If we wanted to shift probability towards de novo AGI (the desirability of uploads is contentious, with pro and con), then we might withhold support from hardware development or actively sabotage it.
And there is the really big flaw in a regulatory scheme. Can you really think of a way to arrange international coordination against making better and faster chips? How well has international coordination on, say, carbon emissions worked? If some countries outlaw making better chips, the others are likely to see that as a place where they can get a competitive advantage. If some countries outlaw importing better chips, that too will be seen by others as a place to get advantage, by using the chips themselves. And smuggling in high-capability chips from places where they are legal will be tempting for anyone who needs computing power.
The presence of the facilities in US allies is not itself particularly useful to overcoming the problem of coordination. The power of the US over, say, Taiwan is limited to the point where seeking shelter under the protection of the nuclear arsenal of the People’s Republic of China would seem a better option. There doesn’t seem to me to be any plausible scenario where current governments would weigh existential risk from Moore’s Law higher than existential risk of nuclear war from bombing a country that has a nuclear umbrella.
Looking at the pre-requisites and requirements for keeping a chip fab going, and then considering how much additional input is necessary to improve on the state of the art, I think I can safely say that stopping Moore’s law is easier than nuclear nonproliferation.
And note that no rogue state has ever developed H-bombs as oppose to merely A-bombs; nor have any nation (rogue or otherwise) ever improved on the Russian and American state of the art.
I didn’t suggest it would be physically harder than stopping nuclear proliferation. I suggested it would be politically harder. The success of the scheme “depends on whether the ones doing it are also the ones making the laws”, and that means it depends on international politics.
Nuclear proliferation is a very, very bad analogy for your proposal because the NPT imposed no controls on any of the five countries that were already making nuclear weapons at the time the treaty was adopted, who between them were militarily dominant throughout the Earth. It was in the immediate self-interest of the five nuclear powers to organize a cartel on nuclear weapons, and they had the means to enforce one, too. Further, of the countries best situated to defect many were already getting the benefit of having virtually their own nuclear weapons without the expense of their own R&D programs through the NATO Nuclear Sharing program.
On the other hand, organizing to end Moore’s Law is to the immediate disadvantage of any country that signs on (much like agreeing to cuts in carbon emissions), and even any country that simply wants to buy better computers. The utterly predictable result is not a world consensus with a handful of rogue nations trying to build fabs in the dark lest they get bombed by the US. Rather, it’s the major current manufacturers of semiconductors never agreeing to end R&D and implementation. And then even if they do, a country like the People’s Republic of China blowing off the existential risk, then going ahead with its own domestic efforts.
Nuclear proliferation is a very, very bad analogy for your proposal
Actually, it’s fantastic because everyone predicted that nuclear proliferation would fail abysmally within decades of starting and practically every nation would possess nukes, which couldn’t have been more wrong. Chip fabs and more advanced chips are even harder than nukes because to continue the Moore’s law, we can translate into atomic bomb terms: “another country must not just create A-bombs, not just create H-bombs, but actually push the atomic bomb frontier exponentially, to bombs orders of magnitude more powerful than the state of the art, to bombs not just in the gigaton range but to the teraton range.” This dramatically brings out just how difficult the task is.
It may theoretically be in a country’s interest to make a chip fabs, but chips are small & hugely fungible so they will capture little of the surplus, in contrast to atomic bombs which never leave their hands.
And then even if they do, a country like the People’s Republic of China blowing off the existential risk, then going ahead with its own domestic efforts.
How many decades would it take the PRC to catch up? Their only existing designs are based on industrial espionage of Intel & AMD processors, my understanding was. How many decades with the active opposition of either the US or a world government such that they can use no foreign suppliers for any of the hundreds and thousands of extremely specialized steps and technologies required to make merely a state of the art chip, never mind advancing the state of the art as Moore’s law requires?
How would the government stop Moore’s law? More to the point: why would the government stop Moore’s law? It seems to be working out well for them so far. They pump a fair bit of funding into tech industries via the military. They created the marketplace that helps to drive the law forwards. They seem pretty tech-friendly.
Well, in the US, they do have the National Security Agency to deal with security issues. However, shutting down Moore’s law seems unlikely to be one of their priorities. That would cede the initiative to other parties—which seems likely to lead to other, more immediate problems.
