Re:Learning to Reason with LLMs
by p1necone on Thursday September 12, 2024 @05:10PM
(#41525622)
For basically all the existing data we have, efficiency improvements always result in more work, not less.
Humans never say "oh neat I can do thing with 10% of the effort now, guess I'll go watch tv for the rest of the week", they say "oh neat I can do thing with 10% of the effort now, I'm going to hire twice as many people and produce like 20x as much as I was before because there's so much less risk to scaling now."
I think there's enough unmet demand for software that efficiency increases from automation are going to be eaten up for a long time to come.
Humans never say "oh neat I can do thing with 10% of the effort now, guess I'll go watch tv for the rest of the week", they say "oh neat I can do thing with 10% of the effort now, I'm going to hire twice as many people and produce like 20x as much as I was before because there's so much less risk to scaling now."
I think there's enough unmet demand for software that efficiency increases from automation are going to be eaten up for a long time to come.