Responsible Nanotechnology

Fast Takeoff: Errors, part 2

A comment by Todd Andersen got me thinking about the variety of factors that affect error rate in a molecular manufacturing system.

Todd is right that error checking of finished sub-blocks may be a way to reduce error rates. Whether this works depends on the details of the process. In some processes, it may be relatively easy to build a 100 atom part (of which maybe 98% will be perfect), then test and throw out the 2%, then stick the perfect parts together to make a perfect meta-part.

In other processes, an error will not only destroy the workpiece, but also the tool. So if a 10,000-atom tool is destroyed by each error, then the success rate needs to be 99.99% or better.

In a multi-stage process, each stage may have a different error rate for a different reason, and need different error handling. (Note that, unlike in chemistry, many kinds of errors can be corrected, reducing the eventual error rate.)

Tools that can only be built by molecular manufacturing may reduce the error rate drastically. For example, sorting rotor cascades can give you any purity you need, and atomically perfect sliding seals can exclude all contaminants. In that case, it may be a quick step from just barely good enough, to so good you don't even have to think about it.

A more general argument is that once you have general-purpose molecular manufacturing, you can probably build improved versions of your tools right away. So... although I can't prove that any given molecular manufacturing pathway will have a fast takeoff thanks to error rates crossing a threshold of significance, it does seem pretty likely.

This follows a general trend of argument: the difference between unfeasible and adequate is generally bigger than the difference between adequate and excellent. Feed "excellent" into an exponential growth equation, and you get a fast takeoff.

In future posts in this series, I'll be following this trend of argument in other areas, including range of machinery that can be built, size of design space that's accessible, and fabrication speed.

Chris Phoenix

Fast Takeoff: Errors

The ultimate goal of molecular manufacturing is to build useful products in useful amounts.

In order to build a "useful" amount of almost anything molecule by molecule, the process will have to be fully automated. This means that machines will have to do whatever sensing and actions need to be done in order to form the product.

All of the operations will have to be very reliable. How much is "very"? Well, for comparison, I've been told that the ribosomes (protein-making machinery) in the human body have an error rate of about one per ribosome. Every ribosome is imperfect in a different way, and most of them work well enough despite these flaws.

The fault-tolerance of ribosomes is kind of impressive, but today's machinery is reasonably fault-tolerant as well. If you disconnected a random wire or unscrewed a random fastener in your car, it would probably still work well enough.

We don't know exactly how low the error rate needs to be, in order to make machines that work well enough. But let's pick a number - let's say that we want an error in less than one-millionth of the automated bond-forming operations. This is probably the right number, give or take two orders of magnitude. For a million-atom machine, one hundred errors would likely be too many, while two orders of magnitude in the other direction would mean that 99% of the machines were perfect.

In today's "wet chemistry," a yield of 99% is excellent. An error rate of one in a million is four orders of magnitude beyond that. Clearly, a different approach would be needed to achieve such a radically low error rate.

Recently, Svein Ove objected to my characterizing the difference between 99% and 99.99% as less than 1%, since it's also two orders of magnitude. In my response, I argued that the difference would indeed be huge in a purification context, but it is less significant in the context of finding designs. In this post, talking about error rates, we're getting pretty close to talking about purity - but there are some important differences.

For example, if error rates depend on reaction rates at individually addressed sites, then we can switch from talking about how to prevent errors, and talk instead about how to speed up desired reactions. Mechanically guided chemistry can speed up reaction rates at the desired site by many orders of magnitude, both by increasing the effective concentration (as Drexler has recently discussed) and by lowering energy barriers. Each order of magnitude speedup corresponds directly to an order of magnitude less error. (There are exceptions, such as undesired reactions happening between the desired reactants. But these depend on the chemistry, and can be made extremely low in some chemistries.)

Biology uses all sorts of error detection and correction machinery, which adds significantly to its complexity. Engineered error detection and correction can be somewhat simpler. Since biology depends largely on diffusion and random processes, its processes must depend on the probability that two molecules will bump into each other in the right way to detect whatever needs detecting. In an engineered system with relatively rigid machines, it is possible to sense a condition in a single deterministic operation, correct it in another operation, and know with a very high degree of confidence that these operations have taken place in a timely fashion. Even if the correction operation is only 99% effective, it needs to be repeated only three times to achieve 99.9999% effectiveness. (Again, there are assumptions here, such as that sensing never causes new damage.)

