The fad is usually due to the whole idea of "nanobots" and "nanites" that will totally infect everything and kill cancer along the way. We can barely engineer a basic cell right now without having to severely use Mother Nature as a crutch (base DNA template, snip off as many bases as possible and see what's the least is needed to form a viable cell). A gray goo type of scenario is about as likely as AGI emerging from Minecraft and hijacking unsecured 3D printers to replicate themselves.
Wow, it has been that long. I met Drexler when he was first saying that stuff. My big question was "how do you power this stuff". Even if you had nanomachines, building something big with a chemically powered system might take as long as growing a tree. Maybe you could hook in external electrical power somehow. But the "clouds of nanomachines floating around" concept seemed to have a serious power problem.
"Nanotechnology" today is mostly surface chemistry. Surface chemistry is important, but boring. IC technology, modern battery technology, and membranes for separating things like salt and water are all based on clever tricks in surface chemistry. It's called "nanotechnology" to make it sound cool.
Amusingly, the IC fab people, who really do put atoms almost exactly where they want them, don't use the term.
Today's paperclip monster is yesterday's gray goo. Today's AGI is yesterday's nanoscale economy. All run into obvious mundane practical challenges like power, resources, scaleability, stability, error correction, and so on.
Of course if you assume those problems are easily solvable implementation details you can produce all kinds of dramatic speculations - which (IMO) are mostly nonsense.
You know everybody, don't you? Just last month Eric Volpe was telling me how you showed up at his house and gave him paper tape for his teletype!
> Even if you had nanomachines, building something big with a chemically powered system might take as long as growing a tree.
Suppose you have a tree-sized vat of milk --- 5 tonnes, say --- which has been boiled to sterilize it and allowed to cool down to 44°, the optimal temperature for Streptococcus thermophilus, where it is being maintained by a thermostat as it is stirred continuously with an impeller. You introduce a single Streptococcus bacterium, two microns in diameter, into the vat. If it happens to survive, how long does it take to convert the entire vat into yogurt, metabolizing most of the lactose into lactic acid?
Our initial seed bacterium is about 4 x 10¯¹⁸ m³ in volume. If we suppose that S. thermophilus doubles its population every 30 minutes under these circumstances, which is quite close to the truth, the answer is about 30 hours, entirely under the strength of the clouds of nanomachines floating around in the milk, despite the absence of external electrical power.
If we instead grant ourselves the luxury of dumping an entire 250-ml cup of seed yogurt into the vat, it only takes a bit over 7 hours.
You will note that this is noticeably faster than the time needed to grow a full-sized tree, even in a 44° rain forest; clearly there was plenty of chemical energy to transform a sufficient fraction of the vat into peptidoglycan to give it structure. This suggests that the tree's biology is not optimized to take advantage of the abundance of chemical resources available in such a quotidian object as a sterilized vat of milk; instead, it arises from a history of substantially more hostile environments, where such rapid growth will merely exhaust the available resources and then allow leaf-cutter ants to harvest your tender, unprotected leaves. There is no particular reason that nanotechnological factories need to operate under such conditions.
This is far from a wild speculation; yogurt-making is a well-understood industrial process, practiced for centuries if not millennia, which can be easily scaled down to tabletop experiments, and often is.
Of course, the gelled mass of bacterial corpses we call yogurt has substantially less macroscopic structure than a tree, and is also mechanically weaker, but that's not primarily a matter of available energy or speed of communication; it's a matter of evolutionary optimization. Organizing an ad-hoc network of 10²⁰ computers into a useful morphogenetic process, maybe one that builds a circulatory system into your house so it can keep getting nutrients until it finishes solidifying, is a nontrivial challenge, but it hardly seems like an insuperable one.
Zyvex's "convergent assembly" proposal provides an alternative that avoids the power problem in a different way and is less exposed to the vagaries of noisy self-replication and emergent morphogenesis.
I agree that most current "nanotechnology" is clearly humbug, sharing nothing with Drexler's (or Smalley's) ideas that it doesn't also share with asbestos or wood smoke.
AI and blockchains are a fad, and nanotech is, at the moment, a marketing fad as stated.
