Accelerating Future

(Michael Anissimov, Richard Jones)

Nanotech expert Dr. Richard A.L. Jones contributed “Rupturing The Nanotech Rapture” to the IEET’s Special Report on the Singularity, that topic that all the cool kids, like the Institute of Electrical and Electronics Engineers, are talking about.

According to his bio:

“Dr. Jones is a professor of physics at the University of Sheffield, in England, and senior nanotechnology advisor for the UK government’s physical sciences and engineering funding agency. His book Soft Machines: Nanotechnology and Life (2004) argues that nanotechnology needs to learn as much from biology as from engineering.”

Jones does us transhumanists the favor of engaging with us directly, unlike practically all other critics. I appreciate this. Jones regularly shares his ideas on nanotechnology and other topics at the Soft Machines blog.

In his piece, “Rupturing the Nanotech Rapture”, the only article in the issue to address molecular nanotechnology (MNT), Jones does a wonderful job summarizing the general arguments used by transhumanists and others for why MNT may have a huge impact on humanity’s future, then criticizes them. His essay begins:

How to usher humanity into an era of transhumanist bliss: first, end scarcity. Second, eradicate death. Third, eliminate the bungled mechanisms that introduce imperfections into the human body. The vehicle for accomplishing all three? Molecular nanotechnology—in essence, the reduction of all material things to the status of software.

This is a nice summary of some of our goals, but it’s worth pointing out that progress towards these goals is already being achieved incrementally using technologies other than MNT. If MNT proves unworkable, then progress towards these goals will just continue using other available technologies, as it has for centuries.

If the three ambitions (superabundance, radical life extension, biological reengineering) were achieved, this would help usher humanity into an era of bliss for everyone, not just “transhumanist bliss”. Nearly everyone in the world would appreciate it if scarcity were lessened, or death rates were vastly lowered, or some of the biological mechanisms that cause disease were removed. It’s just that most people are far more pessimistic than transhumanists about the timescales on which these might be achieved.

This highlights a general point regarding transhumanist ambitions — most of what we want (abundance, life extension, body modifications) has a huge potential market, demanded by every human being on Earth. One might argue that much of civilization up to this point has consisted of efforts to bring these objectives closer to reality — automated manufacturing for abundance, hygiene and medicine for life extension, and technological artifacts for extending the capabilities of the human body. Transhumanists argue for the continuation and deliberate acceleration of these trends through select technologies that might be helpful in this regard. One of the most promising such technologies is molecular nanotechnology, although there are many others — numerous strands of research in biotech, infotech, cognitive science, AI, self-replicating systems, robotics, etc.

These “high-risk, high-payoff” research paths must be pursued, instead of sole focus on incremental updates to older and surer research paths. This view was echoed in a recent white paper by the American Academy of Arts and Sciences, which argued for the “encouragement of high-risk, high-reward, potentially transformative research”, an issue “central to the nation’s research effort”. Psychological intertia in the scientific community, of the type described in Lee Smolin’s excellent book The Trouble with Physics, is to blame for timid attitudes about bold new research paths.

Regarding the alleged potential of MNT, Jones writes, “At a stroke, any material or artifact—a Stradivarius or a steak—could be available in abundance.” This is not correct, and no one argues this nowadays. This language (”Stradivarius or a steak”) is a harkening back to the 80s and early 90s, when MNT was introduced as a general concept. At the time, thinkers were just beginning to explore the idea, and noticed that if you could construct anything atom by atom, perhaps one could synthesize a steak from an MNT-based factory just as easily as a violin.

However, subsequent analysis and practical thinking has demonstrated this to be unlikely. The earliest production units based on MNT (sometimes called “nanofactories”) are very unlikely to be capable of manufacturing steaks. Steaks consist of numerous extremely complex organic molecules whose synthesis from constituent atoms would be enormously difficult. A nanoscale assembler programmed only to pick-and-place carbon atoms in a 3D covalent matrix (diamond) might be useless in the construction of a steak, whose numerous varied organic compounds would look like a tangled mess on the nanoscale. More mundane and likely applications of the first nanofactories might be the mass production of simple containers, greenhouse materials, engines, cell phones, and computers. The attribution of belief in a “machine that makes anything” to MNT advocates has sometimes been used as a red herring to protest against the feasibility of MNT research, but only journalists inaccurately summarizing the ideas of scientists have ever said anything along these lines. MNT’s advocates have instead suggested specialized machines for different tasks, which could cooperatively construct a wide range of objects beyond anything that can be made using present-day manufacturing techniques.