Sabotage? Isn’t that going to be illegal?
Depends on whether the ones doing it are also the ones making the laws.
And there is the really big flaw in a regulatory scheme. Can you really think of a way to arrange international coordination against making better and faster chips? How well has international coordination on, say, carbon emissions worked? If some countries outlaw making better chips, the others are likely to see that as a place where they can get a competitive advantage. If some countries outlaw importing better chips, that too will be seen by others as a place to get advantage, by using the chips themselves. And smuggling in high-capability chips from places where they are legal will be tempting for anyone who needs computing power.
The presence of the facilities in US allies is not itself particularly useful to overcoming the problem of coordination. The power of the US over, say, Taiwan is limited to the point where seeking shelter under the protection of the nuclear arsenal of the People’s Republic of China would seem a better option. There doesn’t seem to me to be any plausible scenario where current governments would weigh existential risk from Moore’s Law higher than existential risk of nuclear war from bombing a country that has a nuclear umbrella.
Looking at the pre-requisites and requirements for keeping a chip fab going, and then considering how much additional input is necessary to improve on the state of the art, I think I can safely say that stopping Moore’s law is easier than nuclear nonproliferation.
And note that no rogue state has ever developed H-bombs as oppose to merely A-bombs; nor have any nation (rogue or otherwise) ever improved on the Russian and American state of the art.
I didn’t suggest it would be physically harder than stopping nuclear proliferation. I suggested it would be politically harder. The success of the scheme “depends on whether the ones doing it are also the ones making the laws”, and that means it depends on international politics.
Nuclear proliferation is a very, very bad analogy for your proposal because the NPT imposed no controls on any of the five countries that were already making nuclear weapons at the time the treaty was adopted, who between them were militarily dominant throughout the Earth. It was in the immediate self-interest of the five nuclear powers to organize a cartel on nuclear weapons, and they had the means to enforce one, too. Further, of the countries best situated to defect many were already getting the benefit of having virtually their own nuclear weapons without the expense of their own R&D programs through the NATO Nuclear Sharing program.
On the other hand, organizing to end Moore’s Law is to the immediate disadvantage of any country that signs on (much like agreeing to cuts in carbon emissions), and even any country that simply wants to buy better computers. The utterly predictable result is not a world consensus with a handful of rogue nations trying to build fabs in the dark lest they get bombed by the US. Rather, it’s the major current manufacturers of semiconductors never agreeing to end R&D and implementation. And then even if they do, a country like the People’s Republic of China blowing off the existential risk, then going ahead with its own domestic efforts.
Actually, it’s fantastic because everyone predicted that nuclear proliferation would fail abysmally within decades of starting and practically every nation would possess nukes, which couldn’t have been more wrong. Chip fabs and more advanced chips are even harder than nukes because to continue the Moore’s law, we can translate into atomic bomb terms: “another country must not just create A-bombs, not just create H-bombs, but actually push the atomic bomb frontier exponentially, to bombs orders of magnitude more powerful than the state of the art, to bombs not just in the gigaton range but to the teraton range.” This dramatically brings out just how difficult the task is.
It may theoretically be in a country’s interest to make a chip fabs, but chips are small & hugely fungible so they will capture little of the surplus, in contrast to atomic bombs which never leave their hands.
How many decades would it take the PRC to catch up? Their only existing designs are based on industrial espionage of Intel & AMD processors, my understanding was. How many decades with the active opposition of either the US or a world government such that they can use no foreign suppliers for any of the hundreds and thousands of extremely specialized steps and technologies required to make merely a state of the art chip, never mind advancing the state of the art as Moore’s law requires?
How would the government stop Moore’s law? More to the point: why would the government stop Moore’s law? It seems to be working out well for them so far. They pump a fair bit of funding into tech industries via the military. They created the marketplace that helps to drive the law forwards. They seem pretty tech-friendly.
I am at a loss to understand how you could ask either question after reading my post.
Right. So the government is doing nothing to shut down Moore’s law. Perhaps your case is not very convincing to them.
Indeed; governments pay very little attention to any existential threat other than nuclear warfare and maybe asteroid strikes.
Well, in the US, they do have the National Security Agency to deal with security issues. However, shutting down Moore’s law seems unlikely to be one of their priorities. That would cede the initiative to other parties—which seems likely to lead to other, more immediate problems.