Comment by Borealid

No bait-and-switch here. The discussion started with the idea that LLMs are helpful at improving LLMs. The goal of improving an LLM is to make the LLM more "intelligent" (whatever humans think that means). If they're helpful at that, they have to keep becoming more "intelligent".

I didn't present any evidence that more-intelligent means more-persuasive. If you don't think it does, then there is nothing to think about in the point I made, as no matter how "smart" the models get they will likely remain subject to human wills.