Two months ago I attended Eric Drexler's launch of
MSEP.one. It's open source software, written by
people with professional game design experience, intended to catalyze
better designs for atomically precise manufacturing (or generative
nanotechnology, as he now calls it).
Drexler wants to draw more attention to the benefits of nanotech, which
involve large enough exponents that our intuition boggles at handling
them. That includes permanent health (Drexler's new framing of life
extension and cures for aging).
He hopes that a decentralized network of users will create a rich
library of open-source components that might be used to build a nanotech
factory. With enough effort, it could then become possible to design a
complete enough factory that critics would have to shift from their
current practice of claiming nanotech is impossible, to arguing with
expert chemists over how well it would work.
Drexler hopes to gamify the software enough that people will use it for
fun. My cursory impression, based on playing with it for less than an
hour, is that it's not very close to being fun. I don't know if
that's even a reasonable goal. The software feels more professional and
easy to use than what I recall of similar software 20 years ago. Using
it still seems like work. I expect it will be hard to get many people to
use it without paying them.
Protein-based nanotech versus Diamondoid-style
MSEP.one currently supports development of diamondoid-style nanotech,
which produces pretty pictures, and seems somewhat likely to enable the
simplest and most understandable atomically precise factories. The
downside is that designs of this type are not very close to being
buildable with current tools. There isn't even a clear path to building
the appropriate tools.
There is at least as much interest in building atomically precise
systems out of proteins and/or DNA. The tools to build those designs
mostly work today. MSEP.one does not yet support such designs. The main
downsides are that they're harder to design and visualize. Proteins in
particular are big messy-looking blobs. It typically requires
sophisticated software to determine whether two of them will fit
together well. I haven't looked much at how hard it is to design a
protein to fit an arbitrary shape, but my impression is that it requires
a painful amount of trial and error.
Yet the presence of life indicates that protein-based engineering can
generate a wide enough variety of atomically precise systems that it
ought to be possible to use protein designs as a path to something like
diamondoid nanotech. The feasibility of other paths seem less clear.
This Manifold
market
shows a 51% chance that a protein-based path will be used, with no other
path having more than 13%.
Nanotech Risks
I was fairly eager 20 to 25 years ago to help develop Drexlerian
nanotech. Now I feel a fair amount of reluctance to do so.
At least half of that reluctance is due to concerns about how nanotech
will affect the risks associated with AI.
Having powerful nanotech around when AI becomes more competent than
humans will make it somewhat easier for AIs to take control of the
world. I only see a small chance of nanotech making a decisive
difference here, but small probabilities of altering which species
controls galaxies can be pretty important.
A more likely risk is that the availability of nanotech would influence
whether one nation or company could conquer the world if it gets a lead
in AI capabilities. I can easily imagine that AI will have politically
destabilizing effects. Nanotech seems likely to add fuel to such
destabilization.
So it seems safer to postpone the development of nanotech until after
the most chaotic aspects of AI development have been resolved.
Awareness
Drexler's response to these concerns is to focus on the risk that world
leaders will be caught by surprise at how powerful generative nanotech
is. There's also a large chance that AI will develop nanotech soon
after AI surpasses human levels. Greater awareness before then of
nanotech possibilities will lead to better contingency planning.
I have some fairly large concerns that world leaders will wait until the
last moment to react to evidence that AI is becoming powerful, resulting
in policies that are as poorly planned as the response to COVID.
But how much of a difference would it make for them to know that
generative nanotech is consistent with the laws of physics, and
therefore somewhat likely to be available shortly after AI surpasses
human-level intelligence? (Manifold suggests that a Drexlerian assembler
will be developed about a year after
superintelligence.)
If they're in denial about what an AI could do with tactics such as
drone swarms and blackmail, why are more powerful technologies likely to
be persuasive?
A more basic concern is whether new information about nanotech will
cause any reaction at all. Freitas'
Nanomedicine has plenty of details about
how to design better medical devices. Approximately no skeptics are
willing to carefully read enough of the book to evaluate whether those
designs are physically possible. Skeptics find it much easier to dismiss
the designs as impractical to build or needing too much further
research.
Who would we convince?
However, I don't assume that better designs should be directed at
people who currently identify as nanotech skeptics. It's more likely
that a key shift in opinion will be centered on people who currently
aren't paying enough attention to nanotech to have much of any opinion
on it.
Back in the dot-com bubble, I had some contact with VCs who had a decent
vision of how valuable nanotech could be (and also some lesser VCs who
struggled to imagine it being worth a few billion). AFAICT, the best VCs
rejected nanotech startups because nanotech was well over 5 years away.
It's somewhat plausible that that kind of VC can be persuaded that
nanotech now could be profitable with less than 10 years of effort.
Those VCs could influence some other world leaders to think about
contingency plans related to nanotech.
The downside with convincing VCs is that it requires rather more than
providing evidence that AI will eventually build nanotech. It requires
at least outlining a path that will get us to commercial nanotech in the
early 2030s. The would seem to speed up the availability of nanotech,
possibly more than we're prepared to handle.
Weighing the Tradeoffs
I have the skills, and possibly the time, to make a modest impact on
Drexler's project.
