likewise, if we want to create legitimate artificial intelligence, we need to get a better grasp of quantum physics, first. these absurdly large ai data centres are built on top of 100,000+ chips running in parallel. your brain is more powerful because it uses quantum physics to manipulate millions of hydrogen atoms in the space of your skull. we don't need to go down some pseudoscientific penrosian path of woo and metaphysics. that's not what i mean. quantum physics is clearly wrong in an absolute sense, but it's also the most accurately false model we currently have to describe how the world operates at a small scale. your brain uses quantum mechanics because everything uses quantum mechanics. quantum mechanics is not the exception; newtonian physics is the exception. however, our basic systems of electronics are still built on the exception, which we know is wrong more generally, and not on the general rule. there is no way to scale this for serious efficiency, it needs to be redesigned to better utilize physics, by using evolution as a model.
100,000 human brains running in parallel would be real intelligence with minimal energy costs.
i like the idea of harnessing anti-matter for energy from orbit. but you eventually even run out of space in orbit, even if you can get the energy.
ai as we understand it is a dead end and doomed to stagnate as a glorified clippy with the cognitive sense of forrest gump. it's really just a search engine. regulations should be focused on human behaviour. this is just another tool for bad actors; it's not fundamentally different than existing tools and it never will be.