Although, the frustration with blockchain technology is more along the lines of "this probably is wrong in principle" rather than "this is probably hundreds of years away" as in the case of AI and nanotech.
Perhaps it was true when this was written, but it's certainly not true now that "mechanical objects on microscales do not exist". The whole field of MEMS is about them, which includes such prosaic things as smartphone gyros/accelerometers.
Even then it was not true, however nanoscale manufacturing as envisioned by Drexler definitely does not exist.
MEMS objects are made by lithography not pushing atoms around, and are extremely simple, usually just single chemical compound. (Sometimes a doped semiconductor.) Said simple structures are often very dissimilar to complexity envisioned by Drexler.
By the way, first M in MEMS is micro. It's not NEMS. NEMS also exists, but again, it's made with lithography and chemistry, not any level of direct molecular manipulation.
Heck, pharmacy chemists make nanostructures. (Liposomes and other vesicles specifically, for mechanical purposes. Peptides. Even protein assemblies.)
I think the bright spot in all this is that people learned from what happened to Drexler's program for responsibly developing advanced nanotech and so far have kept the AI safety program from the same fate. Drexler wrote a recent book on that subject, btw: https://www.fhi.ox.ac.uk/reframing/
At first I thought it was unjustifiably unkind of you to refer to the ignorant downvoters as bacteria, but now I see that wasn't what was meant.
The alternative to constructing atomically precise manipulators with other atomically precise manipulators is to construct them with macroscopic machinery such as scanning tunneling microscopes. Like writing a compiler in machine code with front-panel switches or copying an operating-system kernel with a handheld hole punch and paper tape, this may be feasible, but it will be extremely difficult; once we have a working atomically precise manipulator, it will be an enormously easier way to build more of them with it.
That's the reason for Drexler's prudent concern for safety protocols in advance of the actual development of the machinery.
Of course, there's no reason that the constructor itself need be a micron-scale device; the convergent-assembly proposals from Zyvex and Nanorex instead use macroscopic machines with a large number of nanometer-scale atomically-precise parts.
How can one stay on top of and get involved in the latest work in MEMS?
Can someone elaborate on the difference between licensed software for MEMS multiphysics simulation like Comsol vs open source software simulated collaboration?
Have open fab tool projects been helpful in creating equipment that could help amateurs do more scientific research at this scale?
How could one attempt to experimentally validate Drexler's latest proposed 3d printer design with x-y-z axis photosteppers? [1] (Slide 14)
Silicon atom mechanosynthesis experiments have succeeded, but it seems there is not enough funding for direct to DMS experiments with pico-stabilized tooltips. Would it not be a good use of a billion dollar fund to experimentally research the DFT-simulated diamondoid minimal tooltip set? [2]
18 comments
[ 21.1 ms ] story [ 1117 ms ] threadhttps://www.youtube.com/watch?v=zMRrNY0pxfM
"Nanotechnology" today is mostly surface chemistry. Surface chemistry is important, but boring. IC technology, modern battery technology, and membranes for separating things like salt and water are all based on clever tricks in surface chemistry. It's called "nanotechnology" to make it sound cool.
Amusingly, the IC fab people, who really do put atoms almost exactly where they want them, don't use the term.
Of course if you assume those problems are easily solvable implementation details you can produce all kinds of dramatic speculations - which (IMO) are mostly nonsense.
> Even if you had nanomachines, building something big with a chemically powered system might take as long as growing a tree.
Suppose you have a tree-sized vat of milk --- 5 tonnes, say --- which has been boiled to sterilize it and allowed to cool down to 44°, the optimal temperature for Streptococcus thermophilus, where it is being maintained by a thermostat as it is stirred continuously with an impeller. You introduce a single Streptococcus bacterium, two microns in diameter, into the vat. If it happens to survive, how long does it take to convert the entire vat into yogurt, metabolizing most of the lactose into lactic acid?
Our initial seed bacterium is about 4 x 10¯¹⁸ m³ in volume. If we suppose that S. thermophilus doubles its population every 30 minutes under these circumstances, which is quite close to the truth, the answer is about 30 hours, entirely under the strength of the clouds of nanomachines floating around in the milk, despite the absence of external electrical power.
