Teleconference Transcription – Dr. Eric Drexler, Chairman, Foresight Institute.

This evidence was taken by teleconference at a meeting of the working group on October 29th 2003. For reasons of clarity, the evidence session was recorded and transcribed, and is presented here in its entirety. It has been approved by Dr. Drexler.

Participants:
AD Ann Dowling - Working Group Chair
ED Eric Drexler
AS Anthony Seaton
NG Nicole Grobert
RW Roger Whatmore
RA Raymond Oliver
JR John Ryan
JP John Pethica
MW Mark Welland – Discussion Chair
AWD Andrew Dunn
RP Richard Ploszek
ST Saul Tendler
NP Nick Pidgeon

AD Thank you Dr. Drexler for sending us this useful statement that has been circulated to all members of the Working Group here. I plan now to hand over to Mark Welland who is going to lead the discussions and questions

MW I thought we could start, we have a number of these 6 items that we have been asked to specifically address, so perhaps we could start by the first one which is 'What is Meant by Nanotechnology?' and having read through your submission, it seems that you have a sort of perhaps a split between current nanotechnology and what you ultimately call zeta-technology and I wonder if you could explain whether that's right or not, and perhaps explain in a bit more detail how you see that split?

ED Yes, I think that understanding the state of public discussion of what's termed nanotechnology really begins with understanding that the word has two distinct but slightly overlapping interrelated meanings. One is the meaning that it had when it was first introduced to the broad public, and the meaning that still, I think, animates much of the discussion that you see in the press and elsewhere, which is the label for the Feynman vision of nano-machines building complex systems with atomic precision at low cost. A future technology that is an outgrowth of a long series of developments in laboratory techniques and then their application in a systems engineering effort that has not yet been undertaken.

The other meaning, which is the established meaning in the scientific and technical community, particularly when it comes time to seek grants, is essentially any technology that has interesting nano-scale features, and the latter technologies are built in a technology base that's leading towards the Feynman vision, but is really a distinct and much broader field of Inquiry.

MW So you make a clear distinction then between this second technology, the sort of current nano-scale view, and this Feynman vision?

ED yes, they are distinct both in the time-frame in which they become feasible, one is happening right now and the other is years in the future.

MW Would you like to guess how many years in the future?

ED That's a very difficult thing to do. We understand the physical principles involved in these advanced systems very well, they are usually modelled, the fundamental components are usually modelled by computational means, the analysis has been deliberately constructed to rest solidly on established scientific knowledge. It's really an exercise in understanding the implications of what we already know. But trying to understand how long it will take people to solve a set of difficult, practical, implementation questions coming out of present laboratory techniques, is first subject to a lot of questions that are harder to model and second subject to questions of human will and decision to proceed on focussed systems engineering research efforts. We don't know what's going on around the world in various laboratories. We don't know the state of intentions of all of the governments. We don't know what will be happening as decision-makers come to understand the strategic consequences of the technology that will emerge from this effort. And so that leads to great uncertainty. On the one had if no one were ever to attempt to achieve the goal it wouldn't happen spontaneously. On the other hand, if there were focussed effort with resources commensurate to the expected benefits and the strategic risk reduction, I think many analysts feel that the time frame would be measured in a modest number of years, to get to a certain threshold, certainly not the full range of applications.

JP This is John Pethica here. I would like to ask for a little clarification taking that slightly further down the road. You said, you have effectively phrased what you call the Feynman vision as 'molecular manufacturing' if I have got that right there from your write-up to us?

ED Essentially yes.

JP Yes, I wonder if you could help us a little bit to understand what the distinction between that and what is commonly known as chemistry is? In what way is what you are describing not ordinary chemistry?

ED There is no distinction in terms of the fundamental physical principles involved, both involve intermolecular forces and the playing out of their consequences in physical systems in which molecules are changing their arrangement and patterns and bonding.

However, chemistry today is based primarily on disordered systems, typically in the solution phase, in which molecules collide as a result of thermal motion, and in which reactions occur because of the different propensities of molecules for reacting in the presence of random collisions occurring in all possible collisions and orientations.

