We owe pretty much everything that we are and have to innovation. That is, to our ancestors’ efforts (intentional or not) to improve their behaviors. But the rate of innovation has not been remotely constant over time. And we can credit increases in the rate of innovation to:
I commented on this post before it was ported to the Substack with my suggestion what the next meta innovation could be. I found the corresponding post on Wayback Machine:
Unfortunately, the Disqus there doesn't load and it seems all comments are gone.
I probably posted as Gunnar Zarncke. Robin, if you have access to the old Disqus threads, can you post my comment? I want to use it for a reply to Zvi's Response to Quintin Pope's Evolution Provides No Evidence For the Sharp Left Turn
"…all four seems to be due to better ways to diffuse, as opposed to create, innovations. Lump 1 was clearly the introduction of natural selection, where biological reproduction spreads innovations. Lump 2 seems somewhat clearly cultural evolution, wherein we learned enough how to copy the better observed behaviors of others."
I am not so sure about your emphasis on diffusion over creation of innovation. Natural selection works by generating minor variation in the copying process and then having differential selection. This is both a system of innovation discovery as well as propagation. They are tied together. As a side note, sex is a breakthrough in innovation (via recombination) and seems to have emerged around the time of nucleated cells and to be extremely important (though not essential) to multicellularity.
Similarly, cultural diffusion is creative in the act of being propagated. Minor variations are introduced in the act of copying others, and a new type of selection is created via the various choices we make in what to copy or preserve.
My point is that I would not put so much emphasis on diffusion breakthroughs over innovation breakthroughs. Indeed, I would actually add a third dimension to breakthroughs. Namely, breakthroughs in coordination or organization. I think the longer history of the universe can be modeled using a conceptual framework of meta breakthroughs in innovation, propagation and coordination.
Examples of the latter include cells, nucleated cells, multicellularity, social colonies (like ants), forager bands, villages, states, empires, global market networks and global scientific communities.
If the industrial revolution was perhaps enabled by Gutenberg's printing press, what's interesting is that it took ~300 years to pay off in terms of observable economic growth (disclaimer: I don't know this history well). Likewise, maybe the technology that enables the next meta-innovation lump already exists but just hasn't paid off yet in terms of economic growth.
Put that way, there's an obvious candidate: the internet. Intuitively, arXiv, GitHub, blogs, etc. seem like a really big deal. Maybe they just haven't yet spawned their equivalent of the steam engine, whatever that turns out to be.
you make a copy of a worker and put them in the second factory at zero marginal cost, this is a cheaper way to do knowledge diffusion than before copying workers cheaply was possible.
1) Knowledge can be transmitted at all - via genes.2) Knowledge can be transmitted culturally - via language (and thus faster than reproduction cycles).3) Knowledge can be transmitted persistently - via artifacts and institutions (and thus grows with population).4) Knowledge can be generated systematically and with low error (and thus reliably building on top of each other becomes possible).
My guess for the next stage is
5) Knowledge can be transmitted systematically and without loss.
Right now, copying knowledge is inefficient. Data can be copied efficiently but each human has to learn to apply it again. The process starts with birth again. But there is no way to copy humans *with* their knowledge and the ability to apply it. I think ems are such a way, but large ML models that can act on the world - AGI - would also count: These can be copied arbitrarily too. We see precursors for this: Programs - still implemented by humans but running without us can be copied too. It is a bit like stage 3 where some parts of the transmission process were automated too.
Interesting claim that all previous meta-innovations sped up the rate of diffusion of innovations - I can see how this is true of life & culture, but how is this true of farming, or of the industrial revolution? Both seem to me to be object-level innovations in energy production, not meta-level innovations in how innovations spread.
See "SLOW TUESDAY NIGHT" by R. A. Lafferty for a vision of a vastly accelerated human society. For instance:
"The panhandler was Basil Bagelbaker, who would be the richest man in the world within an hour and a half. He would make and lose four fortunes within eight hours; and these not the little fortunes that ordinary men acquire, but titanic things."
Say two different factories simultaneously figure out a slightly different better way of doing something. How do they spread both of those innovations?
I'm confused by this reply. If you can copy ems (or computer based AI), why is this prohibited from counting as "diffusing innovations"? You set up a factory in one place, you can literally make an exact copy of all the workers and set up a second factory with equal output (so there's only capital costs of duplicating equipment, computer chips).
Can you give an example of diffusing innovations where this would NOT count?
My guess would be you think "diffusing innovations" perhaps requires hybridizing the new innovation with a different knowledge base, so the new knowledge could feed into more knowledge. Is that right?
"Faster diffusion means faster diffusion of false positives as much as true positives." That has always been true. Even so, we've seen huge increases in overall diffusion rates.
Faster diffusion means faster diffusion of false positives as much as true positives. We're already at the limits of how fast we can handle the information that is diffusing.
A lot of computation can be packed into more "selective" diffusion, which often means less rather than more diffusion. So it's not clear that we could do a _lot_ better on diffusion rates.
In biology, cancers "diffuse" the fastest, but they're clearly not the highest form of life. Viruses and bacteria also have the capability to double their "empire" in relatively short time frames, but ascending to become incorporeal replicating parasites (on a substrate of Ems?) hardly sounds like a glorious future to aspire to.
I commented on this post before it was ported to the Substack with my suggestion what the next meta innovation could be. I found the corresponding post on Wayback Machine:
https://www.overcomingbias.com/2022/10/what-will-be-fifth-meta-innovation.html
Unfortunately, the Disqus there doesn't load and it seems all comments are gone.
