Well, arguably from the get go an orbital datacenter would be better. If launch costs were low enough I would say that as much industry as possible should be moved off planet, and we should make earth into a garden?
We're still doing all the mining and manufacturing on Earth, and there's so much empty land. The final product of self-powered datacenter is among the lowest priority of things to get away from us, and not an effective place to spend environmental mitigation money. And then when they re-enter they pollute pretty badly.
You came up with an extremely arbitrary choice. I pointed out that if you were in a position to make that choice, you have much better options somewhere in between.
Then you talked about the more general idea of moving all industry off the planet, and sure that's nice but this is not the industry you want to start with, it should be one of the last ones you worry about.
If someone's offering to build space datacenters for free, probably go ahead and take them up on the offer. But if you have any influence on the decisionmaking, don't push for space datacenters, just push for non-polluting datacenters. You'll get better results for cheaper.
You're still not getting it. You still have to ask permission from places. That's the point. In orbit, nobody is going to care. And you're using words like "should" - should has no place in this discussion. It's an "is-ought" thing. You're saying "they should behave this way" and I'm saying, "they're going to behave this other way."
You haven't reasoned through any of this, you're saying that they "should" do something because it will be cheaper/better/whatever. And that's not how it works - it was never how it worked, but it was less obvious in prior times.
There aren't really any better options. Your bottlenecks are power, land costs, technical stuff, and bureaucracy. In space you basically shortcut all of those problems except the technical ones, and (and this is unbelievably important too), it brings the costs of other things you're doing on orbit down. Sure you "could" build data centers in Arizona or whatever, but they're not going to because they'd have to ask permission, it'd be a permitting nightmare, etc.
These guys are also thinking about opportunity cost. So, say you can build AGI++ you just need more compute. What is the opportunity cost of building terrestrially if it takes you 4 years to build this with an extra year for legal requirements?
> You still have to ask permission from places. That's the point. In orbit, nobody is going to care.
Launches need permission too.
> And you're using words like "should" - should has no place in this discussion. It's an "is-ought" thing. You're saying "they should behave this way" and I'm saying, "they're going to behave this other way."
My "should" is a summary of the incentives if they acted reasonably. Without a crystal ball, don't be so confident they'll do something different.
And here's a version with no is or ought: Anyone whose goal is avoiding datacenter downsides, that could say "we forbid polluting and power grid draining datacenters here" but chooses to say "we forbid any datacenters here, go to space", is a moron. And anyone that doesn't have the power to say either is irrelevant.
> You haven't reasoned through any of this, you're saying that they "should" do something because it will be cheaper/better/whatever. And that's not how it works - it was never how it worked, but it was less obvious in prior times.
Companies almost always do what's cheaper. Your logic is incomprehensible here.
> There aren't really any better options. Your bottlenecks are power, land costs, technical stuff, and bureaucracy. In space you basically shortcut all of those problems except the technical ones
If the satellites self-powering counts as not buying electricity, then so does self-powering on land. Land in the middle of nowhere is very cheap, and so is the bureaucracy per acre when you're buying large chunks. It's less bureaucracy than rocket launches. And the technical stuff is a lot simpler on land than in space; suggesting you avoid technical stuff by using rockets is crazy.
> Sure you "could" build data centers in Arizona or whatever, but they're not going to because they'd have to ask permission, it'd be a permitting nightmare, etc.
A baseless claim.
When you're happy with the middle of nowhere, building on land only requires a handful of approved sites on the entire planet. That's very achievable, and achievable faster than it takes to set up the production lines.