There is a more fundamental reason why error rates will suddenly fall from non-useful to useful ranges (for any given approach). A practical molecular manufacturing system will require machines building machines that build more machines, and so on for many generations to produce enough machines to be useful. If the error rate is slightly too high, this multi-generational approach cannot work. Reduce the error rate by merely half, and suddenly the number of machines in each generation can grow exponentially. So, if the magic number is 99.9999%, then it will probably be a lot harder to get from 99.99% to 99.9998% than from 99.9998% to 99.9999%.

The other question that I'm sure some readers will be asking is whether such low error rates are possible at all. Our experience with chemistry, as well as our experience with mechanical devices, gives us the intuition that such low errors are not possible, but transistors can supply the opposite intuition: the transistors in your computer do the right thing more than 99.999999999999999999% of the time.

You may wonder: what about physical limits? Well, even at room temperature, thermal noise will not break most common organic bonds quickly enough to worry about. The theoretical error rate required by entropy, Heisenberg, and so on, is vanishingly small. Thermodynamics says that every process is imperfect, but it also says that externally supplied energy can be used to improve local order. There is nothing in physics that poses any sort of problem for a mere 99.99999999% correctness rate at room temperature - which is surely better than we need.

Chris Phoenix

Fast Takeoff: Introduction

When and why will molecular manufacturing revolutionize the world? From a technical point of view, the answer has a few subtle but easily understandable aspects. Understanding those points will let us project from current and near-future technology developments, to understand how far we are from a molecular manufacturing breakthrough.

Over the next few weeks, I'll be writing a series of posts exploring the various aspects of fast takeoff. I'll be covering design spaces, product design, factories-building-factories, product performance, the economics of competing technologies, and whatever else seems necessary to understanding the difference between a cool technology and a revolutionary one.

By the time I'm done, CRN's new website design should be live, and I'll convert these posts into new content. (Yay!)

Here's a teaser: A very basic and primitive computer-controlled molecular manufacturing system might have a million atoms (or molecular building blocks). If 99% of those atoms can be placed by the system, then 10,000 atoms must be placed "by hand." That's a very large molecule, or a very large number of scanning probe operations. Probably, a system like this would not be revolutionary - too hard to build, to design, or both. But a system that could handle 99.99% of its atoms would only need 100-atom "inputs" per copy. That is quite feasible by today's standards. So a difference of less than 1% can make the difference between a laboratory demo and a revolutionary manufacturing system.

Chris Phoenix

Confronting the Dangers of Technology

Someone commented on my recent post, The Dangers of Prometheus: "The existence of knowledge doesn't mandate its scope of use. Problems such as global warming come from irresponsible employment of technology, not technology itself."

Sure, there's no mandate that technology has to make things either better or worse. But it's safe to say that irresponsible use is guaranteed. The question is whether we can find and implement enough responsible uses to tip the balance.

The more powerful a technology is, the more important it is to balance or avert - not just the sum total of positive and negative - but each and every of the negatives that's unacceptably catastrophic.

Molecular manufacturing is one of the most powerful technologies I can imagine. It can lead to many outcomes that are unacceptably catastrophic. CRN exists to get people working on how to avoid the catastrophic outcomes of molecular manufacturing.

Of course, molecular manufacturing will also have beneficial outcomes. I'm happy to talk about them as well. Many of those beneficial outcomes will happen by themselves, and others will be promoted by a variety of organizations. But there's no guarantee, and indeed it will often not be the case, that the positives and the negatives neatly match up by themselves. Without deliberate and careful work on avoiding the negatives, molecular manufacturing is easily powerful enough to create unthinkable disaster.

Chris Phoenix

Funny Nano Song

Here's a funny song (and puppet show) about how great nanotechnology is. The song was apparently written for a contest, and is currently in top place (click on the "Top Rated" tab). Other videos on the site are more serious (but probably not as cautionary as CRN is).

