International collaboration on ASI development has several benefits over current race dynamics competition.
People in decision-making positions should strongly consider endorsing it.
https://t.co/4mio1EuYRc
@liron This:
- Avoids race dynamics
- Avoids covert and rushed development
- Improves collaboration on alignment/safety
- Minimizes defection possibilities (if talent and compute are pooled and well tracked)
- Improves transparency
- Is more likely to avoid a "winner take all" scenario
@ozdalva@teortaxesTex Sure, but if he wastes money on sub-optimal ways of getting compute, he's giving others free advantage on that front, for what? Short term gains that disappear if it turns out orbital compute doesn't work? I don't think Elon is that dumb or short-sighted.
@ozdalva@teortaxesTex If compute matters, and putting them in space is the only way you get more compute, that's what you do.
If there are better ways, they will do those too, but apparently Elon thinks this is the most viable, we'll see. Again, I certainly haven't explored all possible ways.
@ozdalva@teortaxesTex Then they lose money, which they have plenty of.
But if they're right, they win everything.
Easy bet to make, unless you can think of something better, and by the looks of it, all major players are thinking the same way.
Also it's not like they're only investing in hardware.
@ozdalva@teortaxesTex Achieving AGI means winning everything, it's not a bet about ROI in the classical sense of actual economic return, it's basically infinite return. For that goal, any loss is acceptable, and I think that's exactly what the main players are doing now. It's not about money.
@ozdalva@teortaxesTex I don't think compute demand will plateau at all, the opposite.
Sure, it might not be as cheap as to build on Earth if there were no limitations on Earth, but since there are, it is a good way to increase compute. There might be better alternatives, but I haven't explored them.
Depends on how much it falls, if it's currently overvalued because of LLMs (I don't think it is, even if we don't reach ASI, LLMs are already incredibly valuable).
If AI demand falls drastically, then indeed something weird happened I think, because the way it's going, I only see demand increasing in the near and medium term, unknown what happens post-ASI.
@ozdalva@teortaxesTex I wrote limitations, not just cost.
Land and building permits, and power generation (which is not an issue in space with solar) are the main bottlenecks apparently, all solved in space.
Cost is irrelevant if you can't even build on Earth.
@joseph_h_garvin@LinusMixson@teortaxesTex@LeoVasanko Seems very unlikely, unless we get something far better for compute, but then he'll just use that in the orbital data centers.
If demand for compute in general falls, then something really strange happened.
@tim_tyler@RokoMijic@PhilipJohnston@ApoStructura I'm not sure a Moon base helps much with this. Maybe in the long term, if you build enough manufacturing there to offload launch costs from Earth.
@teortaxesTex@beffjezos That might be better than nothing (still not convinced), but not ideal to leave it up to "whoever wins the race (by a large-enough margin) gets to rule us".