Does this mean that, like, Elephant A has a name for Elephant B, or that Elephant A has a name that Elephants B-Z all know? Both are amazing, to be sure, I just wasn’t clear on this.
Looks like there is some potential evidence for the latter but they failed to prove it.
Perhaps we shouldn't read too much into that failure: it's a miracle of science that they managed to prove *any* part of this.
I disagree strongly with Guardian's decision to describe this as "AI." That's a pretty bad label in general but in the recent discourse it usually refers to large transformer-based neural networks, and this is something totally different, though I don't know how computationally intensive.
Until recently they were calling this sort of thing "machine learning" (and a lot of people would say that's still too anthropomorphic). But it seems the dam has broken: LLMs are "AI" and I guess everything else in the world is too.