Nanotech Design Problems

No, not GreyGoo. GreyGooIsEasilyContained, so it's a silly thing to worry about. But to build nanotech on any scale that will suit intentional purposes beyond the simple filters and surface treatments we contemplate today - to build real DrexlerianNanotech - is going to require us to solve some engineering problems that we've never been able to solve before.

The point is we're not going to solve these problems overnight. And until we do, all those EnginesOfCreation? scenarios Drexler dreamed up are completely unrealistic. Nanomanipulators are going to look more like trees with tinier twigs on tinier twigs until you get down to the nanoscale - all connected back to some big humming machine on our scale - than any kind of free-floating submersible nanobots you can tell to go increase ATP production in mitochondria #345,678,901,322 through #345,799,999,999.

Physically the nanobots are possible. In engineering terms, however, it'll take thousands of years, if ever, for us to build 'em based on modern software engineering techniques. We need to rethink the VonNeumannArchitecture big time - we need self-aware self-directed self-evolving machines - before we can do anything like what Drexler proposes to do.

self-aware self-directed self-evolving machines? No, thank you.

This all seems to suggest that an organizational form larger than pure "nanotech" is needed. Maybe some cell-like form is needed structurally: see AboutCells.

That would be a bad idea for several reasons. The crucial difference between nanotech and biological systems is that nanotech exists to construct stuff whereas biological systems construct stuff to exist. When building something using self-replicating nanotech, we don't want that thing to be self-replicating nanotech. Physical cells would make achieving this entirely too difficult.

On the other hand, logical cells are a very, very good idea. And DNA is also a very, very good idea. Why is that? Well, it's because two of the above problems cancel each other out if we just think about them the right way. Ask yourself this, why would you ever want to communicate with individual nanomachines? You wouldn't, the notion of addressing individual nanomachines is entirely ridiculous. Now ask yourself why would you ever want an individual nanomachine to stay in one particular place over time? Again, you wouldn't.

What you want is for a nanomachine at position (x,y,z) to act in a way that's appropriate to this position. If it gets displaced from its position then that's not a big deal so long as it acts in a way that's appropriate to its new position and that another nanomachine replaces it in its old position. And this can be achieved by broadcasting uniquely identifiable signals in three dimensions with time-encoded information so that the nanomachines can triangulate. Basically, you need GPS. And of course, using DNA (globally distributed information used differently by each cell).

How would the nanomachines receive such signals?

Do you not realize this question is answered elsewhere on this very page?

Adding up all these problems seems to point to the need for something almost as complex as a living organism to maintain its existence in such environments.

Not necessarily. Don't underestimate humans as engineers - we have made horseless carriages, moon rockets, and the global computing network you're using right now. An obvious non-bio-style solution to these problems is a kind of UtilityFog - a geodesic mesh of nanowires and nanobots, each in logical contact with its neighbours, and each capable of exerting tiny forces on its neighbours. Then problem #1 above is no problem - the bots will simply communicate via IPv7. Problem #3 is no problem because the bots can use electrical signals. And problem #4 is no problem because each bot can measure the forces being exerted on its local mesh, and by integration determine where it is.

Problem #2 though seems like a toughy ...

Umm... a utility fog certainly seems to be getting near the scale of a typical cell. Although obviously of a different nature, we couldn't really claim that we're doing that much better than nature anymore :)

[Everything we do is part of nature, so we can't claim we're doing anything better than nature. We're good at finding simple, alternative ways to accomplish the same things DNA accomplished millions of years ago. Only if we're really lucky will our nanotech work like DNA's existing nanotech.]

Coordinating billions of logical processors.

Yes, the most powerful supercomputer in the world has "only" 5,120 processors. But when I look at photographs of it (, I get the impression perhaps there are other reasons why there are no million processor computers (WhyAreThereNoMillionProcessorComputers?). Finding enough room for them all, and cost jump to mind. Supposedly nanotechnology will eliminate those 2 barriers.

Navigating by touch.

My understanding is that most industrial robots in use today don't use cameras or lenses. Perhaps nanobots don't need cameras or lenses, either.

Communicating via molecular signals

Yes, communicating is very important. If you skim down to the "The Broadcast Architecture" section of we see one way of transmitting information from a single macroscopic computer to the nanobots (sound waves). Another proposed method: photons. Single molecules in the human eye are sensitive to red photons. Other single molecules in the human eye are sensitive to green photons. Give each nanobot a few of these molecules, and broadcast messages to it by blinking large red and green lights.

That's all the communication you need for some kinds of nanobots.

Of course, a way for nanobots to communicate with each other and back to the macroscopic computer would be tremendously useful.

Useful? Maybe. Important? No. Necessary? Hell no. At least, not for an assembler. Now for a disassembler, that's a whole different problem.


EditText of this page (last edited February 21, 2006) or FindPage with title or text search