Summary of response:

1) All three goals are already being pursued.
2) Transhumanists advocate accelerating research that is already happening.
3) Molecular nanotechnology would be helpful but not necessary to achieve major life extension, abundance, and heal the ailments of the human body.
4) If MNT doesn’t work, we’ll use something else.

Regarding the relationship between those that consider the possibilities of MNT and/or some abrupt technological transition event, Jones writes:

This vision holds wide currency among those anticipating a singularity, in which the creation of hyperintelligent, self-replicating machines triggers runaway technological advancement and economic growth, transforming human beings into cyborgs that are superhuman and maybe even immortal. Some of these futurists are convinced that this renaissance is just a few decades away.

Yes — it’s only natural that those considering the longer-term future of manufacturing technology would also be interested in the future of cognitive technology. It is true that some futurists consider runaway economic growth triggered by artificial general intelligence to be possible or even likely in the coming century or even decades. However, not all of these futurists consider molecular nanotechnology an essential mechanism in this runaway technological growth. Considering that a singularity would supposedly introduce smarter-than-human entities, they could conceivably come up with technologies we haven’t even imagined.

On a separate issue, I doubt that the creation of superintelligent, self-replicating machines would necessarily transform human beings into cyborgs. People might not want to be “transformed into cyborgs” right away, or ever. If “hyperintelligent machines” are under human control, or merely under the control of a democratic system, whether biological or otherwise, then it would seem that they could simultaneously foster increased economic growth while respecting human preferences and their obvious right to retain whatever type of body they want.

Despite the proclamations of Ray Kurzweil and some others, the creation of recursively self-improving, smarter-than-human intelligences — an “Intelligence Explosion” — could be more mundane in appearance than is commonly believed. It all depends on the goals of the intelligence in question (which could conceivably be an agglomeration of millions or billions of human preferences or votes). If these goals include the fulfillment of human happiness, then an abrupt, visible transition could be undesirable. These superintelligences might do a lot of mental processing in large mainframes located deep underground, monitor the world’s environment using quadrillions of inconspicuous nanoscale sensors, or otherwise avoid getting in natural humanity’s way. Our wishes will hopefully play into the decisions, as these superintelligences, if built right from the start, are likely to be deeply integrated with human preferences, and some of them may even be humans themselves.

When will superintelligence be created? Based on calculations of the complexity of the human brain and other factors, philosopher Nick Bostrom argues it will occur “in the first third of the 21st century”. I think that within two or four decades is a reasonable estimate, though many will disagree. What Dr. Jones will find over the coming years, if he hasn’t already, is that intelligent and respectable scientists will continue to propose these timescales and present cogent arguments to back them up. Dismissing such people as nutty transhumanists will eventually seem disingenuous.

Some futurists that predict a major near-future transition justifiably attract ridicule. Ray Kurzweil, the most prominent, has a demonstrated tendency to extrapolate with great certainty, push a spiritual-mystical philosophy alongside predictions, present his own predictions with an air of inevitability or predetermination, and engage in other controversial actions that leads to an “either you love him or you hate him” dynamic. Some mystics, far less scientific and careful than Kurzweil, predict a major apocalypse in the year 2012, based on the turnover of the Mayan calendar, and even point to artificial intelligence as a possible cause of this allegedly imminent transition.

Moravec’s air of imminence has also led to failures in his credibility. John McCarthy and Marvin Minsky’s predictions about the near-future likelihood of human-equivalent AI in the 60s also led to widespread skepticism about artificial intelligence.

However, we’re seeing more and more serious and thoughtful people — many of them un-famous, or unassociated with such predictions in public — treat scenarios of human-level AI within the next few decades with at least worth taking seriously, if not considering them outright likely.