In molecular manufacturing the anticipated systems, the class of systems that are analysed and proposed to be built, involve guiding the trajectories of molecules so that, for example, unwanted chemical reactions are prevented by simply preventing the molecular collisions that could cause them and desired reactions are facilitated by bringing the reactive molecules into proximity in a selected position and orientation. So this is a radically different paradigm of how to organise the motion of molecules to bring about chemistry.

JP Can I just amplify that then to be clear, because that's a very good way you've put it, is one of the words you used here to make this distinction is 'mechanosynthesis' are we to understand that mechanosynthesis essentially means a mechanical manipulation of placement of every molecule and prevention of that process for the ones that you don't want being placed? Would that be a reasonable summary of mechanosynthesis?

ED That's the essential idea yes.

JR Eric, John Ryan here, could I ask you a question about a statement you make in your submission which is that the actual core concepts of molecular manufacturing have withstood scientific scrutiny for more than a decade? What sort of scrutiny are you referring to? Are you referring to peer reviewed publications in the mainstream literature, are you referring to peer reviewed grant applications to the mainstream funding agencies? What is the scrutiny that you are referring to?

ED Well, I think it's clear from the rather heated discussion that we have seen regarding the feasibility of these technologies and the desire of many scientists to distance their work from this vision, by denying the feasibility of the vision. I think it is true that there has been a very vigorous effort to discredit in any way that people can, and I believe that if you review what is said about the subject you will find that no one has settled on a standard flaw to point to, no one has been able to say "this doesn’t work because…". And quite apart from the question of, you know, specific literatures and specific formal processes, the fact that in this internet era we have scientists putting their reputation on the line saying that this doesn’t work and unable to state what sounds like a scientific reason, I believe that's really strong evidence of having withstood scientific criticism.

JR But could I just ask you again the point I am trying to elucidate here is what is the scientific scrutiny? Is it on the basis of this debate you refer to in the public domain? Or is it scrutiny following the usual peer review process?

ED Well I would say that its strictly of peer review papers and a variety of journals including the Journal of Nanotechnology, the papers coming out of their conference series run by the Foresight Institute since 1989, results of which have appeared in special issues of that journal. Scrutiny of a technical book on the subject that I published back in 1992, nanosystems from John Wyley and Sons. So there has been a very large target for people to criticise, a large set of specific statements and proposals, discussion of physical processes, thermal noise, quantum mechanical effects, chemical processes, molecular machinery and so on, and a large set of people who have been attempting to critique these ideas and as I say the result is a visible absence of successful criticisms.

MW Eric in reference in linking these different scenarios that you say scientists are trying to say are not possible, can we talk about something which I know you are aware of has been in the press in the UK and that's 'grey goo'.

ED Yes

MW And we would like to understand where did the term 'grey goo' come from? And what is your take on it?

ED Many years ago, when I began thinking about the potential for artificial molecular machine systems, at the time I was a graduate student at MIT, back in the late 1970's, I was focussing on a biomimetic perspective on these molecular machine systems, and on the question of how one could build from initial small-scale machinery to larger systems. In that context it was natural to think about self-replicating machines. Biology demonstrates that molecular machine systems can be made to self-replicate if they are properly constructed and it provided one way of achieving the scaling from an initial small set of machines to larger machines. This also points to the usability of the class of systems that had some of the properties of living systems but not all of their constraints, in terms of their natural ecological checks, for example. And that, in turn, suggests that it is physically plausible to build structures that self-replicate unchecked in the natural environment, and would be enormously destructive.

On the other hand, the motivations for doing that are very slight, enormous damage but no benefit. If one wishes to consider negative applications of the technology, weapon systems are a far greater concern because people have an actual motivation to build them, and therefore I am inclined to say that what has been labelled 'grey goo' is a legitimate concern. It is physically possible, attempts to deny the physical possibility will inevitably fail and I think discredit those who attempt to deny that physical possibility. But when people say "this isn't what we should be worrying about" I think they are right. I believe it's very much the wrong issue to focus on for a variety of practical and sensible reasons.

