Dwarkesh Patel says AI solving major open problems humans have not resolved registers as surprising during an intermediate period before full intellectual superiority

No, and this is a ridiculous argument. We will forever be above, alongside, and beneath all "technology" in terms of capability... as long as each technology is both:

1) more capable than humans at X task 2) under our control

Because every time (1) happens, it "beats" humans... but every time (2) happens afterwards... then humans get the power of (1) anyways... raising the bar for what "human" capability actually is

A dump-truck is SUPERHUMAN in its ability to carry rocks... but the human is in the driver's seat... so now the human can carry rocks.

An airplane is SUPERHUMAN in its ability to fly... but a human controls where it flies... so now humans can fly.

(repeat argument for every technology)

All technology *augments* and *empowers* humans as long as we're in *control*. So all these predictions about AI "solving an important open problem that humans couldn't" is either:

A) Offensive: because the human prompting the AI isn't getting credit for getting the AI to do the thing they needed (much less... the people behind the training data, algorithms, etc.) B) Nihilistic: Implies that (2) isn't going to happen... which is a political statement about power far more-so than a technological statement.

The real prediction of the future is something totally different: there'll be competition for who can use AI to do crazy things better than anyone else...

Think less Terminator... more like... Neo in the matrix...

And yes... some people will be very good/intuitive at AI.