The Symbol Pushing of Understanding

I could say, maybe I could train the system up to give really good answers. But it's not doing it for the right reasons. You could have some like, you know, wiccup completely accurate Wikipedia exhaustive Wikipedia or something like that. That knows all the truth is and knows all the false things. So I don't think that this turning to truth conditions is yet an answer that we should be satisfied with. And then we should reframe the question or at least given us some rails. We've got something like, okay, if that's the kind of thing that looks like it's one of the marks of someone who understands. What is the kind of mechanism that we think