This is the first time I'm reading about this "em" emulated brain concept. I see a much more "near", pertinent issue in the tendency of today's corporate and financial systems to reward businesses and individuals who already treat the masses of living people as "biological androids". Like Huxley's Brave New World. You don't have to hold your breath for three-to-fifteen decades to profit from the economics of this. It's already here and booming.
Oh yeah, and the budget train-wreck isn't two decades out, either. It's already happened and they have been keeping things on life support ever since 2008.
It is interesting to consider how the industrial-era political ideologies would deal with the em-econ. Liberalism would be concerned with ensuring the rights and freedom of persons, likely struggling to handle the political and economical quirks of copyable persons. Conservatism would want to preserve implicate social knowledge in the face of the change (it would of course have serious problems with the whole transition). Socialism would want to get the means of production under control by the "right" group, whether that is copy cooperatives, some avant garde etc. None of them seems to have speed and flexibility as a core ideological goal - liberalism would of course claim it would achieve it through harnessing individual initiative and institution-building, while socialism might think it could achieve greater coordination. But they all seem to be seriously hamstrung by their devotion to various pre-existing social institutions and assumptions about the nature of persons, societies and means of production.
For decisionmaking, demarchy might become practical. Imagine randomly selecting ems and making a parliament copy of them, acting as a representative body (normal demarchy has problems with having to call citizens to a duty that interferes with their lives). This might be a group generating the values input into futarchy.
Thanks for pointing me towards that report. I haven't yet found the time to read it in detail, but based on my initial skim, it sounds like the issue of raw computing power may not be as large as I thought. I'm bothered, though, by the current lack of progress on modeling even simple nervous systems. One of my undergraduate professors had spent a good portion of of his life trying to figure out the nervous system of the nematode worm, and the problem turned out to be far more difficult than his team initially expected. They're still working on it. If three to fifteen decades is your estimate for time until human brain emulation, when do you think we can expect simple invertebrates to be worked out?
I have to agree with Wei Dai. Its seems reasonable that China and India could be fueled purely by catch-up.
First, there have been a host of other countries which were accelerating towards US per capita GDP and suddenly slowed down as they approached, from the USSR to Western Europe to Japan. Clearly the USSR had much worse policies than the US but that did not stop a significant amount of catch-up.
It might have seemed at one point that Germany and Japan had better policies and would finally be the ones to past the US. However, that didn't happen either.
Second, pure theory suggests something that looks a lot like what we are seeing - immense capital deepening and technology transfer.
If China were at the technological frontier in many industries then it would seem more reasonable that they would surpass the US. However, right now they are not. The fast growth is due to high rates of capital investment and using an increasing number of mature technologies.
True but that tiny amount of genetic code "executes" on a very powerful and complex "computer" the molecules and forces etc. of reality. I personally don't think we have to simulate quantum effects etc. but to discover the high-order effects that is necessary for intelligence we haven't even scratched the surface yet. Still intelligences seems to be the result of enormous inter-connectivity and complexity so you cannot simplify only increase efficiency.
So exa-flop computers will probably be needed to even begin the development and experiments, even mature efficient ems will be many peta-flops because of the fundamental complexity of intelligence.
You answer your own question, the "program" may be fractal but intelligence is not a program, it is an evolving system that executes that program. And the most simple fractal algorithm can consume all conceivable computing power very fast.
To be honest I don't fear Robin's emu-calypse in the short term (the coming decades he claim).1. He underestimates the complexity of intelligence, we will probaly need high exa-flop computers just to begin productive research and development.2. The most efficient em will need many petaflops if not exa-flops to function, because of the fundamental nature and complexity of intelligence.3. End of Moore's law in silicon after 2020, there is going to be a stagnation in computing power through the 20's as everybody flails about, and no nothing is ready to smoothly take over, investment has almost been exclusively in silicon to continue squeeze everything out of it. The incentive to seriously invest in the next stage will only appear when silicon hits its fundamental limits. So push back the computing power extrapolations by at least 10 years if not 20 or more.
Ok, but why do you think this is "probable"? Or did you not mean to imply that in the post? The outside view suggests that China and India won't be any different from the other countries that temporarily had high growth rates in the past as they played catchup.
Brain organization is fractal, and stored in a tiny amount of genetic code. So it is likely we could emulate the relevant parts without modeling it at the molecular level.
We already have petaflops computers, and even most pessimistic estimates assume exaflops will be achieved before 2040. That should be more than enough, assuming higher-level insights are made during that time period. It is that last part that is truly unpredictable.
Question about brain emulation feasibility: are there any reputable back-of-the-envelope estimates for how much computing power (and how much of an increase over current computing power) would be needed to run a brain emulation in real time? Based on my knowledge of neuroscience and computer science, I'm skeptical that such emulations will be available in the foreseeable future: definitely not three decades, even fifteen sounds like a wild guess. As I understand it, with current technology, we have trouble doing decent simulations of large molecules, the best computer simulations of what happens when a nuclear bomb goes off take forever to run, and workable computer simulations of even simple nervous systems don't exist, period. A human brain is more complex than any of these things, and should therefore be harder to emulate.