However, trying to predict the arrival of human-level AI and the feasibility or merits of molecular nanotechnology are two completely different issues, even if they are often discussed by the same groups.

Summary of response:

1) Not all who consider AI likely also consider MNT likely, or vice versa.
2) Superintelligence could be possible without everyone “becoming cyborgs”.
3) Near-future predictions of human-level AI are not going away.
4) …even though some futurists have made these predictions controversial.

Jones continues:

But in academia and industry, nanotechnologists are working on a very different set of technologies. Many of these projects will almost certainly prove to be useful, lucrative, or even transformative, but none of them are likely to bring about the transhumanist rapture foreseen by singularitarians. Not in the next century, anyway.

The qualifier here is interesting. Will the transhumanist rapture occur next century, then, if not in this one? If I were a critic of transhumanism, I’d probably say the same thing — arguing primarily on grounds of practical difficulties and time frames rather than outright physical implausibility. Unfortunately, only a few of transhumanism’s critics are so reasonable.

Jones goes on to say:

However, it is a very long way indeed from a top-notch tennis racket to smart nanoscale robots capable of swarming in our bodies like infinitesimal guardian angels, recognizing and fixing damaged cells or DNA, and detecting, chasing, and destroying harmful viruses and bacteria. But the transhumanists underestimate the magnitude of that leap. They look beyond the manipulation of an atom or molecule with a scanning tunneling microscope and see swarms of manipulators that are themselves nanoscale. Under software control, these “nanofactories” would be able to arrange atoms in any pattern consistent with the laws of physics.

It is the argument of some futurists that others overestimate the magnitude of this leap. It’s especially ironic considering that some companies that build “top-notch tennis rackets” — like Zyvex — are investing serious money into the precursors of molecular assemblers, such as high-precision nanomanipulators. At the Center for Responsible Nanotechnology conference in Tucson last year, the CEO and Founder of Zyvex, James Von Ehr, told us he had a research team working towards an nano-assembler.

A working nano-assembler would be the first step towards a self-replicating nano-assembler. If you had self-replicating nano-assemblers, with a bit of work you might be able to get them to replicate into a cooperative array of atomic manipulators. With the right programming, this array could be used to build products like cell-sized robots, consumer electronics, and more. Of course, you’d have to design a cell-sized robot before you could build it.

One working nano-assembler to lead to all that. If you can build one, you can build a quadrillion. The argument here is over whether you could build one at all.

Companies like Zyvex aren’t the only groups working towards nano-assemblers. Earlier this year, Jason Gorman, of the Intelligent Systems Division at the US government’s National Institute of Standards and Technology (NIST), announced that he and his colleagues had built a “proto-prototype nano assembler” which “consists of four Microelectromechanical Systems (MEMS) devices positioned around a centrally located port on a chip into which the starting materials can be placed”, which can then be used to “cooperatively assemble a complex structure on a very small scale”. The press release states, “For the last two decades, those researchers who recognized the potential have taken diminutive steps towards such a nanoassembler.”

As stated before, the notion of a nanofactory that can “arrange atoms in any pattern consistent with the laws of physics” is a long-term vision of the potential of nano-assemblers, not a short-term prediction. The first nanofactories would only be capable of building simple objects with well-specified designs and a limited range of materials. The feedstock material might be something like propane, at first used just to build products out of carbon. You ask for a product that calls for atoms like silicon or iron, and at you’re out of luck. Even still, a self-replicating nanofactory that could quickly build carbon products in an automated fashion would have a tremendous worldwide economic impact.

Cell-sized robots have already been designed — mechanical red blood cells (respirocytes), gene delivery vectors (chromallocytes), mechanical immune cells (microbivores), and more. What obviously remains are to actually construct and test these designs. Even if they don’t work perfectly, the designs can be modified until successful. Obviously, a huge number of biological entities, from molecule-sized to cell-sized, regularly traverse the body and perform a wide variety of essential functions, so we know such a thing is possible in principle. We need to solve challenges like the risk of immune rejection, signaling, sensors, navigation, and so on. Engineers will look to biology for inspiration.

