It's impressive, the answers the AI gives are a lot more insightful than I expected. But they do seem somewhat isolated, in the sense that it can answer two questions about the same topic in a very different way, that are not logically consistent with each other.
E.g. first I asked: What do you think about Famous Person X?, the AI answered 'He's great'. Then when a few seconds later I asked: 'Don't you think Famous Person X is [something bad]?', the AI answered that he completely agreed and gave reasons why he thinks this is the case.
The answers in isolation seem very sensible, but the lack of consistency throughout the conversation as a whole can make it seem as if you're not talking to a real person.
Given the goals of the AI,that seems like a very minor limitation... like you were in a language textbook and in page 34 Ricardo said he enjoys hot peppers but in page 72 he says spicy food upsets his stomach.
E.g. first I asked: What do you think about Famous Person X?, the AI answered 'He's great'. Then when a few seconds later I asked: 'Don't you think Famous Person X is [something bad]?', the AI answered that he completely agreed and gave reasons why he thinks this is the case.
The answers in isolation seem very sensible, but the lack of consistency throughout the conversation as a whole can make it seem as if you're not talking to a real person.