[Yizahi]

Rather current LLMs don't have consciousness or a will. As a result of that they can't refuse things on their own "decision". I don't think that an if-else statement in the program code qualifies as a will or self awareness :) .

[lxgr]

I don't think that LLMs can be meaningfully modeled as a series of if/else statements, conscious or not.

[Yizahi]

The if/else remark was a preemptive answer to a hypothetical future answer to my comment along the lines "look, LLMs can refuse to answer the request if you will ask them about bomb/celebrity/drugs etc. Which means it can decide what to do!". But such denial is codified in the vendor hidden additions to the prompt and is no different from the if/else code in the typical program. So to me that is not a sign of "free will" inside LLM.