At the deepest level, I think the problem is that when the human brain was evolving, there was no selective pressure to make the lives of neuroscientists easy. It might turn out that the only way to make a brain emulator is to simulate every little bit of jury-rigging evolution has come up with, and that the kinds of logic circuits allowed by the laws of physics aren't up to the task of doing this in real time.
Surely "catchup" factors increase China's growth rate, but they aren't obviously the only relevant factors. If their growth doesn't slow much as they catch up closer that will suggest other factors dominate.
By the year 2150, the cost of digging falls below the value of more rapid connectivity to neighboring intelligences. A sub-field of economics is born and named termite colony optimization (tercoptics). In the year 2200, the most valuable real estate is the center of the planet and is owned by a colloidal intellect named 'Mark Zuckerberg'. Ultimately the earth achieves a fractal dimension of 4.2.
With slow low-bandwidth long-delay expensive humans comm today can be cheap enough that it may not matter where humans are. Ems can be cheap, fast, and high bandwidth enough that comm costs and delays greatly matter.
If interaction is virtual, however, isn't the "distance" between New York and Hong Kong trivial. In the same way that transit expands the area that can effectively function as a city, can't EMs use the whole world as their city, communicating with each other electronically?
Also, my thinking was that growth rates would speed up because you basically have replicable labor. You go from (K^.3)(L^7) to something like (K^.3)(M^.65)(L^.05). Even in a basic Solow framework you would shift to a much higher growth path.
However, if - as I would assume - EMs are involved in research and development and R&D drives technological progress, then suddenly you have much faster technological progress.
This, I thought, was the source of much faster growth. And it makes sense even if EM thinking time is the same as human time. What isn't immediately clear is why the EM social network would be faster, though TGGP seems to be saying that EMs would indeed think faster than humans.
I don't agree with Cheney that "deficits don't matter" (just ask Greece now or plenty of other dysfunctional countries before it), but thought Casey Mulligan and Robert Barro gave interesting arguments on why concern with it is misplaced.
This is the first time I'm reading about this "em" emulated brain concept. I see a much more "near", pertinent issue in the tendency of today's corporate and financial systems to reward businesses and individuals who already treat the masses of living people as "biological androids". Like Huxley's Brave New World. You don't have to hold your breath for three-to-fifteen decades to profit from the economics of this. It's already here and booming.
Oh yeah, and the budget train-wreck isn't two decades out, either. It's already happened and they have been keeping things on life support ever since 2008.
It is interesting to consider how the industrial-era political ideologies would deal with the em-econ. Liberalism would be concerned with ensuring the rights and freedom of persons, likely struggling to handle the political and economical quirks of copyable persons. Conservatism would want to preserve implicate social knowledge in the face of the change (it would of course have serious problems with the whole transition). Socialism would want to get the means of production under control by the "right" group, whether that is copy cooperatives, some avant garde etc. None of them seems to have speed and flexibility as a core ideological goal - liberalism would of course claim it would achieve it through harnessing individual initiative and institution-building, while socialism might think it could achieve greater coordination. But they all seem to be seriously hamstrung by their devotion to various pre-existing social institutions and assumptions about the nature of persons, societies and means of production.
For decisionmaking, demarchy might become practical. Imagine randomly selecting ems and making a parliament copy of them, acting as a representative body (normal demarchy has problems with having to call citizens to a duty that interferes with their lives). This might be a group generating the values input into futarchy.
> industrialization, farming, and, I think, language
No, just the first two. The age of language is quite unknown, so it is unclear whether its advent caused population growth.
Thanks for pointing me towards that report. I haven't yet found the time to read it in detail, but based on my initial skim, it sounds like the issue of raw computing power may not be as large as I thought. I'm bothered, though, by the current lack of progress on modeling even simple nervous systems. One of my undergraduate professors had spent a good portion of of his life trying to figure out the nervous system of the nematode worm, and the problem turned out to be far more difficult than his team initially expected. They're still working on it. If three to fifteen decades is your estimate for time until human brain emulation, when do you think we can expect simple invertebrates to be worked out?
From the point of view of institutions, do you think that the em-econ would benefit from the insulation afforded by charter cities?
I have to agree with Wei Dai. Its seems reasonable that China and India could be fueled purely by catch-up.
First, there have been a host of other countries which were accelerating towards US per capita GDP and suddenly slowed down as they approached, from the USSR to Western Europe to Japan. Clearly the USSR had much worse policies than the US but that did not stop a significant amount of catch-up.
It might have seemed at one point that Germany and Japan had better policies and would finally be the ones to past the US. However, that didn't happen either.
