#well maybe the agis can solve that one for us too if they don't turn us all into paperclips week 1

[not-terezi-pyrope]

Do I think language models will replace computer programmers? I think that even the most cutting edge models (looking at you O1) still lack the capacity for longitudinal planning and systems design, meaning that they are far below competence to automate away even a fairly low skill level human programmer.

However the way things are going is that all this means is that I'll be very surprised for a week when in three years OpenAI drops some sort of proto-AGI that bests all of us.

I would be more scared of this prospect, if the implications of that scenario weren't so far reaching that it's a category error to even start worrying about consequences before knowing specifics

(Also I find programmers generally tend to be more "shrug and bear it" about automation threat, because even when it's our own jobs on the line, it's kinda hard to be a professional automator who is ideologically opposed to automation).