I appreciate this, thanks. Identifying anything like beliefs in an AI that has to be capable of “wearing many masks” definitely seems tricky—features that seem to play semantic roles might only do so relative to a certain prompt or prompting style, in a way that doesn’t really have an analogy in humans.
I appreciate this, thanks. Identifying anything like beliefs in an AI that has to be capable of “wearing many masks” definitely seems tricky—features that seem to play semantic roles might only do so relative to a certain prompt or prompting style, in a way that doesn’t really have an analogy in humans.