This leads to the comment I've seen many mathematicians make that humans should refocus on understanding not just proofs and mechanics.

There’s a long history of revenge porn and other nonconsensual porn showing up there and being hard to remove.

The whole site had been declining badly too, and is loaded with bots. Still an okay place to get news in some areas but pretty meh and the comments are always echo chambers.

Never done it on something that huge but I’ve been burned on one or two kickstarter type things.

If you can’t pay and then take delivery it doesn’t exist.

Backward compatibility is less of an issue when fixing legacy stuff takes $25 in tokens.

But history has already shown us that manipulation of belief and religion is a more effective and cheaper means of control than technocratic methods. A giant technocratic surveillance dystopia is possible, but it costs a lot more than old fashioned dogmatism with enforcement through herd behavior.

So we smashed religion, a kind of anti-Butlerian Jihad we call the Enlightenment, but the human lust for power and control just followed us into the machine age. If we smash the machines, we'll just end up back where we were. It will follow us there too.

Of course you could say this was one of the sub-themes of Dune, since (reading comprehension test!) Paul is not a hero. He's a religious manipulator, though a bit of a tragic villain in the sense that he feels dragged into it by fate.

IRL I don't think humanity will ever escape this cycle until we can define what freedom even means and reduce it to first principles. Does having thinking machines do my thinking for me increase or decrease my freedom? Does having a religious orthodoxy make my life choices for me increase or decrease my freedom? What's the third alternative, and do we really need to smash either religion or technology to get it?

Gotta commit to the bit.

Our country is now a "bit," one that unfortunately kills people. This timeline sucks.

Impressive engineering but utterly impractical. There are semi-practical flying car designs around folding wings or quad rotors (basically a big car sized drone) and the reason we don’t see them has more to do with safety, traffic control, and energy cost. Aircraft, like rockets, tend to get more energy efficient with size, so tiny flying machines are power wasters.

Meanwhile a decent LLM that can do tool use and code and reason tends to weigh in at a minimum of 30B parameters, and you need over 100B to get good performance.

Do LLMs just contain more information or are diffusion models more efficient at compression?

I saw some work on language diffusion models a while back but I think they tend to have problems with reasoning versus transformer type models.

All these trends are happening everywhere to varying degrees with less developed places like much of Africa simply earlier on the curve but following the same trajectory.

So where does it end? I’m skeptical of population doomerism but shrinking population, like too rapid population growth, creates economic and social problems.

On the other hand if we are about to automate tons of stuff it means we need less humans to run things, and if medicine keeps getting better longer life and health spans somewhat offset lower birth rates.

The outer surface would probably be sprayed with something, maybe just reflective dust, to also keep it from subliming into the vacuum.

Extremely repairable, melts at human and equipment friendly temperatures, recyclable, doubles as an emergency water reservoir, and last but definitely not least a fantastic radiation shield.

Would also work for structures on Mars, the Moon, etc, but I’m a fan of the idea of settling space itself not another planet. Gravity wells are a trap.

I agree.

I might be wrong, of course, but I would argue that general intelligence -- meaning intelligence reducible to a single scalar 'g' -- does not exist. All intelligence is a hybrid ensemble and an adaptation to a particular niche or set of niches. I'm a reasonably smart human but I'd make a really dumb tree or dolphin.

To get started, look up the no free lunch theorem.

There's room for debate here though. I think our technological journey with AI is going to help answer some of these questions.

With the Rationalists, I get the sense that 'g' is a faith, a prior they won't examine.

It's like this:

I can write an unbreakable cryptographic system by inventing one, going through the motions of making up attacks and showing that those attacks are impossible, and then concluding my system is unbreakable. Anyone can design cryptography they themselves cannot break.

Ask any cryptographer how well that works.

This is pretty normal human behavior, but it's not rational.

It took us thousands upon thousands of years to arrive at this simple idea because it's a blow to our egos and to primate hierarchy and its pretensions of authority. If I am big monkey, I must be right.

It's like a secular IQ-Calvinist take on infallible religious authority.

An ingredient you're missing though is what I call "intellectual edgelording." All Rationalists don't do this, but it's endemic in the community and has helped give birth to things like neoreaction (monarchy is good actually), extreme "if anyone builds it everyone dies" AI doomerism, and has made them very open to things like "race science" and other "forbidden ideas THEY don't want you to believe!"

They'd never characterize it that way though. They launder it through "giving a fair hearing to ideas that are outside the normal Overton window" or similar framings. But if you look closely, the ideas they give that fair hearing to all have a certain kind of bent: edgy, elitist, or validating to their belief that intelligence can be reduced to a single scalar 'g' and this confers some kind of divine authority.

I think the AI doomerism might stem from this. If intelligence can be summarized by one scalar value than if we build something whose 'g' is on average much higher than ours, it will of course annihilate us. The other unexamined prior there is that intelligence is necessarily supremacy which is necessarily dominating or genocidal. Which leads back to why they fall so easily for fascist-adjacent and racist ideas.

The least charitable take I have on that line of thinking is: they at least secretly believe that lower-IQ people should all die, so they project that belief onto their hypothetical superintelligent AI. It would, of course, come to the same conclusion, right?