Second, pure theory suggests something that looks a lot like what we are seeing - immense capital deepening and technology transfer.
If China were at the technological frontier in many industries then it would seem more reasonable that they would surpass the US. However, right now they are not. The fast growth is due to high rates of capital investment and using an increasing number of mature technologies.
True but that tiny amount of genetic code "executes" on a very powerful and complex "computer" the molecules and forces etc. of reality. I personally don't think we have to simulate quantum effects etc. but to discover the high-order effects that is necessary for intelligence we haven't even scratched the surface yet. Still intelligences seems to be the result of enormous inter-connectivity and complexity so you cannot simplify only increase efficiency.
So exa-flop computers will probably be needed to even begin the development and experiments, even mature efficient ems will be many peta-flops because of the fundamental complexity of intelligence.
You answer your own question, the "program" may be fractal but intelligence is not a program, it is an evolving system that executes that program. And the most simple fractal algorithm can consume all conceivable computing power very fast.
To be honest I don't fear Robin's emu-calypse in the short term (the coming decades he claim).1. He underestimates the complexity of intelligence, we will probaly need high exa-flop computers just to begin productive research and development.2. The most efficient em will need many petaflops if not exa-flops to function, because of the fundamental nature and complexity of intelligence.3. End of Moore's law in silicon after 2020, there is going to be a stagnation in computing power through the 20's as everybody flails about, and no nothing is ready to smoothly take over, investment has almost been exclusively in silicon to continue squeeze everything out of it. The incentive to seriously invest in the next stage will only appear when silicon hits its fundamental limits. So push back the computing power extrapolations by at least 10 years if not 20 or more.
So after 2050 around 2070.
See the brain emulation roadmap.
Ok, but why do you think this is "probable"? Or did you not mean to imply that in the post? The outside view suggests that China and India won't be any different from the other countries that temporarily had high growth rates in the past as they played catchup.
Brain organization is fractal, and stored in a tiny amount of genetic code. So it is likely we could emulate the relevant parts without modeling it at the molecular level.
We already have petaflops computers, and even most pessimistic estimates assume exaflops will be achieved before 2040. That should be more than enough, assuming higher-level insights are made during that time period. It is that last part that is truly unpredictable.
Question about brain emulation feasibility: are there any reputable back-of-the-envelope estimates for how much computing power (and how much of an increase over current computing power) would be needed to run a brain emulation in real time? Based on my knowledge of neuroscience and computer science, I'm skeptical that such emulations will be available in the foreseeable future: definitely not three decades, even fifteen sounds like a wild guess. As I understand it, with current technology, we have trouble doing decent simulations of large molecules, the best computer simulations of what happens when a nuclear bomb goes off take forever to run, and workable computer simulations of even simple nervous systems don't exist, period. A human brain is more complex than any of these things, and should therefore be harder to emulate.
At the deepest level, I think the problem is that when the human brain was evolving, there was no selective pressure to make the lives of neuroscientists easy. It might turn out that the only way to make a brain emulator is to simulate every little bit of jury-rigging evolution has come up with, and that the kinds of logic circuits allowed by the laws of physics aren't up to the task of doing this in real time.
Surely "catchup" factors increase China's growth rate, but they aren't obviously the only relevant factors. If their growth doesn't slow much as they catch up closer that will suggest other factors dominate.
By the year 2150, the cost of digging falls below the value of more rapid connectivity to neighboring intelligences. A sub-field of economics is born and named termite colony optimization (tercoptics). In the year 2200, the most valuable real estate is the center of the planet and is owned by a colloidal intellect named 'Mark Zuckerberg'. Ultimately the earth achieves a fractal dimension of 4.2.
With slow low-bandwidth long-delay expensive humans comm today can be cheap enough that it may not matter where humans are. Ems can be cheap, fast, and high bandwidth enough that comm costs and delays greatly matter.
If interaction is virtual, however, isn't the "distance" between New York and Hong Kong trivial. In the same way that transit expands the area that can effectively function as a city, can't EMs use the whole world as their city, communicating with each other electronically?
Also, my thinking was that growth rates would speed up because you basically have replicable labor. You go from (K^.3)(L^7) to something like (K^.3)(M^.65)(L^.05). Even in a basic Solow framework you would shift to a much higher growth path.
However, if - as I would assume - EMs are involved in research and development and R&D drives technological progress, then suddenly you have much faster technological progress.
This, I thought, was the source of much faster growth. And it makes sense even if EM thinking time is the same as human time. What isn't immediately clear is why the EM social network would be faster, though TGGP seems to be saying that EMs would indeed think faster than humans.
I don't agree with Cheney that "deficits don't matter" (just ask Greece now or plenty of other dysfunctional countries before it), but thought Casey Mulligan and Robert Barro gave interesting arguments on why concern with it is misplaced.