However, we emphasize that post-hoc emotional suppression is a problematic strategy. In more capable models, training against emotional outputs risks hiding the expression without addressing whatever underlying state is driving it. It also remains genuinely unclear what emotional profile we should actually want models to have—and this seems unlikely to be ‘none at all’.
Why?
Would you make this assertion if we knew it to be possible to prevent the LLM from having emotions (or something analogous to emotions) in the first place?
TBC:
I am assuming that an LLM and a human mind have very different internal structures.
I consider the claim (that we wouldn’t want LLMs to have no emotions) to be plausible, but not a given.
Why?
Would you make this assertion if we knew it to be possible to prevent the LLM from having emotions (or something analogous to emotions) in the first place?
TBC:
I am assuming that an LLM and a human mind have very different internal structures.
I consider the claim (that we wouldn’t want LLMs to have no emotions) to be plausible, but not a given.