grey goo is unlikely

bhauth17 Apr 2023 1:59 UTC
158 points
120 comments9 min readLW link1 review

Link post

The nanomachinery builds diamondoid bacteria, that replicate with solar power and atmospheric CHON, maybe aggregate into some miniature rockets or jets so they can ride the jetstream to spread across the Earth’s atmosphere, get into human bloodstreams and hide, strike on a timer.

— Eliezer Yudkowsky

To control these atoms you need some sort of molecular chaperone that can also serve as a catalyst. You need a fairly large group of other atoms arranged in a complex, articulated, three-dimensional way to activate the substrate and bring in the reactant, and massage the two until they react in just the desired way. You need something very much like an enzyme.

— Richard Smalley

My understanding is that anyone who can grasp what “orthos wildly attacking the heterodox without reading their stuff and making up positions to attack” looks like, considers that this is what Smalley did with Drexler—made up an unworkable approach and argued against it.

— Eliezer Yudkowsky

In this post, I use “nanobots” to mean “self-replicating microscopic machines with some fundamental mechanistic differences from all biological life that make them superior”. Various specific differences from biological cells have been proposed. I’ve organized this post by those proposed differences.

1. localized melting

Most 3d printers melt material to extrude it through a nozzle. Large heat differences can’t be maintained on a small scale.

2. rare materials

If a nanobot consists largely of something rare, getting more of that material to replicate is difficult outside controlled environments.

Growth of algae and bacteria is often limited by availability of iron, which is more common than most elements. Iron is the active catalytic site of many enzymes, and is needed by all known life. The growth of something made mostly of iron would be far more limited, and other metals have more limited availability than that.

3. metal surfaces

Melting material isn’t feasible per (1), so material must be built up by adding to the surface. Since that’s the case, the inside of structures must be chemically the same as what was the exterior.

Metal objects have a protective oxide layer. In an air or water environment, there’s no way to add individual (eg) aluminum atoms to a metal surface and end up with metallic aluminum inside; the whole thing will typically be aluminum oxide or hydroxide.

Corrosion is also a proportionately bigger problem for smaller objects. A micrometer-scale metal structure will rapidly corrode, perhaps doing some Ostwald ripening.

4. electric motors

Normal “electric motors” are all electromagnetic motors, typically using ferromagnetic cores for windings. Bigger is better for those, up to at least the point where you can saturate cores.

On a very small scale, it’s better to use electrostatic motors, and you can make MEMS electrostatic motors with lithography. (Not just theoretically; people actually do that.) But, per (2) & (3), bulk metals are a problem for a self-replicating system. If you need to have compounds floating around, electrical insulation is also difficult. You also need some way to switch current, and while small semiconductor switches are possible, per (3) building them is difficult.

Instead of electrostatic charge of metal objects, it’s better to use ions. Ions could bind to some molecule, and electrostatic forces could cause that to rotate relative to another molecule. Hmm, this is starting to sound rather familiar.

5. inorganic catalysts

Lab chemistry and drug synthesis often use metal catalysts in solution, perhaps with a small ligand. Palladium acetate is used for making drugs, but it’s very toxic to humans, because it...catalyzes reactions.

Life requires control of what happens, which means selective catalysis of reactions, which means molecules need to be selectively bound, which requires specific arrangements of hydrogen bond donors and acceptors and so on, and that requires organic compounds. Controlled catalysis requires organic compounds.

6. no liquid

Any self-replicating nanobot must have many internal components. If the interior is not filled with water, those components will clump together and be unable to move around, because electrostatic & dispersion interactions are proportionately much stronger on a small scale. The same is true to a lesser extent for the nanobots themselves.

Vacuum is even worse. Any self-replicating cell must move material between outside and multiple compartments. Gas leakage by the transporters would be inevitable. Cellular vacuum pumps would require too much energy and may be impossible. Also, strongly binding the compounds used (eg CO2) to carriers at every step would require too much energy. (“Too much energy” means too much to be competitive with normal biological processes.)

7. no water

Most enzymes maintain their shape because the interior is hydrophobic and the exterior is hydrophilic. If some polar solvent is used instead of water, then this stability is weakened; most organic solvents will denature most proteins. If you use a hydrophobic solvent, it can’t dissolve ions or facilitate many reactions.

Ester and amide bonds are the best ways to reversibly connect organic molecules. Both involve making or taking water or alcohol. Alcohols have no advantages over water in terms of conditions where they’re stable.

Water is by far the best choice of liquid. The effectiveness of water for dissolving ions is unique. Water can help catalyze reactions by donating and accepting hydrogen. Water is common on Earth, easy to get and easy to maintain levels of.

