I broadly agree with this, but I think there are phase transitions that will accelerate the rate of technological growth, like true "Einstein's in a data center" and/or whole brain emulation (e.g. see Robin Hanson's argument on the latter).
> Indeed, rain-fed agriculture still provides virtually all of our food, as it has for millennia. It looks unlikely that vertical farming will change that.
I feel like we have discovered the real upgrade for this (CO2 electrolysis using solar or fusion electricity, some biochemical path to higher-order hydrocarbons and ultimately precision fermentation/cellular agriculture to make foods), but the required technology is far away from becoming cost competitive.
I think the issue is at least partially marketing. Some flavor chemicals are made artificially. And there is stuff like the artificial saturation of fat.
The issue is at least in part that people don't want to buy lab made food. And food standards laws aren't too keen either.
Yeah over time I've become more excited about precision fermentation and industrial enzymes to make high-value foods (and somewhat less excited by solar fuels per se). But true, the industry is having problems scaling right now, let's hope they figure something out!
You may like the tidbits on ag innovations I put at the end of this post:
> The way I think about innovation is Trying Stuff. The best way to determine if something works is to try it in the real world.
Humans often innovate under conditions where trying something is easier than doing things the theoretical way. There is a spectrum of potential innovation strategies. Evolution uses a Huge amount of trial and error with basically no intelligence. An ASI might use a huge amount of intelligence to be almost always right first time. And humans are somewhere in between.
> The implication is that brilliant theory can only get you so far.
You have observed that trying stuff, with minimal theory, does work, eventually. You have "deduced" that theory, with minimal trying stuff, doesn't work. This isn't a valid deduction.
Now is there some finite upper bound on the amount of invention that can be done per unit of trying stuff. Possibly. But is that bound anywhere near current human innovation? I doubt it.
The gap between the amount of stuff that humans need to try, and the amount of stuff that a supersmart ASI might need to try can be finite, but large and very significant.
> A data center full of Nobel Laureates would spend most of their time waiting for results, not making discoveries
In the limiting case of LOADS of intelligence just waiting for more data, you have an exhaustive list of all possible theories, each with a probability. You might not know which theory is true. But you know exactly which theory is the most likely. You know exactly what the cheapest experiment to distinguish between 2 theories is. And you know exactly how to interpret any possible result.
Humanity isn't close to this.
> This is true of all technology. We find an effective way to do something, tweak it a bit, and then kind of stop
Except that the long term history of humanity looks superexponential. Cavemen spent 1000s of years gradually refining stone axe tech. By the end of the stone age, there were no more large gains to be made in stone axe technology. There were huge gains to be made in doing something that isn't stone axes. And, due to a larger more educated population, better communications, better tools etc, we can progress through aircraft tech much faster than we progressed through stone axe tech.
> Things like cups and chairs and pens are already optimized. There is little need for further cup research.
Well firstly, I suspect it's physically possible to have nanotech that makes a chair (or a lot of much more valuable things) out of garbage in minutes.
There is also a constant trickle of little background innovations. Some bigger ones, like how cups and chairs are often plastic now. But also smaller stuff. If you ask an expert, they will probably tell you about all sorts of tiny tweaks, both in the items themselves and in the manufacturing process.
I won't be able to respond to this in full, but a few clarifications:
1. I agree that we could innovate faster (with AI, automation, etc.) and that innovating faster would be good.
2.
> You have observed that trying stuff, with minimal theory, does work, eventually. You have "deduced" that theory, with minimal trying stuff, doesn't work.
I am not making the second claim. Theory clearly contributes to the innovation process.
3. Some of your points about background innovations and elsewhere seem consistent with what I'm saying about stacking many s-curves. The fact that innovation jumps around so much is a symptom of diminishing returns in a particular domain.
4. My core point is in the limit of lots of innovation (via AI, robots, brains in datacenters, testing every possibility etc) is convergence on a right way to produce goods and services. We're stuck with the rules physics and economics and biology have laid out.
This point is orthogonal to all the ways we might use AI or do more theory or automate science. Those things change the rate at which we converge, not what we converge to (unless we change something fundamental about our circumstances).
> We're stuck with the rules physics and economics and biology have laid out.
Physics, sure. Biology, not really. DNA isn't special in the laws of physics. There are many many different ways to arrange atoms into self replicating machines. DNA based life is just what evolution came up with, one solution out of many.
> Those things change the rate at which we converge, not what we converge to
Sure. But this "convergence" point contains uploaded minds, drexler style nanotech, dyson spheres etc.
Invention isn't just trying stuff, nor is it just theory, it is its own thing that comes from finding useful new connections between things. It is a specialized talent that depends on having wide knowledge of disparate fields as well as gaps or problems to be solved and spotting just those connections that could fill the gap or solve the problem.
Invention is an art. For most things there is no one best way to do things any more than there is one best painting. Technology is an ecosystem with many niches which open more often than they disappear. Is there one final type of organism?
Every improvement in technology since the sharp rock originated in /individuals'/ insights and inventions, not market forces, teams, society, governments, corporations, investors or managers. Yet there is no reward at all for invention itself, no jobs, no career path for inventors, no division of labor that lets inventors invent.