JP This is John Pethica again. Can I just take that a little bit further then that question about grey goo, because it is of such publicity-related things over here? You have described it, not specifically what it is or what it might be, but rather as a potential undesirable end product. Are you leaving it merely as that, or does 'grey goo' actually mean something? Would you like to say what it actually is a little bit more?

ED What the term denotes is a problem, and emergent problem that would arise if someone undertook the difficult engineering effort of building a self-replicating machine system that was able to work using natural sources of raw materials and energy and built it in such a way that it could use a wide range of substances, and such a way that it had no limitations on its replication and then proceeded to turn something like that loose in the natural environment. The suggestion is that such devices could become extremely numerous and convert a large amount of material into copies of themselves, causing great ecological damage. So that is the concept.

JP So essentially it is an undesirable outcome, rather than something which is say related to entropy maximisation say?

ED Precisely. It's an undesirable outcome of the construction of a particular class of obnoxious system. Not the result of an accident but the result, the hypothetical result of a kind of deliberate abuse of an advanced technology that we are quite far off of having today, and it is quite far off from anything that is associated with laboratory nanotechnology today. This is why it's highly inappropriate for concerns about grey goo, which as I have said, are not the right concerns to focus on in any context that I know of, it is highly inappropriate for those concerns to be attached to present day distantly related nanotechnology research.

JR John Ryan here again. Could I just take this forward just a little bit further. The self-replicating machines that we have just been discussing, presumably they would evolve into lager scale systems, into microscopic, indeed into macroscopic systems, just as the biological world we inhabit has evolved? Presumably that's part of the model?

ED That's a natural expectation if one starts with a biological perspective. However, it turns out to be the case that self-replication and evolution can be separated. In fact, if one examines biological systems, they have many structures that are very different from those one finds in the mechanical world, or the world of software, which seem to be closely related to the historical requirements that these biological systems be evolvable. And in fact they evolved to be evolvable. I think a natural consequence of the nature of machine systems is for them to be less evolvable. One could get into a series of technical discussions of what those qualitative differences in organisation are. More concretely it is possible to build the equivalent of a genetic system that works through an encoding and decoding system, that effectively ensures that the only forms of mutation are equivalent to randomisation of the genome. And that precludes the incremental change that is required for evolution to work. So the short answer is that if one were to build self-replicating systems of this sort, which is not necessary with the implementation of molecular manufacturing, and I think it's something to be avoided unless one very thoroughly understood what one was doing and had some strong reason to do so, which at present is not clear. If one were to go down that path, it would be possible to build self-replicating machines that would not evolve.

JR Could I just clarify a point here? You've talked about mechanical nanosystems as if they were quite distinct from the naturally occurring biological nanosystems that we know about. The mechanical nanosystems are composed of atoms and would presumably have properties not unlike the biological nanosystems that we know about. Is that true?

ED There are different levels of organisation to consider here. At the level of the properties of atoms and bonds and fundamental physics, there's of course no difference. At the level of the organisation of those parts to achieve results the differences can be enormous. Certainly there is potentially a continuum between biological systems and other systems. But, moving towards the non-biological extreme one can describe systems that are as unlike, say, a living cell, and their pattern of organisation, as a modern factory is unlike a cell. If the factory were like a cell, one would see machines floating around inside the building, bumping into each other at random, parts moving around at random and sticking to machines, adhering to machines. Products from one machine diffusing to another machine. This would be a very different physical system in its organisation than what you actually see in a factory which consists of parts bolted down in fixed relationships to each other, with systems of conveyance carrying parts, and distinct organised channels for conveying power and information. A much more rigid style of organisation that we see in the engineered and mechanical world of the macro scale, that style of organisation can be carried out on a nano scale as well.

RO Eric, this is Ray over here. I'm a simple Scottish Chemical Engineer, so can you give me an example of how and where you have tested some of your ideas quantitatively?

