> Recently, the use of Large Language Models, relying on AI tools to write reviews came up in EM discussions. My opinion on it is that I try to judge the output

> Recently, the use of Large Language Models, relying on AI tools to write reviews came up in EM discussions. My opinion on it is that I try to judge the output and not how it was created. If the resulting text is solid feedback, then I’m happy for you to save a bit of time by using ChatGPT. If not, you need to work more on it, regardless of the tools used. I get the point that the author is trying to make, but frankly since LLMs are expert are producing output that looks right, I would be feeling uncomfortable giving feedback that was partially written by ChatGPT. Even if as the author you feel like it conveys what you meant, you are still introducing new words around your feedback that are not your own and that will likely lead you towards impersonal somewhat generic feedback. Not saying that's what would necessarily happen, but I'd rather get 4 real lines from my manager than a half-page worth that was seeded from those same 4 lines.