The way that I prefer to think about this is exactly the opposite: given that there is no “magic” involved because it is all just a big dumb tensor network underneath the hood, it is incredibly remarkable how smart it can be in practice.

There is genuine progress in admitting we don’t know what the fuck we’re doing.

That’s not really what The Bitter Lesson says, but it’s not not what The Bitter Lesson says.

at least the bottom panel looks fun