NO expert here.
However, I have to wonder why they call these LLMs, "AI". It's a neat trick, mathematically and statistically predicting "which word should come next.." But the training is on the Western Civ canon of literature and available texts. I can't see how that can ever lead to "consciousness" no matter how many NVIDEA chips are used.
I've had many long interactions regarding what I think are good topics to measure "thinking" and the AIs have done remarkably poorly at thinking, but remarkably good at regurgitating history. In all the cases I pursued, the main response has been justifying canned history. After long convos, I would then suggest a different analysis of the history they were spouting, and suddenly, they would agree with me - completely contradicting their previous half hour position. My conclusion was these things are GREAT at accessing historical content very thoroughly, and they can offer "standard analysis" that has been suggested by others, But in no case do I ever find them offering "NOVEL thinking."