Comment by Retr0id
4 hours ago
> But now either the AI can handle it or it can pretend to handle it. Frankly it's pretending both times, but often it's enough to get the result we need.
This has been how I think about it, too. The success rates are going up, but I still view the AI as an adversary that is trying to trick me into thinking it's being useful. Often the act is good enough to be actually useful, too.
The first anthropomorphization of AI which is actually useful.
It's not even an anthropomorphization, the reward function in RLHF-like scenarios is usually quite literally "did the user think the output was good"