I mean, see the outcry when OpenAI briefly nuked GPT-4o in ChatGPT; people acted as if OpenAI had killed their friend. This is of course all deeply concerning, but it does seem likely that the personified LLM is a more compelling product, and more likely to encourage dependence/addiction.

▲

skohan 6 days ago | parent [-]

I wonder to what extent the 4o rollback was motivated by this exact case

	▲	rsynnott 6 days ago \| parent [-]
		As in the removal of 4o, or its reinstatement? Like, the model involved here was 4o AFAICS; if it was related to this case you'd expect them to remove it and bury it, not remove it and return it a few days after.