a macro scale. If you want to make molecular machines that do things, you have to move things, you have to transmit power. Standard mechanical engineering approaches seem to be a very good way to go. If you think you can do better, then you are more radical. People will come along and say, “This is silly. A biological approach is better.” Fine, they are more radical. They think there is a shorter pathway, better functionality. That’s great.
I’m trying here to give a sense of where things can go. If someone has a better approach, I believe certainly at the detail level there will be vastly better approaches. That will be all to the good with respect to progress on the path. All these were simulated, by the way, using the software package I mentioned earlier. You can see that structures like that are in the same general class as moving parts like this. The atomic detail here is rather fine-grained rather than being the coarse atoms that you saw a moment ago. But rotating mechanical parts with chemically active sites on them can do things like bind molecules, run them through a series of processes where you activate them to make highly reactive molecular fragments, and then put them on building blocks to make larger building blocks, working up to the macro scale.
If you look at where that goes in building larger architectures, small mechanisms of the sort on the left there, moving down to the invisible molecular scale, can put pieces together that are put together to make larger pieces, that are put together to make larger pieces, that finally work up to the macroscale. It turns out this is a balanced production process, in that all of the productivities of the different layers match up in terms of the throughput of mass per unit time. Molecules go in one end and come out the other and they follow some path. If they go at a centimeter per second, then it takes them a hundred seconds to get from one end to the other; to go from molecules to a three-dimensionally precise structure that can be very intricate. That is a little different from macroscopic manufacturing in terms of the time for throughput of simple molecules to product, orders of magnitude better.
I would also note that if you look at productivity, something like this can apparently produce its own mass in material structured to be of similar quality in a time on the order of an hour. If you ask how long does it take a semiconductor fabrication facility to produce its own mass in semiconductor chips, well, I haven’t done that calculation but it’s never going to happen. One thing you don’t see here is swarms of nanobugs making things. This is an industrial style production line but scaled down. In fact, here is an example of a molecular manufacturing system in this conceptual design. You can see it is a very threatening thing. It has four little rubber feet, you can see a grill there for a fan on the right, plugged into the wall, touch screen and raw materials coming from little bottles. What we are looking at here is an appliance, but one that could make remarkable things from simple raw materials, and do so quickly and inexpensively.
That is the kind of picture of molecular manufacturing today. It is the one that has been around since 1992. When you hear about “nanobugs,” that’s an idea that is fourteen years out of date. Please sync it. Self-replicating nanobots are not part of anybody’s picture.
So, backward chaining, how do you get there? If you wanted to build something like this, you could build one if you had one half the size. You could build one of those if you had one half the size, etc. That is until you get to a minimal size. What’s that? Something like one tenth to one micron in scale. That would involve having things like this in that kind of volume to make your small mechanism that could then make the larger ones. This is something we have in our sights in the research community. A stretch by anyone’s standards, but a reasonable project to aim at. We are already making things that are larger and comparable in complexity, in terms of number of atoms.
Why do you expect acceleration? Well, cycle times will shrink. It takes weeks to make these structures, but they are fairly simple. It is going to take longer to design new instances of things like ribosomes, but ones that extend capability. Once you have those, design will be relatively simple and direct. When you have ones that are good enough to make not just polymers but things like this, you are starting to move into a mechanical engineering world where things are very designable, much more predictable, like these. One can do the design before having the tools, so that when you have the tools you are ready to roll and do a fast cycle of tests and redesign. So far, it’s all been stealth. It is microscale, making relatively small changes in the world, then scale up can be done rapidly, and you have a dramatically superexponential process where this microscale technology erupts into the macro-world and changes manufacturing on a large scale.
What can this buy? Moore’s law is going up to the billion transistor range. A technology base like this can give you a billion CPUs, not transistors, and put them in an air-cooled laptop computer. Being able to make small things on this scale has applications in medicine. You saw this image earlier. It functions much like a white blood cell, designed by Robert Freitas. There in front of it you can see a flu virus to scale, perhaps an H5N1 Virus. To deal with that, you would program the thing to say “Here is what it looks like. Grab those, chew them up, and dispose of them.” Much better than trying to develop a vaccine that arrives too late.
I said that you could make things on a macroscopic scale. This kind of fabrication facility scales up, and you can make pieces that get put together. You could make materials that are fifty times the strength to weight ratio of what the space shuttle was made out of. It makes it much easier to make things that go up high and fast. It provides a technology base that can clearly open the space frontier in a decisive way. We have looked like we are stuck on this planet so far, because we are still in the hot air balloon stage of space flight.
However, this leads to limits. Trends don’t tell you what the limits are. They tell you what you are doing within limits. Projecting forward, you run into a wall eventually in many different directions: the second law of thermodynamics, quantum measurement limits, speed of light limits. At a key conceptual point of thinking about this kind of problem is that technology, as it progresses, is always expanding capabilities. Science, as Mark Miller pointed out to me some 25 years ago, is different. Its discoveries can revise limits up or down. It’s about finding what is out there, not about achieving more. We can see here various limits that were discovered in the last 200 years that appear to be solid.
Physics sets limits. I say “physics,” I don’t say “our present understanding of physics.” Whatever the limits are, they are there. We may or may not understand them well enough to understand their consequences in a particular case. Nonetheless, physical law, and I think it is worth guessing that we understand it quite well with respect to the engineering domain of light and matter, provides a stable framework for thinking about the long-term future.
So the picture looks like this: super-exponential hitting a wall. That is a quality of technology in areas of complexity. The picture can look much more complex with a slower growth. Somewhere in there I think you get machine intelligence, but this story does not rely on it. Rather, downstream it supports it by providing a better substrate for computation.
I have reviewed the points: atoms like bits, rapidly advancing research, shorter cycle times and super-exponential development, a deep and pervasive transformation of technology, but physical limits. A key point with respect to development right now is that while the technology base is advancing rapidly, thus far it has mostly been an unfocused effort. People are making more and more nanoscale things, finding out how to put them together, but only beginning to get a focus on productive nanosystems as an objective. In the United States, one sign of progress is that bogus criticism has fallen out of fashion. It used to be that we had criticism at the intellectual level of “You can’t do productive nanosystems because you can’t turn lead into gold.” This is about rearranging atoms, not making new ones. That level of absurdity of criticism has fallen out of fashion.
In the U.S. we now have a technology roadmap project underway, co-sponsored by Foresight Institute and the Battelle Memorial Institute. It is an international technology roadmap for the development of productive nanosystems. Battelle manages five of the U.S. national labs. It is a heavyweight organization. Outside the U.S. I think there is less confusion in many respects. One indication is that journalists ask better informed questions. Another one is that R&D agencies are seeking planning advice on how they can take their expanding nanotechnology research and move it around a little bit to focus on these systems goals of productive nanosystems.
Quite striking, I have seen some speeches recently from national leadership that referenced the concepts. Nanosystems: Molecular Machinery in Manufacturing and Computation is a technical, mathematical work in applied physics. That full title in quotes has been on the lips of the president of India in many of his speeches recently and this is a man who speaks to quite technical audiences about these things. He is articulate, lyrical, and talks about this as part of the future of India, which is increasingly a technology powerhouse.
In closing, I would just like to say that this is a vision of technology development, a transformative super-exponential one, that is not something that is shared only in this auditorium. It is not something that is a United States phenomenon. It is a world phenomenon and it will be driving accelerating change. It will change everything from the bottom up. Thank you.