I probably posted as Gunnar Zarncke. Robin, if you have access to the old Disqus threads, can you post my comment? I want to use it for a reply to Zvi's Response to Quintin Pope's Evolution Provides No Evidence For the Sharp Left Turn
https://www.lesswrong.com/posts/Wr7N9ji36EvvvrqJK/response-to-quintin-pope-s-evolution-provides-no-evidence
"…all four seems to be due to better ways to diffuse, as opposed to create, innovations. Lump 1 was clearly the introduction of natural selection, where biological reproduction spreads innovations. Lump 2 seems somewhat clearly cultural evolution, wherein we learned enough how to copy the better observed behaviors of others."
I am not so sure about your emphasis on diffusion over creation of innovation. Natural selection works by generating minor variation in the copying process and then having differential selection. This is both a system of innovation discovery as well as propagation. They are tied together. As a side note, sex is a breakthrough in innovation (via recombination) and seems to have emerged around the time of nucleated cells and to be extremely important (though not essential) to multicellularity.
Similarly, cultural diffusion is creative in the act of being propagated. Minor variations are introduced in the act of copying others, and a new type of selection is created via the various choices we make in what to copy or preserve.
My point is that I would not put so much emphasis on diffusion breakthroughs over innovation breakthroughs. Indeed, I would actually add a third dimension to breakthroughs. Namely, breakthroughs in coordination or organization. I think the longer history of the universe can be modeled using a conceptual framework of meta breakthroughs in innovation, propagation and coordination.
Examples of the latter include cells, nucleated cells, multicellularity, social colonies (like ants), forager bands, villages, states, empires, global market networks and global scientific communities.
awesome summary and good guess. Can you recommend some good books about ems? I recently enjoyed the Bobiverse and plan to read The Age of Em.
If the industrial revolution was perhaps enabled by Gutenberg's printing press, what's interesting is that it took ~300 years to pay off in terms of observable economic growth (disclaimer: I don't know this history well). Likewise, maybe the technology that enables the next meta-innovation lump already exists but just hasn't paid off yet in terms of economic growth.
Put that way, there's an obvious candidate: the internet. Intuitively, arXiv, GitHub, blogs, etc. seem like a really big deal. Maybe they just haven't yet spawned their equivalent of the steam engine, whatever that turns out to be.
(I see you mention similar ideas in 6a.)
you make a copy of a worker and put them in the second factory at zero marginal cost, this is a cheaper way to do knowledge diffusion than before copying workers cheaply was possible.
Let me see if I can break this down o
1) Knowledge can be transmitted at all - via genes.2) Knowledge can be transmitted culturally - via language (and thus faster than reproduction cycles).3) Knowledge can be transmitted persistently - via artifacts and institutions (and thus grows with population).4) Knowledge can be generated systematically and with low error (and thus reliably building on top of each other becomes possible).
My guess for the next stage is
5) Knowledge can be transmitted systematically and without loss.
Right now, copying knowledge is inefficient. Data can be copied efficiently but each human has to learn to apply it again. The process starts with birth again. But there is no way to copy humans *with* their knowledge and the ability to apply it. I think ems are such a way, but large ML models that can act on the world - AGI - would also count: These can be copied arbitrarily too. We see precursors for this: Programs - still implemented by humans but running without us can be copied too. It is a bit like stage 3 where some parts of the transmission process were automated too.
Interesting claim that all previous meta-innovations sped up the rate of diffusion of innovations - I can see how this is true of life & culture, but how is this true of farming, or of the industrial revolution? Both seem to me to be object-level innovations in energy production, not meta-level innovations in how innovations spread.
See "SLOW TUESDAY NIGHT" by R. A. Lafferty for a vision of a vastly accelerated human society. For instance:
"The panhandler was Basil Bagelbaker, who would be the richest man in the world within an hour and a half. He would make and lose four fortunes within eight hours; and these not the little fortunes that ordinary men acquire, but titanic things."
Say two different factories simultaneously figure out a slightly different better way of doing something. How do they spread both of those innovations?
I'm confused by this reply. If you can copy ems (or computer based AI), why is this prohibited from counting as "diffusing innovations"? You set up a factory in one place, you can literally make an exact copy of all the workers and set up a second factory with equal output (so there's only capital costs of duplicating equipment, computer chips).
Can you give an example of diffusing innovations where this would NOT count?
My guess would be you think "diffusing innovations" perhaps requires hybridizing the new innovation with a different knowledge base, so the new knowledge could feed into more knowledge. Is that right?
If you call it innovation when GDP rises, what do you call it when GDP drops?
Also, less out of skepticism than curiosity, do you have a reference for the assertion that GDP "is generally an acceptable measure for this purpose"?
"Faster diffusion means faster diffusion of false positives as much as true positives." That has always been true. Even so, we've seen huge increases in overall diffusion rates.
Faster diffusion means faster diffusion of false positives as much as true positives. We're already at the limits of how fast we can handle the information that is diffusing.
A lot of computation can be packed into more "selective" diffusion, which often means less rather than more diffusion. So it's not clear that we could do a _lot_ better on diffusion rates.
In biology, cancers "diffuse" the fastest, but they're clearly not the highest form of life. Viruses and bacteria also have the capability to double their "empire" in relatively short time frames, but ascending to become incorporeal replicating parasites (on a substrate of Ems?) hardly sounds like a glorious future to aspire to.
That is generally an acceptable measure for this purpose.
I may be wrong, but my sense is you've had a number of other ideas that got a very significant amount of traction, while this one not so much.
There are some plots of GWP out there. For example: https://www.openphilanthrop...