There's somehow this belief that "newer models will disprove <insert LLM criticism here>" despite the "newer" models being... just a scaled-up version of a prev

There's somehow this belief that "newer models will disprove <insert LLM criticism here>" despite the "newer" models being... just a scaled-up version of a previous model, or some anciliary features tacked on. An LLM is an LLM is an LLM: I'll believe it when I see otherwise.