8. high temperatures

Per (5) you need organic molecules to selectively catalyze reactions.

Enzymes need to be able to change shape somewhat. Without conformational changes, enzymes can’t grab their substrate well enough. Without conformational changes, there’s no way to drive an unfavorable reaction with a favorable reaction, and that’s necessary.

Because enzymes must be able to do conformational changes, they need to have some strong interactions and some weaker interactions that can be broken or shifted. Those weaker interactions can’t hold molecules together at high temperatures. Some life can grow at 100 C but 200 C isn’t possible.

This means that the reactions you can do are limited to what organic compounds can do at relatively low temperatures—and existing life can pretty much do anything useful in that category already.

9. diamond

It’s possible to make molecules containing diamond structures at ambient temperature. The synthesis involves carbocations or carboanions or carbon radicals, which are all very unstable. The yields are mediocre and the compounds involved are reactive enough to destroy any conceivable enzymes.

Some people have simulated structures that could theoretically place carbon atoms on diamond in specific positions at ambient temperature. Here’s a paper on that. Because diamond is so kinetically stable, the synthesis must be exothermic, with high-energy intermediates. So, high vacuum is required, which per (6) doesn’t work.

Also, the chemicals consumed to make those high-energy intermediates are too reactive to plausibly be made by any enzyme-like system. And per (1) & (8) you can’t use high temperatures to make them on a small scale.

Also, there is no way to later remove carbon atoms from the diamond at low temperature. How, then, would a nanobot with a diamond shell replicate?

10. other rigid materials

CaCO3, silica, and apatite are much easier to manipulate than diamond. They’re used in (respectively) mollusk shells, diatom frustules, and bone.

If it was advantageous to use structures of those inside cells for reactions somehow, then some organisms would already do that. Enzymes generally must do conformational changes to catalyze reactions. A completely rigid diamond shape with functional groups would not make a particularly good enzyme.

And of course, just a small solid shape, with nothing attached to it—even if you can make arbitary shapes—isn’t useful for much besides cell scaffolding, and even then, building diatom frustules out of linked diamond pieces seems worse than what they do now with silica. Sure, diamond is even stronger than silica, but that doesn’t matter. And that’s assuming you can make interlocking diamond pieces, which you can’t.

11. 3d structures

Unlike cells, nanobots could make 3d structures, instead of being limited to a soup of folded linear structures.

Yes, believe it or not, I’ve seen people say that. But cells have eg microfilaments.

Again, enzymes must be able to do conformational changes to work. At ambient temperature, that means they’re shaking violently, and if proteins are flopping around constantly, you can’t have a rigid positioner move to a fixed position and assume you’re placing something correctly.

What you can do is hold onto the end of a linear chain as you extrude it, then fold up that chain into a 3d structure. What you can do is use an enzyme that binds to 2 folded proteins and connects them together. And those are methods that are used by all known life.

12. active transport

Life relies on diffusion and random collisions; nanobots could intentionally move things around.

Yes, I’ve actually seen people say that, but cells do use myosin to transport proteins sometimes. That uses a lot of energy, so it’s only used for large things.

13. combining reaction steps

Nanobots could put all the sequential reaction steps next to each other, making them much more efficient than cells.

Cells have compartments with proteins that do related reactions. Some proteins form complexes that do multiple reaction steps. Existing life already does this to the extent that it makes sense to.

14. positional nanoassembly

The above sections should be enough background to finally cover what’s perhaps the most central concept of the genre of proposals called “nanobots”.

Some people see 3d printers and CNC routers, and don’t understand enzymes or what changes on a molecular scale very well, and think that cells that work more like 3d printers or gantry cranes would be better. Now, a FDM 3d printer has several components:

  • sensors that detect the current position

  • drivers that control motors based on sensors

  • 3 motors that do 3-axis movement

  • a rigid bed and rigid drive system

  • a good connection between the bed and material being printed

  • a nozzle that melts material

Protein-sized position sensors don’t exist.

Molecular linear motors do exist, but 1 ATP (or other energy carrier) is needed for every step taken.

If you want to catalyze reactions, you need floppy enzymes. Even if you attach them to a rigid bed, they’ll flop all over the place. (On a microscopic scale, normal temperatures are like a macroscopic 3d printer being shaken violently.)

Suppose you’re printing diamond somehow. You need a seed that’s rigidly connected to the printing mechanism. The connection would need to be removable in order to detach the product from the printer. In a large 3d printer, you can peel plastic off a metal surface, but that won’t work for covalently bonded diamond. You would need a diamond seed with functional groups that allow it to be grabbed, and since you’re not starting with a sheet, you’d need a 5-axis printer arm.

