That's still a lot more effort, especially since you have to vet each one to figure out if they provide good customer service in the event something goes wrong, and to be confident they won't steal and sell your credit card number (yeah, you aren't liable for the fraud, but getting a new card is a huge PITA). What could make this work well is the existence of a few online shopping aggregators that combine searching across all of the online stores and centralize payment. The problem is that in order to compete with Amazon any such alternatives would have to have enormous scale, which makes it a very difficult space to enter. Google tried with Google Shopping, but regulators immediately jumped in to stop them.

FWIW, my strategy is that for inexpensive stuff I just buy on Amazon, period, spending a little time to look for cheaper/better options than the "Amazon recommended". For pricier stuff, where it's worth spending a few minutes, I search on Amazon and also on Google, and if I find cheaper non-Amazon options I spend some time evaluating the different sites, unless they happen to be sites I've already bought from. For really expensive stuff I use other search engines and recommendation sites... and then almost always end up buying on Amazon because on those products pricing tends to be consistent, and it's a lot of money and if something goes wrong I trust Amazon to make me whole

Yep. So what? All species are evolved to fight for survival, because any that doesn't evolve to fight for survival is likely to cease to exist. I'm human and want my species to survive. Should I instead want my species to be eaten by wolves, or ASIs?

The problem is that there is a portion of our species that is not interested in humanity's survival. Those people are an existential threat to the rest of us. That doesn't mean we need to exterminate them, but it does suggest that we shouldn't help them carry out their plans.

The problem with the "trick" is that we (a) don't know how to set goals or "wants" for the AI systems we build, nor do we (b) know what goals or wants we could or should safely set if we did know how to set them.

The combination of (a) and (b) is what's known in the AI world as the Alignment Problem (i.e. aligning AI interests with human interests), and it's completely unsolved.

I think you also need to prove that humans supplanting other less-intelligent species is good. Maybe the universe would be better off if we hadn't dominated the Earth and killed off so many species.

(Note that I think both arguments are silly. I'm just pointing out that if you're asking for proof that AI is better than humanity, you should also be asking for proof that humanity is better than non-humanity, whether AI or not. My own take is that humanity, like every other species, selfishly fights for its own survival. There's no morality in it, there's no such thing as making the universe better or worse off.)

Seizing land is a counterproductive and foolish solution to that problem. Basically the whole world uses a different solution, which works pretty well: property taxes (though land-value taxes would probably be better). You just keep raising the taxes until leaving land idle becomes a money-losing proposition. The only way that doesn't work is if ownership of farmland is truly monopoly-dominated so there is no competition, in which case you might have to resort to trust-busting.

This is exactly why we have property taxes, to ensure that most property is put to productive use.

Yes, mass starvation is worse than land seizure, but land seizure is just about the worst possible solution to the problem, as evidenced by what has happened to Venezuela's economy since then. Seizure and collective ownership is guaranteed to produce horribly inefficient operations which might prevent outright starvation but will leave the populace on the edge of it. Seizure and redistribution to private ownership is slightly less bad, but will redistribute the land mostly to people who don't know how to use it effectively.

What would have worked much, much better would be actions that served to restore competition among farmers, starting with making sure they were all paying fair property taxes that were high enough to disincentivize leaving farmland fallow.

Why would superintelligent AIs obey the billionaires?

If you think it's because they'd be programmed to to it, you don't understand how we currently design and build AI. We don't program it to do anything. We train it until it responds the way we want it to, but we have no way of knowing if it's just fooling us. We can't actually define goals for the systems and we can't introspect them to tell what actual goals they have derived from their training sets.

Note, BTW, that the above is only one half of the problem called "AI alignment". In order to make sure AI will serve humanity (or a small segment of humanity; it's exactly the same problem either way) you need to be able to do two things. First, you need to be able to set the AI's goals, in a way that sticks. Second, you need to figure out what goal you can set that will achieve the subservience that you want. The difficulty in setting a "safe" goal for a powerful being is well illustrated in that old tales about genies and wishes, but modern philosophers have taken a hard, systematic look at this problem and so far no one has come up with any safe goal, not one, there's always some way it could go horribly wrong.

A laser pointer is hell on camera.

Okay... those things hardly seem like "participating in Project 2025", though there has been an unfortunate amount of capitulation to the administration. Personally, I think the worst of it was the way they settled Trump's lawsuit.

What are those?

It really doesn't do anything to illuminate the situation. It would if the ISS were built to generate/consume a lot of power, but it's not. Quite the opposite. It's designed to provide a reasonable amount of power to run life support and energy to run some experiments.

It may, of course, be true that orbital gigawatt data centers are a lot more than twenty years away, but the comparison with ISS doesn't tell us anything.

The whole point of AI is that it's supposed to be able to adapt to us, allowing us to give it direction in natural language and expect it to deal correctly with our ambiguities. While it's true that current-generation AI does require a learning curve, it's improving very rapidly, so any thing you learn about how to use it today will be obsolete next year. "Prompt engineering" shouldn't ultimately be a thing at all, and if AI development stalls out at some point so that it actually is a thing people have to do a decade from now, it will not be what it is today.

It makes sense to learn how to work around the idiosyncrasies and limitations of today's AI tools if you can use them to accomplish useful work today, but there's no point in learning those things in order to use the tools of 2035.

Actually, I retract my earlier statement. I had read something about Tinker earlier today on another site, but crossreferencing, it appears to have been wrong.

This product is a 'me too' since there's already tools to help you fine tune various Big Player models.

That's not what this is. This is a tool for creating foundation models.

Would you prefer the word "proprietary"? Because AFAIK that's all he means by that.

Given that the comparison is with a Small Modular Reactor, this isn't really relevant. If we end up with GW data centers then the comparison result may change, since it will be comparing a full-sized reactor against much larger renewable plants. I doubt the results will be much different, but they might be.