Fast Takeoff: Exponential, or faster?

Molecular manufacturing will require R&D in many areas. Progress in these areas will happen at different speeds. MM will not have a big impact until a number of capabilities come together.

Hervé Musseau asks whether progress toward MM will happen exponentially, like the Human Genome Project, or according to the 80/20 rule where the last 20% is hard.

It definitely won't be exponential, overall; too many technologies have to come together, and each will develop at its own rate. But the flip side of that is that, once the last gating technology is developed, years of delay will be followed by rapid integration.

At least, I think the integration will be rapid, and this series of posts is written to explore the reasons why it will be.

One thing that could slow down MM is if there's a gating technology that no one identifies until they start trying to integrate. But that is looking less and less likely, for the simple reason that MM is still widely thought to be many decades away. I expect that the final push to integrate capabilities and develop a world-changing manufacturing capability will not get underway until it's blindingly obvious that it will work.

Even if someone started trying to build a diamondoid nanofactory today, we would see almost no progress for a while - and then almost overnight they'd go from a few machines that almost worked, to a fully functional nanofactory.

I don't think this fits either exponential metrics or the 80/20 rule. I don't think standard measures of technological progress will work here, since ultimately the impact of a nanofactory is less about its technology than about its products and their earth-shaking implications.

Chris Phoenix

Comments

[Tom Craver]

I don't necessarily agree that one would go almost overnight from machines that don't work, to a system of machines capable of producing accurate copies of themselves with minimal to zero human intervention.

It's far more likely that there will be a series of increasingly complex machines, each stage adding to or multiplying existing capabilities - creating basic capabilities, reducing error rates, increasing processing speed, becoming more automatic - probably with multiple loops within that sequence.

I could see that process taking years - but with it being fairly evident (except to the inevitable naysayers) from the start that it was eventually going to reach the point of rapidly self-copying systems.

We're really not to that starting point yet - we think it's possible, but there's no clear development path as yet, no equivalent of lithography that can be iteratively refined - at least so far as we know.

And I think it is unlikely that a secret government project could get going and make rapid progress, without conditions being so ripe for progress that multiple open projects also get underway - at least there'd be rumors of research being suppressed.

[Hervé Musseau]

It's nice (although a little strange too) to see oneself quoted, and I'm glad you found my question worthy of a follow-up entry.
I find it strange that you say it will not be exponential, yet say that years of development (into component techs) will be followed by rapid integration (into true MM), or speak of overnight progress. This sounds like the definition of an exponential field to me.
Progress will be slow at first, because there are many fields that must progress and come together, but once we have made sufficient progress in all those enabling steps and start putting things together, things will accelerate (your overnight expectation) and lead to the first nanofactories. And if I read your previous post correctly, those first factories that can produce a limited array of items will rapidly be used in conjunction to increase the range of what's feasible.

[jim moore]

CP "if someone started trying to build a diamondoid nano-factory today"

I think if you started today you would not call it a diamondoid nano-factory you would call it something more like Graphene Fabbing. (I know diamondiod includes graphene, it a matter of emphasis)

By starting with graphene you get an atomically "perfect" sheets of carbon you can cut with hot iron probe into almost any pattern you want. You use a system of pegs (carbon nano-tubes) to align the graphene sheets into 3 D parts. Then you chemically attach the carbon nano-tube "pegs" to the top and bottom sheet of graphene. And boom you have the bottom level parts for a nano-factory.

(more processing is of course possible, for example putting patterns of nitrogen and boron at the edges of the sheets could be interesting or taking a cut sheet of graphene and turning it into cut sheet of graphane so that it becomes an insulator would be really useful)

Now of course this method of making nano parts imposes design constrains on the products it can make but it should still be a very large and rich design space that is accesable with near term technology.

[Chris Phoenix]

Hervé, there are things that are faster than exponential. What I'm saying is that the impact of nanofactories may well be one of those things.

More precisely, I'm expecting some very fast exponential curves (number of nanofactories in the world, mass of their products, value of their products) to start at some definite point (when the first nanofactory is built) so that progress prior to that point cannot be used to predict the (very steep) slope of the exponential takeoff.

And of course, those measurable exponentials may have non-exponential effects, perturbing various world systems into new domains.

Hm, I think a couple more blog posts will come out of this thread...

Chris

[Hervé Musseau]

Good, I'm eagerly waiting for those new blog posts then.
I suppose you will make a distinction between the enabling part (up to the first nanofactory), and then the growth from there (spreading of the technology).
Both lines are interesting, of course. Since your blog series is about fast takeoff, I suppose you will start at the beginning, with the roadmap to the first factory, which advances are required in each domain involved, where are the potential problems, how it all comes together, and of course what is the likely timeline.

[todd andersen]

Jim Moore comment above makes the starting point for the MM factory. I would like to make a case for continuing his approach. I like the idea of using a 4 axis computer controlled robotic arm. The first arm would be build in the middle of the grapheme sheet the 4 axis are controlled by simple on off switch a pathway of 1 carbon nano tube when power is applied the arm moves in a defined way. The swivel of the arm will require more control perhaps 3 carbon switch the wires would run under the top sheet between a layer of insulator to a control point all the wires are run at time increment one and all the building blocks for the additional robotic arms are layer out on the grapheme sheet.

Once this is complete then the robotic arm follows a program and builds the arm next to itself then the 2 arms build the 2 arms next to them then the 4 arms build the arms next to them it is at this point the arm in the middle of the build can not reach a new arm to build and can not do anything this kills the expediential effect. But the outside arms can build more arms and in time a 1000 by 1000 arm sheet can be built this could give is a starting point to begin building a more complex factory. As we would now have 1 mill controlled pick and place tools to build a device.

todd

[Loki]

My intuition is that a 2-D initial molecular manufacturing system, as proposed above, is the way to go. It allows for simpler control and better visibility while debugging the process.