Summary of response:

1) Building a nano-assembler is a huge challenge, but both industry and government agencies are working on it.
2) If you successfully construct a reprogrammable self-replicating nano-assembler, you’re already 90% of the way to a nanofactory.
3) The gap between tennis rackets and swarms of nanobots will not be measured in years or centuries, but more likely decades. Once you have an array of reprogrammable nano-assemblers, you could use them to build cell-sized robots.
4) The first cell-sized medical robots might be highly limited in their capabilities, but will improve over time.

Now, some quicker responses:

Rather than simply copying existing materials, the transhumanists dream of integrating into those materials almost unlimited functionality: state-of-the-art sensing and information processing could be built into the very fabric of our existence, accompanied by motors with astounding power density.

Embedded computing and sensors already exist and are developing independently of any progress towards nano-assemblers. Work towards the miniaturization of motors is also underway. These are not transhumanist dreams, but R&D goals already being funded to the tune of hundreds of millions of dollars or more.

Singularitarians anticipate that Moore’s Law will run on indefinitely

No one ever argued this. Obviously, exponential functions cannot continue forever. In the words of Gordon Moore, “No Exponential is Forever…But We Can Delay ‘Forever’”. Moore’s law does seem to be quite robust. It has been ongoing for almost 50 years and will continue for at least another 10. Although clock rates are not increasing exponentially, performance per dollar has. However, if we don’t change computational substrates soon, improvements could slow down.

Although Moravec and Kurzweil may be overenthusiastic in projecting Moore’s law forward until 2050, they do not speak for all “singularitarians”, many of whom reject such long-term projections.

These minuscule robots, or nanobots, need not be confined to protecting our bodies, either: if they can fix and purify, why not extend and enhance? Neural nanobots could allow a direct interface between our biological wetware and powerful computers with vast databases.

Maybe we could leave our bodies entirely. Only the need to preserve the contents of our memories and consciousness, our mental identities, ties us to them. Perhaps those nanobots will even be able to swim through our brains to read and upload our thoughts and memories, indeed entire personalities, to a powerful computer.

Yes, if nanobots can be fabricated, they will be used for human enhancement, or even to take a crack at mind uploading. We will see.

Jones goes on to describe many of the basic ideas of molecular assemblers, presenting cogent arguments in favor the plausibility of their construction:

It’s a seductive idea, seemingly validated by the workings of the cells of our own bodies. We’re full of sophisticated nanoassemblers: delve into the inner workings of a typical cell and you’ll find molecular motors that convert chemical energy into mechanical energy and membranes with active ion channels that sort molecules—two key tasks needed for basic nanoscale assembly. ATP synthase, for example, is an intricate cluster of proteins constituting a mechanism that makes adenosine triphosphate, the molecule that fuels the contraction of muscle cells and countless other cellular processes. Cell biology also exhibits software-controlled manufacturing, in the form of protein synthesis. The process starts with the ribosome, a remarkable molecular machine that can read information from a strand of messenger RNA and convert the code into a sequence of amino acids. The amino-acid sequence in turn defines the three-dimensional structure of a protein and its function. The ribosome fulfils the functions expected of an artificial assembler—proof that complex nanoassembly is possible.

He then points out:

If biology can produce a sophisticated nanotechnology based on soft materials like proteins and lipids, singularitarian thinking goes, then how much more powerful our synthetic nanotechnology would be if we could use strong, stiff materials, like diamond. And if biology can produce working motors and assemblers using just the random selections of Darwinian evolution, how much more powerful the devices could be if they were rationally designed using all the insights we’ve learned from macroscopic engineering.

Not just according to singularitarian thinking! There are numerous scientists and engineers out there who consider the possibilities of nanorobotics and are intrigued by the way that different properties of biological and nonbiological materials would come into play in this field.

Jones then presents challenges for would-be builders of medical nanobots:

1) “In the domain of the cell, water behaves like thick molasses, not the free-flowing liquid that we are familiar with.”
2) “This is a world dominated by the fluctuations of constant Brownian motion, in which components are ceaselessly bombarded by fast-moving water molecules and flex and stretch randomly.”
3) “The van der Waals force, which attracts molecules to one another, dominates, causing things in close proximity to stick together. Clingiest of all are protein molecules, whose stickiness underlies a number of undesirable phenomena, such as the rejection of medical implants. What’s to protect a nanobot assailed by particles glomming onto its surface and clogging up its gears?”