If we instead grant ourselves the luxury of dumping an entire 250-ml cup of seed yogurt into the vat, it only takes a bit over 7 hours.
You will note that this is noticeably faster than the time needed to grow a full-sized tree, even in a 44° rain forest; clearly there was plenty of chemical energy to transform a sufficient fraction of the vat into peptidoglycan to give it structure. This suggests that the tree's biology is not optimized to take advantage of the abundance of chemical resources available in such a quotidian object as a sterilized vat of milk; instead, it arises from a history of substantially more hostile environments, where such rapid growth will merely exhaust the available resources and then allow leaf-cutter ants to harvest your tender, unprotected leaves. There is no particular reason that nanotechnological factories need to operate under such conditions.
This is far from a wild speculation; yogurt-making is a well-understood industrial process, practiced for centuries if not millennia, which can be easily scaled down to tabletop experiments, and often is.
Of course, the gelled mass of bacterial corpses we call yogurt has substantially less macroscopic structure than a tree, and is also mechanically weaker, but that's not primarily a matter of available energy or speed of communication; it's a matter of evolutionary optimization. Organizing an ad-hoc network of 10²⁰ computers into a useful morphogenetic process, maybe one that builds a circulatory system into your house so it can keep getting nutrients until it finishes solidifying, is a nontrivial challenge, but it hardly seems like an insuperable one.
Zyvex's "convergent assembly" proposal provides an alternative that avoids the power problem in a different way and is less exposed to the vagaries of noisy self-replication and emergent morphogenesis.
I agree that most current "nanotechnology" is clearly humbug, sharing nothing with Drexler's (or Smalley's) ideas that it doesn't also share with asbestos or wood smoke.
Although, the frustration with blockchain technology is more along the lines of "this probably is wrong in principle" rather than "this is probably hundreds of years away" as in the case of AI and nanotech.
MEMS objects are made by lithography not pushing atoms around, and are extremely simple, usually just single chemical compound. (Sometimes a doped semiconductor.) Said simple structures are often very dissimilar to complexity envisioned by Drexler.
By the way, first M in MEMS is micro. It's not NEMS. NEMS also exists, but again, it's made with lithography and chemistry, not any level of direct molecular manipulation.
Heck, pharmacy chemists make nanostructures. (Liposomes and other vesicles specifically, for mechanical purposes. Peptides. Even protein assemblies.)
It is not. I'm disappointed to see such lazy dishonest rhetoric on the front page of a site for the intellectually curious. The thesis in question: https://e-drexler.com/d/09/00/Drexler_MIT_dissertation.pdf
I think the bright spot in all this is that people learned from what happened to Drexler's program for responsibly developing advanced nanotech and so far have kept the AI safety program from the same fate. Drexler wrote a recent book on that subject, btw: https://www.fhi.ox.ac.uk/reframing/
We need to clean this place up.
The alternative to constructing atomically precise manipulators with other atomically precise manipulators is to construct them with macroscopic machinery such as scanning tunneling microscopes. Like writing a compiler in machine code with front-panel switches or copying an operating-system kernel with a handheld hole punch and paper tape, this may be feasible, but it will be extremely difficult; once we have a working atomically precise manipulator, it will be an enormously easier way to build more of them with it.
That's the reason for Drexler's prudent concern for safety protocols in advance of the actual development of the machinery.
Of course, there's no reason that the constructor itself need be a micron-scale device; the convergent-assembly proposals from Zyvex and Nanorex instead use macroscopic machines with a large number of nanometer-scale atomically-precise parts.
Silicon atom mechanosynthesis experiments have succeeded, but it seems there is not enough funding for direct to DMS experiments with pico-stabilized tooltips. Would it not be a good use of a billion dollar fund to experimentally research the DFT-simulated diamondoid minimal tooltip set? [2]
---
[1] https://www.energy.gov/sites/prod/files/2016/06/f33/Keynote%... [2] http://www.molecularassembler.com/Papers/MinToolset.pdf