So society does not get what it does not pay for, still less what it punishes. We punish inventors harshly, that's why there is an apparent lack of innovation.
As I wrote:
It's about the same penalty as a first-time DUI to file for a patent. Patents are around $10k, which only gives a license & duty to sue infringers, suing will cost about $1M - $2M per case, and has a poor win rate, at that. At least 98% of patents never earn out their filing fees, let alone attorney's fees.
The number of inventors that make a living at inventing is pretty close to zero, at least two orders of magnitude lower than the number of patent attorneys. There is no market for patents, let alone inventions.
Every invention is competing not only with every other method of doing the same thing that has ever been published, and specifically against the ways that market incumbents have sunk capital into, capital which will be stranded by innovation, but also against every potential way of accomplishing the task that might be conceived before the invention has a chance to earn out the investment required. The transaction costs for what sporadic patent deals do occur are far more than the inventor can expect to get. Any valuable patent will get reexamined until it's invalidated. [....]
And the worst thing is, even if you don't patent, but instead publish an invention, you effectively not only lose all rights, you taint the IP so that no one can seriously invest in it -- it's unpatentable (worse, the patent will probably issue then get invalidated after expensive litigation) and not a trade secret. The smart move for inventors is to keep valuable inventions secret, even if they can't develop them, so as not to spoil things for someone who could make a go of the idea later.
Which brings me to another problem, which is that a good inventor can come up with a dozen or more inventions per month, culled from hundreds of ideas, but the system has no real division of labor to allow inventors to just invent - no, to profit from any idea requires all the time and money investment and the traits and skills of a successful startup founder, which are mostly incompatible with such inventors' personality types. Even when such a rare Musk-monster is to be found, even when he isn't crushed by one adversary or another and manages to get access to capital on good terms -- there is only enough time in life to implement a very few innovations; spending the time on entrepreneurial / corporate / social BS instead of on the mind-play of invention costs that inventor -- and society -- nearly his whole output.
That comes far from exhausting the problems with the system, for instance the fact that universities usually claim rights to inventors' work, even for undergraduates who are paying the school. Businesses also regularly claim the right to legally take employees' patent IP without any compensation, even when not developed with company resources or at company direction. Inventors are either shut out of universities and large and tech companies, or they are exploited.
I broadly agree with this, but I think there are phase transitions that will accelerate the rate of technological growth, like true "Einstein's in a data center" and/or whole brain emulation (e.g. see Robin Hanson's argument on the latter).
> Indeed, rain-fed agriculture still provides virtually all of our food, as it has for millennia. It looks unlikely that vertical farming will change that.
I feel like we have discovered the real upgrade for this (CO2 electrolysis using solar or fusion electricity, some biochemical path to higher-order hydrocarbons and ultimately precision fermentation/cellular agriculture to make foods), but the required technology is far away from becoming cost competitive.
I think the issue is at least partially marketing. Some flavor chemicals are made artificially. And there is stuff like the artificial saturation of fat.
The issue is at least in part that people don't want to buy lab made food. And food standards laws aren't too keen either.
Yeah over time I've become more excited about precision fermentation and industrial enzymes to make high-value foods (and somewhat less excited by solar fuels per se). But true, the industry is having problems scaling right now, let's hope they figure something out!
You may like the tidbits on ag innovations I put at the end of this post:
https://splittinginfinity.substack.com/i/166475536/other-ways-to-reduce-ag-emissions
Some other thoughts on precision fermentation:
https://splittinginfinity.substack.com/p/unlocking-precision-fermentation
Biosphere has a cool approach in this space:
https://www.biosphere.io/
> The way I think about innovation is Trying Stuff. The best way to determine if something works is to try it in the real world.
Humans often innovate under conditions where trying something is easier than doing things the theoretical way. There is a spectrum of potential innovation strategies. Evolution uses a Huge amount of trial and error with basically no intelligence. An ASI might use a huge amount of intelligence to be almost always right first time. And humans are somewhere in between.
> The implication is that brilliant theory can only get you so far.
You have observed that trying stuff, with minimal theory, does work, eventually. You have "deduced" that theory, with minimal trying stuff, doesn't work. This isn't a valid deduction.
Now is there some finite upper bound on the amount of invention that can be done per unit of trying stuff. Possibly. But is that bound anywhere near current human innovation? I doubt it.
The gap between the amount of stuff that humans need to try, and the amount of stuff that a supersmart ASI might need to try can be finite, but large and very significant.
> A data center full of Nobel Laureates would spend most of their time waiting for results, not making discoveries
In the limiting case of LOADS of intelligence just waiting for more data, you have an exhaustive list of all possible theories, each with a probability. You might not know which theory is true. But you know exactly which theory is the most likely. You know exactly what the cheapest experiment to distinguish between 2 theories is. And you know exactly how to interpret any possible result.
Humanity isn't close to this.
