Maybe we need to differentiate between 'self-replication' and 'autonomous replication'? Perhaps the term 'recursive replication' is more accurate than 'self-replication', as 'self' seems to imply that a machine is self-contained and capable of independent action?

Presumably to cut off instructions to replicate, there needs to be some form of communications channel which can be broken. That is, the computer sending replication instructions must not be part of the machine being replicated.

These are all good points and I'd moderate the comment up but I don't think I can do so for articles where I have commented (probably a reasonable system). (So someone else with moderator access should uplift it.)

The only thing that I might add is that there needs to be a lot more work by people looking into the convergence of the biotech (wet) and nanotech (dry) approaches to molecular assembly. I'm working on that with the people becoming involved with Nano@Home but it is a very slow educational and developmental process. I think that the dry approach has been over stressed in the classical nanotech literature and that it may be time to look at a combination approach. (I think Eric realized this but very few people have read Nanosystems closely enough (one particularly really has to understand Table 16.1) to understand this.)

Robert

"I think one of the fundamental things which is not understood at this point is that artificial replicating systems, manufacturing systems, are going to bear about as much resemblance to the biological variety as, say, a 747 bears to a duck," Merkle said.

I bet that someone's planning a diamondoid duck the size of a 747 right now.

I tend to agree with Robert here. The metaphor is actually fairly appropriate, and if it has caused a problem then the blame probably lies more with non-technical people making invalid assumptions and extrapolating wildly than with the metaphor itself.

Of course there are differences — there would have to be at first, given that biological systems tend not to be made out of diamondoid at this time. But seriously, the current mechanocentric view of nanotechnology merely reflects our starting point, and I would expect a huge amount of convergence with biotech fairly early on given the intense interest in improving the lives of a certain protein-based species. Doubters should note that even without any convergence whatsoever, there will be interfacing in ever-increasing amounts, and ultimately the issues with replicated biological interfaces are no different to those with biotechnology itself.

Of course, there is always room for improvement in metaphors and analogies, and the most important improvement would probably come from laying to rest the grey goo meme once and for all. Fortunately, this is almost trivially done since engineering is about tradeoffs and the most obvious tradeoff for nanotech manufacturing is keeping the source of build instructions at the point of origin and routing the stream instead of replicating it. Almost everyone will understand (I think) that replicating goo scenarios are impossible under such circumstances, even in the presence of massive failure and mutation. Howard may at least be right in the sense that engineers didn't make that point clear right at the start. What we need now is someone to phrase the idea in a metaphor-compliant way.

The above also ties in with Robert's final statement that "We are going to face the nanotech hazards of biotech much much sooner than we will face the hazards of "classical" nanotech machinery", and not only because of the lead of biotech over nanotech in development. Given the extreme unlikelihood of replication problems in early nanotechnology, I see far more scope for nanotech in helping us control biotechnological dangers than in contributing to our worries.

Howard — the point you are trying to make is unclear. I can't understand why you would claim that Ralph's metaphor is poor. I've been educated in computer science, microbiology, biochemistry and molecular biology and his metaphor is completely reasonable.

The key point that most people do not understand is that biotech is nanotech. And it is to a large extent here now. What Ralph has failed to express is that we really need a way to implement something like a "broadcast architecture" for biotech (and I don't see a really easy way to do that). If the point that you (or Ralph) are trying to make is that it one would never put the source code into the machine then I would suggest the metaphor is flawed. We already have too many examples of where the source code in the machine is an efficient and functional architecture (biological systems are based upon this). To resolve this would require an extensive rethinking process in everything from the production of beer to yogurt.

Ralph and Robert Freitas are in the process of publishing what is likely to be the definitive reference work on self-replicating systems (SRS) ("Kinematic Self-Replicating Machines") and there are aspects of it which cover in detail aspects of biological (vs. say computer or mechanically enabled SRS). However it will need to be set side-by-side with the excellent work being done by several universities and institutes with respect to the risks of bioterrorism (which inherently involve SRS) — and which groups such as the NanoBusiness Alliance and VC firms such as DFJ are well aware of). In short people are starting to think seriously about both the power and risks of SRS.

If the point you are trying to make is that Ralph is claiming that biology cannot be compared to nanotech then I'd be happy to debate with Ralph on that topic. We are going to face the nanotech hazards of biotech much much sooner than we will face the hazards of "classical" nanotech machinery.

Robert

Another headline article containing the word "fear". Can we stop it please?