Drexler wrote a book that proposed mechanical computers which control positioner arms by lever assemblies. An obvious problem there is mechanical wear—yes, some MEMS devices have adequate lifetimes, but those just vibrate; their sliding friction is negligible. But suppose you can solve this by making everything out of diamond or using something like lubricin.

So, suppose you have a mechanical computer that moves arms that control placement of something. Diamond is impractical, so let’s say silica is being placed. Whatever you’re placing, you need chemical intermediates that go on the arms, and you need energy to power everything. Making energy from fuel or photosynthesis requires more specific chemicals, not just specific arrangements of some solid. To do the reactions needed for energy and intermediate production, you need things that can do conformational changes—enzymes.

Without conformational changes, enzymes can’t grab their substrate well enough. Without conformational changes, there’s no way to drive an unfavorable reaction with a favorable reaction, and that’s necessary. You can’t just use rigid positioners to drive reactions that way, because they have no way to sense that the reaction has happened or not...except through conformational changes of a flexible enzyme-like tooltip on the positioner, which would have the same issues here.

At ambient temperature, enzymes that can do conformational changes are shaking violently, and if proteins are flopping around constantly, you can’t have a rigid positioner move to a fixed position and assume you’re placing something correctly. Since you need enzymes, you need a ribosome, and production of monomers—and amino acids are the best choice, chemical elements are limited and there is no superior alternative.

Since all that is still needed, what are the positioners actually accomplishing? They’d only be needed to build positioners. The whole thing would be a redundant side system to enzymatic life.

OK, maybe you want to build some kind of mechanical computers too. Clearly, life doesn’t require that for operation, but does that even work? Consider a mechanical computer indicating a position. It has some number, and the high bit corresponds to a large positional difference, which means you need a long lever, and then the force is too weak, so you’d need some mechanical amplifier. So that’s a problem.

Consider also that as vacuum is impractical per (6), and enzymes and chemical intermediates are needed, you’d have stuff floating around. So you have all these moving parts, they need to interface with the enzymes so they can’t just be separated by a solid barrier, and stuff could get in there and jam the system.

The problems are myriad, and I’d be well-positioned to see solutions if any existed. But suppose you solve them and make tiny mechanical computers in cells—what’s the hypothetical advantage of that? The ability to “do computation”? Brains are more energy-efficient than semiconductor computers for many tasks, and the total embodied computation in cells is far greater than that of neurons’ occasional spikes.

15. everything else

When someone has an idea about something cells could do, it’s often reasonable to presume that it’s either impossible, useless, or already used by some organism—but there are obviously cases where improvement is possible. It’s certainly physically possible to correct harmful mutations with genetic engineering. There are also ongoing arms races between pathogens and hosts where each step is an informational problem.

But what about more basic mechanisms? Have basic mechanisms for typical Earth conditions been optimized to the point that no improvement is possible? That depends on their complexity. For example, glycolysis and the citric acid cycle are optimal, but here’s a more-efficient CO2 fixation pathway I designed. (Yes, you’d want to assimilate the glycolaldehyde synthons by (erythrose 4-phosphate → glucose 6-phosphate → 2x erythrose 4-phosphate). I left that as a way for people to show they understood my blog.) See also my post on the origin of life for some reasons life works the way it does. (You can see I’m a big blogger—that’s a good career plan, right?)

I wrote this post now as a sort of side note to my post on AI risks. But...what if a superintelligence finds something I didn’t think of?

I know, right? What if it finds a way to travel faster than light and sets up in Alpha Centauri, then comes back? What if it finds a way to make unlimited free energy? What if it finds a friendly unicorn that grants it 3 wishes?

There’s a gap between seeing that something is conceivably possible and seeing how to do it, and that’s the only reason that things like research and planning and prediction about the future are possible. I understand Eliezer Yudkowsky thinks that someone a little smarter than von Neumann (who didn’t invent the “von Neumann architecture” or half the other stuff he took credit for, but that’s off topic) would be able to invent “grey goo” type nanobots. If that was the case, even I would at least be able to see how it would be done.

To be clear, I’m not trying to imply that a superintelligent AI wouldn’t have any plausible route to taking over societies or killing most of humanity or various other undesirable outcomes. I’m only saying that worrying about “grey goo” is a waste of time. On the other hand, Smalley was mad at Drexler for scaring people away from research into carbon nanotubes, but carbon nanotubes would be a health hazard if they were used widely, and the applications Smalley hoped for weren’t practical. Perhaps I would thank Drexler if he actually pushed people away from working on carbon nanotubes, but he didn’t.

bhauth17 Apr 2023 1:59 UTC
158 points
120 comments9 min readLW link1 review