@lexpedite
@damienriehl
OK. What is the evidence to the contrary? Since OpenAI has told us zilch about GPT-4, any claims that we might want to make about future systems are pure speculation. I have been saying this for quite a while.
Here is a comment that captures the situation accurately, in my opinion:
a remarkable thing about LLMs is i can't find anyone, novice or expert, proponent or critic, who seems to have a particularly good intuition for how these things will develop going forward. i think part of the reason is that there are so many moving parts and none of those parts are either transparent or intuitive -- unintuitive statistics of large data, massive human annotation/tuning/iteration efforts, huge limitations in evaluation, etc. GPT-5? nothing burger or another leap, who knows
@ltmccarty @damienriehl They are getting better. That's the evidence. It's just a trend, not a guarantee, but so what?
The idea that they have plateaued verges on the delusional. Human beings have never gotten anything that right that quickly.
Jason Morris
L Thorne McCarty