Lanier is right that the phrase "artificial intelligence" isn't very useful for understanding what the recent progress in generative models actually means. I

Lanier is right that the phrase "artificial intelligence" isn't very useful for understanding what the recent progress in generative models actually means. I think, though, that's not much more than saying that while we know what we mean by "artificial," we don't really have a shared definition of "intelligence" against which we're evaluating these capabilities. Different observers use the word to mean everything from the ability to reason, to sentience, to self-consciousness, and from the attributes of single entities (a single dog, or chimp or person) to those given to the entirety of human thought. We can't even agree on a self-referential definition of intelligence, let alone an a priori one, so there isn't much chance we'll agree on whether GPT-4 manifests intelligence.