What was it that got you into researching self-replicating machines?
Like any engineer I look at biology with a mixture of awe (mainly) and contempt (a little). The engineering solutions exhibited by life are profoundly elegant and transcendently beautiful; this is interrupted by the occasional piece of crassness (like our optic nerves running over the _front_ of our retina) that shows that life must have evolved and not been designed. The two most important phenomena in biology are self-assembly and self-copying. For many years I have been interested in self-assembly. Indeed, the first student project I ever set when I was a young university lecturer was to get the students to write a simulation program for self-assembling systems. Recently I started some work on adapting some of George Whitesides’ ideas (he’s at Harvard, incidentally) on self-assembly to make self-assembling self-replicators. I also happen to be the person at Bath University responsible for rapid prototyping. Then, last year, I stripped the threads on a screw in my lathe in my home workshop. I drilled it out, tapped the hole, and replaced it with a piece of studding, using the lathe itself to do all this. This got me thinking about machines that make themselves, and I realised that a rapid prototyping machine is the ideal such machine to start from.
What percentage of household products do you see being made with these machines?
A guess: 25% in the medium (30 year?) term.
As they won’t be capable of producing glass and other parts, there will have to be parts kits for some products. How do you see these being retailed?
There is already a healthy web market for all sorts of engineering components; I see that growing considerably.
There is clearly a trade off in that the more complex the machine, the harder it is to replicate. And the harder it is to replicate the more complex the machine needs to be. Do you see a problem in deciding at what level of capability to release the plans?
We’re going to try to make the simplest machine we can that works. As with all design, we’ll first make parts of it, and only then see simpler ways to achieve the same function. We’ll have to restrain the temptation to perfect everything: good enough is best…
Once they can replicate, do you see a form of evolution occurring? If so, will that be led by AI, or human engineering, or a combination of both?
Yes, and definitely by human engineering to start with. There are some impressive results in AI, and I know some of the researchers. But we’re nowhere near having an AI system into which we can type, “Design me a new pair of pliers,” let alone more complicated things.
Despite the fact that the first unit would be expensive, the nature of these replicators means that additional copies will be relatively cheap. Do you see this property as something that will help poorer countries and economies?
Yes. I’m probably being over-idealistic (I once said that no level of cynicism is too high when it comes to politics…), but I see this as a liberating technology, putting the ability to create wealth into the poorest hands just as much as into those of the rich.
What level of assembly do you think the general public is capable of? Do you think the machines can be assembled by the public?
The first people to take up the technology will undoubtedly be geeks and hackers – no change there from every other piece of new technology. But they will have a strong incentive to improve the design in the direction of ease of assembly, so ultimately I can see any reasonably competent person being able to put one together. How strange would it have seemed twenty-five years ago if one had said that ordinary people would be drilling holes in their bedroom walls to thread ethernet cables through?
Do you envision these machines being able to re-use material from previously made objects that have outlived their usefulness?
Yes – see the end of the background article on the project website, where it talks of recycling.
Do you see this as strengthening our throwaway culture, by giving people the ability to create new objects as and when they’re needed? Or will it undermine it, by allowing people to recycle broken or unwanted items into new ones by using the machines?
The first, then the second. Remember – the machine should be able to make its own recycler…
Do you see your technology leading to devices that would enable space exploration/colonisation?
A bit. But, at least at the start, people will have to put each one together. I think genetic engineering as more important for space colonisation – we should be starting with extremophile organisms and pushing the compass of the extremes.
How do you see this changing the commercial economy? Will it make some businesses obsolete? And how could they evolve to embrace this new technology?
That is something I just don’t know. The most important point to make is that the idea may not take off at all – I have seen many good academic research projects, and vanishingly few of them make it to products. But if it does take off, it will certainly make some manufacturing industries redundant. But people will find ways to make money out of it, you can be sure. Remember, it’s not
employment that creates wealth, it’s wealth that creates employment. And this machine will make wealth…
Do you see yourself as having a role in its development after it’s released into the public domain, or will you move onto a new project?
Probably the latter. If it does not become self-propagating, then it will not have been a success. So, if it is a success, I will not be needed. And, once I’ve had an idea and shown that it can work, I want to go on to something new.
Do Star Trek replicator references annoy you?
No, not at all. Mark you, that insouciance may be a consequence of the fact that the last time I watched Star Trek was in 1969, so the reference is rather lost on me…
Although further away than the sort of technology you talk about, do you see nanotech as a competing technology? Or do you eventually see it becoming a part of your self-replicating machines?
I think the two will merge. We live on a planet most notable for having been covered knee-deep for 3.5 billion years in replicating nanobots. Indeed, our own knees are made from them, too. Microbes are what we will use to base our own nanobots on, and it will be a long time before we can do better than the biological ones. My sort of engineering is moving down from the top, and the biologists and biochemists are moving up from the bottom. Bits are already overlapping in the middle.
How soon do you think these machines will be available?
The BBC asked me that, and I said 4 to 20 years, or – if the idea dies, which it may well – never.
If self-replicating machines were led by AI design, do you think that they would develop into intelligent robots that would disregard Asimov and eventually take over the Earth? Or would it be ok?
The earth is already covered in replicating machines, as I pointed out above. We are not dropping anything very new into the mix, and any that we create will have to live with life or be eaten. I would back biology against engineering to come out on top every time.
The three laws are not evolutionarily-stable (in the John Maynard Smith sense) for any self-replicator. A completely selfless non-violent species could never evolve, and – if it were to be created – it would quickly mutate to have a small fraction of violent selfish individuals and a larger fraction of benign and helpful ones; just as we see in our own species and every other. That’s how the hawks-and-doves game-theoretic matrix must always level out.
Many thanks to Adrian Bowyer for agreeing to this interview.