Chris Phoenix

Protein Stiffness And Scaling Laws

Eric Drexler has recently been pushing proteins as nanomaterials - even to the point of telling people not to work on diamond yet, which seems strange given that serious researchers are already studying diamond mechanosynthesis in the lab.

Drexler likes protein because it's got a huge design space, it's reasonably strong, and it's stiff enough to build mechanosynthetic systems with. As this post on nanomaterials points out, building blocks bigger than atoms can reduce a major practical challenge of molecular manufacturing: the need for high precision in the face of thermal noise. A little change in size - one order of magnitude or so - can make a big difference.

But there's another issue with stiffness: stiff machines are easier to design, and may work better in several ways. Proteins seem to be about 100 times less stiff than diamond. Stiffness scales linearly with size. So, for machines where stiffness is a limiting factor, the machines might have to be 100 times as big.

As long as the size of a machine is proportional to the size of its building blocks, the time to make a machine increases in proportion to its (linear) size. So a machine 100 times as big might take 100 times as long to build. (If the building blocks shrink in proportion to the machine, the time increases as the cube of the shrinkage. You don't want to go there.)

Molecular manufacturing systems using atom-scale building blocks might take 1000 seconds for a general-purpose fabrication system to process its weight in atoms. If the blocks are ten times as big, then duplication of a nanofactory might take 10,000 seconds, or several hours. But if the blocks are 100 times as big, then duplication might take 100,000 seconds, or about a day. RepRap is already somewhat close to that.

Larger building blocks have additional problems, such as having less design flexibility for molecule-handling (which could be quite important in medical and environmental applications). And proteins likely will need a solvent, further limiting machine speed due to fluid drag.

So it is not obvious to me that protein-based molecular manufacturing, though apparently workable, will be nearly as revolutionary as diamondoid molecular manufacturing. It may even have trouble competing with conventional manufacturing or with RepRap.

Protein-based molecular manufacturing certainly appears to be a strong candidate for a bootstrapping technology, on the way to better machines made of crystalline covalent solids. But in that case, there's not much point in advertising the material properties of protein. And there are numerous other candidate bootstrapping pathways. At this point, it's not too early to champion your favorite, but it is too early to denigrate the others.

Chris Phoenix

U.S. The Enemy of Distributed Manufacturing?

What if the U.S. found it in their national interest to prevent distributed manufacturing from being developed?

I was talking with Russell Brand, CRN Strategist on Societal Response, about a book that Josh Hall had recommended, The Next 100 Years. The author, George Friedman, argues that a lot of geopolitics has been centered around control of ocean shipping lanes - England's partial control of the world's shipping thus led to the Pax Britannica. In the past few decades, the Pacific Ocean has become as important as the Atlantic... and the U.S. is uniquely positioned to control both oceans in the next century.

Russell thought for a few minutes, then pointed out that if U.S. power and influence rests on control of global shipping, a technological advance that drastically reduced the quantity of stuff needing to be shipped would seem to weaken the U.S.'s influence.

I can't easily argue with this logic... though I don't much like the implications.

Chris Phoenix

Fun With Organisms

Jittery particles, with non-linear interactions, moved into interesting and sometimes beautiful patterns, on a two-dimensional surface, by very clever inventive people, who can only control them indirectly.

Here's the video.

No, I'm not talking about atoms this time. In fact, the relationship to molecular manufacturing is tenuous at best... but this is well worth watching for the sheer joy in human creativity.

Chris Phoenix

The Dangers Of Prometheus

Josh Hall likes Prometheus; I'm not so sure.

A few days ago, I posted a comment on changes happening at Foresight, including Josh Hall becoming president. I questioned whether Foresight would continue to take a rosy view of nanotechnology. A partial answer has arrived in the form of a Nanodot post from Josh.

Josh asserts that, in forecasting the effects of a technology, it is easy to see the downsides and hard to see the upsides. Fire, for example, is known to be dangerous in that it can burn people; even Homo habilis knew that. But Homo habilis could not conceive of the positive side: the Apollo moon landings, and the global transportation network powered by combustion. A caveman might have rejected fire on the basis of known dangers, without being able to make a well-informed choice.