ED The systems that have been explored through computational modelling and systems engineering design are by their nature and by the choice of the domain that we are examining are systems that cannot be built with present-day laboratory abilities. It's an effort to understand what will become possible as certain abilities mature and as certainly unit operations are combined to build complex systems. And therefore, if one focuses on different aspects of the systems engineering, on designed systems being researched in the literature, the questions resolve into those of physical chemical reactions, specific sorts of mechanical systems, and one is lead typically to examine computational models. So in many cases the arguments rest on observing that at the timescale of the unit operations involved, the specific steps are those that we already see happening in chemistry and biology.

RO Do you have a team working on this Eric.

ED I missed a word there?

MW Do you have a team working on this activity?

ED I have been working in this area since the late 1970's initially as just an individual studying questions, giving talks, getting feedback from a broader community. During the 1990's there began to grow a set of people who were pursuing research in this area, most notably Dr. Ralph Merkle who was with Zyvex, a nanotechnology company, and is now a professor at Georgia Tech but we have a handful of people who are pursuing computational studies and design work, there is currently a series of volumes coming out on nanomedicine by Robert Freitas. But, since the rise of nanotechnology in the new sense, the broad current-day laboratory research sense, there has been a striking polarisation of the scientific community, a striking degree of what I would call a breakdown in scientific communication. We have seen conversations carried out essentially through the press, rather than through, for example, the kind of exchange that we are engaged in right now, where we have scientific questions being asked and answered. A consequence of that was in an environment in which no one who was working on these ideas felt that it was possible to get funding, and as a consequence of that I cannot presently identify a substantial team of people in the Western world who is actively engaged in advancing these ideas. Which I might add, placed us in a position where we have an increasing potential for strategic surprise, should other parties in the world decide to….

RO I have to say I admire your courage for ploughing on with this, given you know, sort of the resources and the constraints on resources. Do you think you are close to making self-replicating machinery?

ED Am I close to making self-replicating machinery?

RO Well, do your calculations say that you are getting close, and therefore, what kind of ….how do you go about that as a molecular manufacturing operation?

ED Well, in terms of self-replicating machinery the place to look right now is the re-engineering of living cells where people are moving towards being able to entirely replace the genome of a small bacterium with a small artificial genome, so I believe the biologically-based and biomimetic approaches are where we are going to see I think an active progress in self-replicating machines and a blurring of the boundaries between nanotechnology and biotechnology.

RO Okay thank you Eric.

ED In terms of getting such results from the kind of systems engineering effort that I am describing that's many years in the future, and will not be a direct consequence of any work that I am doing.

RO Right, thank you.

MW Eric, Mark Welland again, can we….time is pressing on and I want to try and look at the second item which you have already….which is the current state of scientific knowledge, and you've already touched on this in a number of ways I think, and in terms of your vision for the molecular manufacturing and the fact perhaps that it's not as directly, or explicitly funded as it could be, would you like to say what the current state of scientific knowledge is in respect of that technology? And where perhaps you would see either where more scientific emphasis is required, if you could wave a magic wand?

ED Well I think I would say that a sharp distinction should be drawn between understanding the feasibility of a certain broad class of future systems, and understanding how to implement the enabling technologies for those systems. Enabling technologies are characteristically more complex and harder to model than at least some of the future systems. In understanding our knowledge of those future systems there is another distinction and that is between designs that are performed today for the purpose of understanding what is possible, and designs that will be performed in the future to actually implement systems. Those future designs will surely take advantage of detailed scientific knowledge that will emerge along the development path that will emerge in part from the very active ongoing research in diverse nanotechnologies today.

MW Are you saying that the designs that you are currently making in terms of the molecular modelling may well change in light of actual scientific developments?

ED Certainly. I would be very surprised if any but an extremely small fraction of them prove to be in any sense optimal for example, and what we have here is what I call 'exploratory engineering' where the goal is to set lower bounds on future capabilities by designing simple robust systems that rest directly on what we already know. So by construction the analysis of future molecular manufacturing systems is based on present scientific knowledge. It would be an error according to the methodology that researchers in this field have adopted (a small number of researchers), to assume scientific knowledge that we don't already have, and anyone who was able to point to assumptions that were not supported by present scientific knowledge would have identified an error and would have legitimate cause for asking for a revision of conclusions of the research.

