And third is that it can find a bridging narrative, that is to say, because the narratives actually, this is principal component analysis. So you can actually see there are some narratives that are not quite connected to the other narratives. And so by moving the latent space, you can actually synthesize narratives, like fuse this void. Nobody actually said that. But if you’re going to construct a bridging narrative that brings people to something they can live with, even though not perfectly understand, like this is necessary, at least say that this may probably be relevant, right? Then the language model can synthesize that sort of bridge making statements.

Keyboard shortcuts

j previous speech k next speech