Of course, all of these challenges were taken into account in the first serious study of the feasibility of nanoscale robotic systems, titled Nanosystems. The conclusion was that these challenges are significant, but surmountable, if we keep them in mind. We’ll need to build nanomachines using nanomechanical principles, not naive reapplications of macroscale engineering principles.

In a comment on Richard’s blog, I responded to some of these concerns, stating:

“We wouldn’t need to have nanoscale assemblers in the human body. We’d use nanofactories to build heat-tolerant microbots, maybe on the size scale of a micrometer or so, to do medical work.

The human body could easily be too hot or chaotic of an environment for the first generation of molecular assemblers. It makes sense to say that some categories of synthetic nanomachine might not operate well in a biological context, but to say that none of them will is somewhat excessive. More research is needed, of course.”

In response to that, Richard said:

“Michael, I’m aware of course that it isn’t proposed to put assemblers inside the body. My point is that all the other functionalities that are envisaged for MNT-based medical nanobots - powering, sensing, information processing - are all based on the same mechanical paradigm that I argue is inappropriate for the warm wet world. As my reply to Brian should make clear, I do think that sub-micron devices with increasing degrees of functionality for use inside the body will come, and indeed are being developed now. It’s fine by me if we call these medical nanobots, as long as we remember that their operating principles are likely to be very different to those envisaged in MNT.”

It seems here that the essence of Richard’s argument is that medical nanobots are possible, as long as they operate according to biological principles rather than mechanical engineering principles. If that is necessary, it makes perfect sense to me. I propose the following path, then:

Scientists continue to work towards nanoassembly. Those that advocate the mechanical-engineering approach should use it, and those that advocate a softer, biological approach should use that. Eventually, progress in one field might leap ahead of the other and everyone can jump on that approach, if they so wish.

After success with nanoassembly, scientists propose different designs for medical nanobots. Some of these designs will be based on a mechanical-engineering approach, others on a softer, biological approach. Build both and test them. Use the one that works better.

That wasn’t so hard, now was it?

Jones goes on to describe his problems with the mechanical-engineering approach to nanobots, but focuses more on nanobots inside the body or in other difficult environments.

Eventually he says:

“Put all these complications together and what they suggest, to me, is that the range of environments in which rigid nanomachines could operate, if they operate at all, would be quite limited. If, for example, such devices can function only at low temperatures and in a vacuum, their impact and economic importance would be virtually nil.”

In response to this, on his blog, I said:

“I disagree with that last sentence. If you could mass produce solar cells, factories, chemical processing plants, etc., in an entirely or almost-entirely automated way, then I think the manufacturing units would pay for themselves even if they required high vacuum and liquid helium for cooling.”

If the machines used to maintain high vacuum and extreme refrigeration could be manufactured for the cost of raw materials, and energy can be obtained in great abundance from nano-manufactured, durable, self-cleaning solar panels, I am skeptical that this would be as substantial of a barrier as it is to similar high-requirement processes today. A nanofactory the size of a real factory would be able to output its own weight in product every day or less, so even if the whole thing had to be cooled and kept in a state of vacuum, the absolute volume of output involved would be huge.

Jones presents his own vision for the future of nanotechnology at the end of the article, concluding with the statement:

We shouldn’t abandon all of the more radical goals of nanotechnology, because they may instead be achieved ultimately by routes quite different from (and longer than) those foreseen by the proponents of molecular nanotechnology. Perhaps we should thank Drexler for alerting us to the general possibilities of nanotechnology, while recognizing that the trajectories of new technologies rarely run smoothly along the paths foreseen by their pioneers.

Here, here. I am in agreement. Even if MNT doesn’t work out, we should try to achieve the same goals via different means, even if it takes a substantial length of time.

I thank Richard for his thoughtful critique of molecular nanotechnology. He adds a valuable voice to the discussion.