MW I assume that you are not saying that the fundamental basis upon which your molecular models are developed on, in terms of an atomistic view is flawed, it's rather how one might actually assemble them and what the production process might infer in terms of the structure?

ED Well again there is a distinction to be made between advanced systems building further advanced systems where one can assume mechanosynthesis in the production of the systems that are later used to perform mechanosynthesis. This is the implementation the first time, where one would have to start with for example protein engineering, or for example scanning probe manipulation of chemically synthesised building blocks, which would inevitably be a far more difficult and messy process because one is beginning with disordered matter and trying to produce the state of order for the first time. And that's why the early stages and the implementation pathway are so much more difficult to understand and model, than are, at least, the simpler systems that we understand to be possible in the future.

MW So you mentioned two actual processes there, the scan probe microscope and then a protein-based assembly. Do you use those two techniques in terms of your current models? Is that the current scientific knowledge that you feed into your models in terms of how you might actually manufacture these?

ED I would add a third which is organic synthesis of self-assembling building blocks. But, referring to those three as implementation paths, there has been a certain amount of attention, this is a distinct area from the studies of long-term feasible systems, a certain amount of attention, it is a question of how do we get there, and that very much rests on the technologies that you mention there, yes.

MW Right because as you said, your models currently would be modified very significantly in light of future technology, so your current models….I'm just trying to establish exactly how you take into account those three technologies in terms of the models that you currently develop. Are any of the current models that you have able to be assembled by any one or combination of those three techniques in terms of current technology?

ED The systems that are easy to model assume the use of mechanosynthesis to build systems that are achievable with mechanosynthesis. The picture of a simple lower bound set of capabilities for future molecular manufacturing is by construction, by the assumption of the framework of the analysis is in a stage where one has moved beyond the constraints of, for example protein structure when working with the larger class of structures. One has moved beyond the constraints of existing scanning probe systems, in terms of their speed of motion, precision and so on because one is instead working with molecular machine systems. So, the complexities and uncertainty of these pathway or implementation technologies are very strongly de-coupled from this understanding of what future systems will be able to do, so it very heavily affects the question of how competing diverse laboratory groups could move down the implementation path that would move us to the better organized world, this world of systems and capabilities for organising molecular motion. I sometimes think of this as facing a mountain of technology that we are climbing, we are struggling up a slope using existing laboratory facilities, while meanwhile using computational modelling to get sort of an aerial-view of the landscape on the other side of the mountain, so that we have some understanding of where we are going in the long-term, even though that understanding doesn’t directly help one to climb the mountain and deal with all of the cliffs and brambles that one encounters in laboratory work.

AD It's Ann Dowling here, could I ask you perhaps to explain to a non-expert how you go about the computational modelling? Is it on an individual molecular machine basis, or is it aggregate modelling? What kind of modelling do you do?

ED One level involves looking at systems of a few dozen atoms or fewer in the immediate locus of the chemical reaction and for that one uses the standard tool that chemists have developed for modelling chemical reactions. The gold standard would be the high order ab initio quantum chemical methods. In dealing with molecular machines one uses a different set of physical approximations in which one models the forces between molecules using empirically described force fields, these are known as molecular mechanics models, and those can be used to describe systems from a few atoms up through millions, and that is large enough to describe molecular machines with moderate numbers of moving parts describe the geometry of the parts, the forces the deformations the molecular dynamics the patterns of motion.

Looking at a higher level, one transitions to using the kinds of models that are used by mechanical engineers to design systems that contain many components. The mechanical engineers typically don’t worry about where the individual atoms are in their macroscopic machines, and likewise one may not worry about where the individual atoms are in extended molecular machine systems, given that the components have been modelled adequately.