> This is true of all technology. We find an effective way to do something, tweak it a bit, and then kind of stop
Except that the long term history of humanity looks superexponential. Cavemen spent 1000s of years gradually refining stone axe tech. By the end of the stone age, there were no more large gains to be made in stone axe technology. There were huge gains to be made in doing something that isn't stone axes. And, due to a larger more educated population, better communications, better tools etc, we can progress through aircraft tech much faster than we progressed through stone axe tech.
> Things like cups and chairs and pens are already optimized. There is little need for further cup research.
Well firstly, I suspect it's physically possible to have nanotech that makes a chair (or a lot of much more valuable things) out of garbage in minutes.
There is also a constant trickle of little background innovations. Some bigger ones, like how cups and chairs are often plastic now. But also smaller stuff. If you ask an expert, they will probably tell you about all sorts of tiny tweaks, both in the items themselves and in the manufacturing process.
I won't be able to respond to this in full, but a few clarifications:
1. I agree that we could innovate faster (with AI, automation, etc.) and that innovating faster would be good.
2.
> You have observed that trying stuff, with minimal theory, does work, eventually. You have "deduced" that theory, with minimal trying stuff, doesn't work.
I am not making the second claim. Theory clearly contributes to the innovation process.
3. Some of your points about background innovations and elsewhere seem consistent with what I'm saying about stacking many s-curves. The fact that innovation jumps around so much is a symptom of diminishing returns in a particular domain.
4. My core point is in the limit of lots of innovation (via AI, robots, brains in datacenters, testing every possibility etc) is convergence on a right way to produce goods and services. We're stuck with the rules physics and economics and biology have laid out.
This point is orthogonal to all the ways we might use AI or do more theory or automate science. Those things change the rate at which we converge, not what we converge to (unless we change something fundamental about our circumstances).
> We're stuck with the rules physics and economics and biology have laid out.
Physics, sure. Biology, not really. DNA isn't special in the laws of physics. There are many many different ways to arrange atoms into self replicating machines. DNA based life is just what evolution came up with, one solution out of many.
> Those things change the rate at which we converge, not what we converge to
Sure. But this "convergence" point contains uploaded minds, drexler style nanotech, dyson spheres etc.
Invention isn't just trying stuff, nor is it just theory, it is its own thing that comes from finding useful new connections between things. It is a specialized talent that depends on having wide knowledge of disparate fields as well as gaps or problems to be solved and spotting just those connections that could fill the gap or solve the problem.
Invention is an art. For most things there is no one best way to do things any more than there is one best painting. Technology is an ecosystem with many niches which open more often than they disappear. Is there one final type of organism?
Every improvement in technology since the sharp rock originated in /individuals'/ insights and inventions, not market forces, teams, society, governments, corporations, investors or managers. Yet there is no reward at all for invention itself, no jobs, no career path for inventors, no division of labor that lets inventors invent.
So society does not get what it does not pay for, still less what it punishes. We punish inventors harshly, that's why there is an apparent lack of innovation.
As I wrote:
It's about the same penalty as a first-time DUI to file for a patent. Patents are around $10k, which only gives a license & duty to sue infringers, suing will cost about $1M - $2M per case, and has a poor win rate, at that. At least 98% of patents never earn out their filing fees, let alone attorney's fees.
The number of inventors that make a living at inventing is pretty close to zero, at least two orders of magnitude lower than the number of patent attorneys. There is no market for patents, let alone inventions.
Every invention is competing not only with every other method of doing the same thing that has ever been published, and specifically against the ways that market incumbents have sunk capital into, capital which will be stranded by innovation, but also against every potential way of accomplishing the task that might be conceived before the invention has a chance to earn out the investment required. The transaction costs for what sporadic patent deals do occur are far more than the inventor can expect to get. Any valuable patent will get reexamined until it's invalidated. [....]
And the worst thing is, even if you don't patent, but instead publish an invention, you effectively not only lose all rights, you taint the IP so that no one can seriously invest in it -- it's unpatentable (worse, the patent will probably issue then get invalidated after expensive litigation) and not a trade secret. The smart move for inventors is to keep valuable inventions secret, even if they can't develop them, so as not to spoil things for someone who could make a go of the idea later.
Which brings me to another problem, which is that a good inventor can come up with a dozen or more inventions per month, culled from hundreds of ideas, but the system has no real division of labor to allow inventors to just invent - no, to profit from any idea requires all the time and money investment and the traits and skills of a successful startup founder, which are mostly incompatible with such inventors' personality types. Even when such a rare Musk-monster is to be found, even when he isn't crushed by one adversary or another and manages to get access to capital on good terms -- there is only enough time in life to implement a very few innovations; spending the time on entrepreneurial / corporate / social BS instead of on the mind-play of invention costs that inventor -- and society -- nearly his whole output.
That comes far from exhausting the problems with the system, for instance the fact that universities usually claim rights to inventors' work, even for undergraduates who are paying the school. Businesses also regularly claim the right to legally take employees' patent IP without any compensation, even when not developed with company resources or at company direction. Inventors are either shut out of universities and large and tech companies, or they are exploited.
[From: https://enonh.substack.com/p/nobody-wants-geniuses]