Gwern Branwen

We're making progress.

How is that possible?

And then coming up with the answer, well, progress must be possible now because civilization gets destroyed every couple of thousand years.

And all we're doing is reinventing and rediscovering the old stuff.

That was actually his explanation for technological acceleration.

We can't actually have any kind of real technological acceleration.

It must be because the world gets destroyed periodically and we just can't see past the last reset.

Yeah, yeah.

It turns out even Lucretius, around 1700 years before that, was writing the same argument.

He said, look at all these wonderful innovations and arts and sciences that we Romans have compiled together in the Roman Empire.

This is amazing.

but it can't actually be a recent acceleration technology.

Could that be real?

Could there be progress?

No, that's crazy.

Obviously the world was just recently destroyed.

It is, yeah.

So the 10,000-foot view of intelligence that I think the success of scaling points to is that all intelligence is is search over Turing machines.

And I think anything that happens can be described by Turing machines of various lengths.

And all that we're doing when we're doing learning or when we're doing scaling is that we're searching over more and longer Turing machines, and we're applying them in every specific case.