AD Do these calculations enable you to look at the energy input required to produce structure from disorder?

ED Well, in terms of the energy input and questions of maintaining and extending organised structure, one issue is the stability of the structures against thermal damage and radiation damage, that stability is quite high for sub-micron scale volumes of material giving long lifetimes for structures just sitting statically in the environment without active repair. Another question is the energy input required to drive the motions of molecular machine systems. That's actually relatively difficult to model using computational methods in a direct fashion because the losses are (on a short time scale) a very small fraction of the characteristic thermal motion energy. So there one turns to analytical methods of looking for non-isothermal non-adiabatic properties in molecular motion, looking at questions of phonon scattering at interfaces, doing analytical modelling of the interaction of inhomogeneities at the interface caused by rows of atoms in one surface moving over rows of atoms in other surface. So really one ends up using quite a diverse set of models at different levels. It's hard to make a generalisation.

AD Can I just ask you about how you investigate stability? Because it's something that I do for computational based models, and simply letting things evolve in a time domain is usually a very poor measure of physical stability. It tells you something about your code more than anything. How do you go about investigating about whether these structures are stable or not?

ED In a wide range of structures where questions of stability are indeed very hard to answer, for a computation model with any given degree of accuracy I think it's a sound generalisation to say that one can always specify a physical system in which the questions of stability, of whether its behaviour will evolve down one pathway or down another pathway is unpredictable in terms of the model because the results depend on differentials in energy that are within the uncertainty of the model. The goal for engineering work is to choose systems which are deliberately designed to be as stable and unambiguous in their outcomes as possible. In trying to understand future capabilities, this leads to an emphasis on strong covalently bonded solids, which locally resemble ceramics or diamonds. And looking at surfaces that, except in very local and controlled circumstances, are saturated in a covalent bonding sense, and therefore are stable in the way that diamonds and certain ceramic and semi-conductor surfaces are stable.

RW Roger Whatmore here Dr. Drexler, have you considered the issues of information delivery and control to your molecular manufacturing systems - how that would be done?

ED Well there are a variety of answers to those issues depending on the class of system and the state of development of the technology that one is considering. For advanced systems where one has organised macroscopic amounts of material with atomic precision with scattered defects, one ends up looking at the future of computational technology, either mechanical or electronic digital systems and examining what can be done with digital signals on fine wires or analogous signals in moving mechanical rods. For systems on the implementation pathway, one tends to think in terms of providing information in the form of particular changes in the chemical environment, a much slower and lower bandwidth process. Or particular changes in the other physical parameters of the environment such as pressure, which, in the case of ultrasonic signals can extend up to the megahertz range. There is a wide range of answers to those questions, which again depend on the particular engineering context.

ST Saul Tendler, Dr. Drexler, when you are designing systems using biomolecular components, what properties of proteins or nucleic acids do you select to allow the systems to assemble?

ED The difficulties of modelling systems of proteins and nucleic acids in solution are much greater than those of modelling small rigid covalent solids in the vacuum, therefore the detailed modelling which has gone into components for future products of molecular manufacturing has not yet been done for these, as I say, more difficult pathway technologies where one is constrained by the structures of available bio and other polymers and so on. But in trying to understand what our knowledge of protein structure and protein engineering abilities tells us about building blocks that could be developed for the implementation of these intermediate generations of molecular machine systems, systems on the pathway to the advanced systems, one typically would see classic engineering virtues like high stability, and there is an intensive body of work in protein engineering which actually if you searched the citation tree of papers on computational protein engineering the citation tree seems to be rooted in my 1981 paper which introduced the concept as a pathway to what we have been discussing. Any of that, the work that has proceeded in that area, which occasionally cites the utility of it for the longer term goal looks at differentials in protein stability associated with differentials in structure between thermally stable proteins and less stable proteins and so on, and tries to maximise the density of the features that are known to improve stability. And there is a body of mixed computational and laboratory work in that area that has demonstrated some effective design and predictive abilities.

