I think there's still challenges.

The main one I see is the question of how we can use an LLM as humans to produce something harder than the LLM can by itself.

I like to think of it with an analogy. Suppose I was granted the privilege of having a world renowned human expert as my personal assistant, but she had no initiative of her own. How could I collaborate with her?

I would definitely need to keep learning more to understand what she says. I would also need tools to augment my ability to follow along and she would need tools to bring knowledge down to my level of understanding.

One might argue humans are unnecessary in the above process, but I think that's just failing the challenge of keeping up. So I try

▲

0c3ca83 an hour ago | parent [-]

We've largely reached the point where a mathematician in the loop is a net negative in theorem proving.

How long do you think other fields have?

▲

WhyIsItAlwaysHN an hour ago | parent [-]

That's what I'm saying though. It's a new challenge to make the mathematician a net positive again.

We can consider what makes a human a net negative in the loop and augment humans with tools to collaborate.

▲

0c3ca83 35 minutes ago | parent [-]

The processing capacity of the human brain makes the human a net negative.

▲

WhyIsItAlwaysHN 27 minutes ago | parent [-]

Well let's get some more capacity then. I don't see any good brain to computer interfaces developed yet.

	▲	0c3ca83 26 minutes ago \| parent [-]
		Gross. We're building dystopia, and the only "improvement" suggested is adding elements of hell. I don't know how to sustainably withdraw from society, but every day the desire gets stronger.