Writing this post has left me confused about whether I should put
significant effort into using MSEP.one. I'm tentatively leaning toward
no. I welcome input from readers on this question.
Two months ago I attended Eric Drexler's launch of MSEP.one. It's open source software, written by people with professional game design experience, intended to catalyze better designs for atomically precise manufacturing (or generative nanotechnology, as he now calls it).
Drexler wants to draw more attention to the benefits of nanotech, which involve large enough exponents that our intuition boggles at handling them. That includes permanent health (Drexler's new framing of life extension and cures for aging).
He hopes that a decentralized network of users will create a rich library of open-source components that might be used to build a nanotech factory. With enough effort, it could then become possible to design a complete enough factory that critics would have to shift from their current practice of claiming nanotech is impossible, to arguing with expert chemists over how well it would work.
Drexler hopes to gamify the software enough that people will use it for fun. My cursory impression, based on playing with it for less than an hour, is that it's not very close to being fun. I don't know if that's even a reasonable goal. The software feels more professional and easy to use than what I recall of similar software 20 years ago. Using it still seems like work. I expect it will be hard to get many people to use it without paying them.
Protein-based nanotech versus Diamondoid-style
MSEP.one currently supports development of diamondoid-style nanotech, which produces pretty pictures, and seems somewhat likely to enable the simplest and most understandable atomically precise factories. The downside is that designs of this type are not very close to being buildable with current tools. There isn't even a clear path to building the appropriate tools.
There is at least as much interest in building atomically precise systems out of proteins and/or DNA. The tools to build those designs mostly work today. MSEP.one does not yet support such designs. The main downsides are that they're harder to design and visualize. Proteins in particular are big messy-looking blobs. It typically requires sophisticated software to determine whether two of them will fit together well. I haven't looked much at how hard it is to design a protein to fit an arbitrary shape, but my impression is that it requires a painful amount of trial and error.
Yet the presence of life indicates that protein-based engineering can generate a wide enough variety of atomically precise systems that it ought to be possible to use protein designs as a path to something like diamondoid nanotech. The feasibility of other paths seem less clear. This Manifold market shows a 51% chance that a protein-based path will be used, with no other path having more than 13%.
Nanotech Risks
I was fairly eager 20 to 25 years ago to help develop Drexlerian nanotech. Now I feel a fair amount of reluctance to do so.
At least half of that reluctance is due to concerns about how nanotech will affect the risks associated with AI.
Having powerful nanotech around when AI becomes more competent than humans will make it somewhat easier for AIs to take control of the world. I only see a small chance of nanotech making a decisive difference here, but small probabilities of altering which species controls galaxies can be pretty important.
A more likely risk is that the availability of nanotech would influence whether one nation or company could conquer the world if it gets a lead in AI capabilities. I can easily imagine that AI will have politically destabilizing effects. Nanotech seems likely to add fuel to such destabilization.
So it seems safer to postpone the development of nanotech until after the most chaotic aspects of AI development have been resolved.
Awareness
Drexler's response to these concerns is to focus on the risk that world leaders will be caught by surprise at how powerful generative nanotech is. There's also a large chance that AI will develop nanotech soon after AI surpasses human levels. Greater awareness before then of nanotech possibilities will lead to better contingency planning.
I have some fairly large concerns that world leaders will wait until the last moment to react to evidence that AI is becoming powerful, resulting in policies that are as poorly planned as the response to COVID.
But how much of a difference would it make for them to know that generative nanotech is consistent with the laws of physics, and therefore somewhat likely to be available shortly after AI surpasses human-level intelligence? (Manifold suggests that a Drexlerian assembler will be developed about a year after superintelligence.)
If they're in denial about what an AI could do with tactics such as drone swarms and blackmail, why are more powerful technologies likely to be persuasive?
A more basic concern is whether new information about nanotech will cause any reaction at all. Freitas' Nanomedicine has plenty of details about how to design better medical devices. Approximately no skeptics are willing to carefully read enough of the book to evaluate whether those designs are physically possible. Skeptics find it much easier to dismiss the designs as impractical to build or needing too much further research.
Who would we convince?
However, I don't assume that better designs should be directed at people who currently identify as nanotech skeptics. It's more likely that a key shift in opinion will be centered on people who currently aren't paying enough attention to nanotech to have much of any opinion on it.
Back in the dot-com bubble, I had some contact with VCs who had a decent vision of how valuable nanotech could be (and also some lesser VCs who struggled to imagine it being worth a few billion). AFAICT, the best VCs rejected nanotech startups because nanotech was well over 5 years away. It's somewhat plausible that that kind of VC can be persuaded that nanotech now could be profitable with less than 10 years of effort. Those VCs could influence some other world leaders to think about contingency plans related to nanotech.
The downside with convincing VCs is that it requires rather more than providing evidence that AI will eventually build nanotech. It requires at least outlining a path that will get us to commercial nanotech in the early 2030s. The would seem to speed up the availability of nanotech, possibly more than we're prepared to handle.
Weighing the Tradeoffs
I have the skills, and possibly the time, to make a modest impact on Drexler's project.
Writing this post has left me confused about whether I should put significant effort into using MSEP.one. I'm tentatively leaning toward no. I welcome input from readers on this question.