I think this is a one-sided view. One of the biggest concerns of the present time is anthropogenic global warming. Guess what put all the carbon dioxide into the atmosphere? Our control of fire. It remains to be seen whether the progress of economics (coal burning and unsustainable agriculture) will push the global climate past a disastrous tipping point before the progress of technology can pull it back. But if our hypothetical cave dwellers understood that one possible consequence of fire was that most of the planet would turn into desert for 100,000 years, as James Lovelock expects... perhaps they would make a rational choice to reject it.

My point is that it's not only the positive consequences of a technology that are hard to forecast and hard to understand. Even technologies as well-established, beneficial, and canonical as fire can have consequences that we are still struggling to deal with or even comprehend. And some of the potential consequences are disastrous almost beyond imagining.

Of course, nothing is all good or all bad. In the next few decades, molecular manufacturing will probably (depending on how it is deployed) give humans the ability to undertake planet-scale engineering. This will make it relatively feasible to moderate the planet's climate and chemistry. It will also make it quite easy to destroy the planet's climate - and I'm not talking about gray goo, but about deliberate applications of high-throughput high-performance manufacturing.

To us today, writing and reading these words on a computer in a comfortable climate-controlled environment, it seems inconceivable that we might want to reject the gift of fire. To our descendents fifty generations from now, whether they are scratching out a living in the few remaining habitable square miles near the Arctic Circle (Lovelock's prediction), or struggling to cope with the nano-built flying land mines that have already killed 99% of the population, the answer may not be so obvious.

I am not arguing for stopping technology. I don't think we can, and I don't think we should try. But let's not pretend that technology is going to make things better. It will give us new problems and opportunities, even as it solves some old problems. The best we can do is to try to guide technology in directions that are less destructive than they might be, and keep looking for new options to solve problems that (to paraphrase Einstein) can't be solved by the systems that created them.

Chris Phoenix

Fun With Electrons

Researchers at Stanford University have managed to make images of the letters "S U" appear in a pattern of electrons. The letters appear to be about 1.5 nm wide, so the features in the letters are smaller than an atom. This is, so to speak, a new low in scanning-probe imaging.

A video of their work is posted on Youtube. It shows several images of the letters, and a brief explanation by one of the researchers. It also includes some interesting "raw" footage of what the letters look like before they are image-processed.

An irregular conductive material has a very intricate pattern of electron density. By placing carbon monoxide molecules on a slab of copper and moving them around with a scanning tunneling microscope, they have been able to create a pattern of "reflected" (their word) electrons that (with the right color palette) contains a pattern recognizable as S. Choose a different sample voltage on the same spot, and you get the letter U.

The letters are fairly crude, and surrounded by visual noise. At first sight, I thought they had taken a semi-random pattern and hunted till they found something that looked like letters with enough image tweaking. But then I calculated: the letters are about 5x7 pixels, counting the dark border around them. That means they got 70 pixels the way they wanted them in the same spot. This pattern could not happen by chance, nor be found by searching random data. They must have actually made the letters by careful placement of carbon monoxide molecules.

I disagree with one thing they said: they claim that this is a very compact way to store information. But the area where the letters appear is influenced by a large number of molecules spread over a large area. That is the information storage. What they have found is a very compact way to display information. They can pack a lot of electron patterns into a very small area, and control the patterns. (It may be that I'm misunderstanding this. The final sentence of their abstract says, "Our electronic quantum encoding scheme involves placing tens of bits of information into a single fermionic state." If they're actually changing the quantum state of the electrons, rather than simply using the electrostatic properties, then they have a point. But you have to change the sample voltage to "develop" the two images, which implies to me that they are not actually modifying the electrons but simply pushing them around.)

I imagine that this work will turn out to be useful for investigating chemical bonding. Chemical bonds are patterns of electrons, and are influenced by nearby patterns. So the ability to create an arbitrary electronic environment, and then manipulate molecules inside it, and watch them react, should provide new theoretical insights or at least fine-tuning of what we understand about how bonds are made.

I'll register one more bit of skepticism. They say this is the first suggestion that information could be stored smaller than one bit per atom. But I read a suggestion at least a decade ago (in EE Times, IIRC) that many bits of information could be stored by exciting the orbitals of a single atom. Of course, that was theoretical, and this is a (rather impressive!) lab demonstration.

Chris Phoenix