Self-assembly is then a matter of designing complimentary interfaces that again have these stabilising features, whilst simultaneously trying to design the interfaces so that they don’t adhere to other unwanted surfaces which leads to a rather complex world of constraints which I think people will be some time in sorting out in development of practical systems.

ST Why would you design interfaces when you can take advantage of natural protein recognition templates?

ED I think that taking advantage of existing protein building blocks and existing interfaces for the assembly is a sound design strategy, I expect that it will be exploited. On the other hand we can look at these systems, generalise from them to their principles and make other structures that embody the same principles which result in similar behaviours. The only advantage there is that one can expand the set of building blocks and make parts that fit into novel structures not seen in biology.

NP Dr. Drexler, it's Nick Pigeon here, you've mentioned already that grey goo might not be something to worry about, or maybe not worry about in the immediate term, but you have said that some of these proposals in the longer term might be quite transformative on society. And I wonder, obviously you've thought a lot about this, if you could share with us some of your visions in terms of some of the social and ethical questions that we should be raising in our report, from your perspective to this.

ED My interest has focused on long term consequences both including accidents and dealing with issues of military application and strategic stability and with broader issues of potential social disruption, it's that set of concerns that has motivated me to work in this area for so many years. I have had a sense of responsibility to try to communicate what I have understood of future possibilities, to (I hope) increase our foresight and ability to make wise choices. Broadly speaking I think that the problems resolve into three sets. One is a set of problems associated with accidents and there it's useful to distinguish unprecedented runaway accidents involving self-replicating machines, the classic grey goo fear, from everything else, and to observe that such accidents need not happen because one need not build systems that remotely resemble the systems that would cause the problem. There is simply no reason to do that any more than there is a reason for people to accidentally have nuclear explosions happening in their house, people don't assemble anything resembling a critical mass of plutonium or uranium in their house, you just don't do that because there is no reason to it, and it's difficult and expensive.

NP I could put to you there may be other forms of accident actually, funnily enough, while you are worrying about the grey goo, then maybe other more mundane things will creep up on us?

ED Yes, well, I think that the classic range of accidents and unexpected consequences is enormous and we are certainly not going to see those go away. On the other hand….

NP There's uncertainty over that I guess?

ED Yes, I think that history pretty well guarantees that people will make a muddle of things in various ways. It is worth noting that what molecular manufacturing is about fundamentally is getting more precise control over the structure of matter, greater efficiency, more precision, fewer unintended consequences at the physical level, that of course leaves room for enormous unintended consequences at the social and political level. So those are issues to keep in mind in thinking about the spectrum of things being associated with accidents. The second sphere….

MW Actually we are quite short on time as well because we've got somebody else to talk to so I'm sorry….

ED Let me just very briefly summarise. The second sphere is associated with strategic issues, strategic competition, the fact that these technologies have very powerful abilities in the military sphere, and therefore there are a set of risks associated with deliberate competitive action and strategic stability. The further concerns are problems of abundance, we have seen those increasingly in society as technology has evolved, has then evolved advanced molecular manufacturing abilities will push us much further in that direction.

MW Sorry, what do you mean by that? By the term abundant?

ED Well the pursuit of low cost high nutritional density tasty food has been quest of the human race for a long time. In much of the world we have done a very good job of providing that, and there has also been a variety of problems of obesity and health problems. We have made inexpensive automobiles, and the result is that we have automobiles all over the landscape and automobile accidents and traffic jams.

MW Yes so it seems the relation between the technology and the human systems. Sorry somebody else wanted to come in did they? Okay yes….although I guess you could argue that some of those are not necessarily unique to nanotechnology. We’ve had a previous conversation today and thinking about the progress of technology more generally and those things come with that, rather than this particular technology.

ED Many of these problems are as a result of simply going further along paths that we are already pursuing and they raise known forms of difficulty in a more perhaps a bigger or larger scale fashion.

AD I think I have to break in there now. Thank you very much for that Dr. Drexler it's been interesting talking to you. And thank you for giving us your time.

ED Thank you. I very much appreciate the efforts of this group.

MW Thank you.