The Limits of Artificial Intelligence

I'm very cautious about saying that there are limits to what AI can do, because I think this is a place where our human intuitions may fail. It does seem at the moment that large language models have some sort of quite fundamental flaws in the way that in the way they work. But it's very possible that we simply lack the imagination to understand when you have trillions of parameters and trillions and trillion and trillions of words and texts... What human-like intelligence can be produced from those very inhuman processes? And could a machine with access to every human text ever written reconstruct that observation? You'd pretend potentially yes.