Yes, this is what you’d expect from the maths for LLMs, which is what we’re talking about here, of course. LLMs being a subtype of neural nets, which themselves are a subtype of AI.
Neural nets are functions of universal approximation. Very broadly, that means that to get arbitrarily close to the function being modelled, the number of neurons in the net tends to infinity.
Thus, when pundits declare LLMs will improve until they take over the world, there’s a whole bunch of us just rolling our eyes and pointing at the maths and laughing.
As I like to say: marketing is a psychopathy.