ChatGPT and AI is just word predicting program. Thats all it is. People who think its sentient think so because AI repeats human conversations. So yes, AI sounds like human because it actually just repeats what some humans have said. Its based on plenty of human text, and then it repeats from human text what is most appropriate to repeat. So yes, when you talk to AI, you are actually talking to human in a sense that it repeats what humans said in given scenario.
However, ChatGPT 5 is said to change everything. I am not so sure about that, because I find local models somewhat better than ChatGPT. ChatGPT must be instructed every time I use it, and also is censored much more. So what is exactly going to change other than less false claims by ChatGPT?
I dont think false claims by AI can be solved other than manually removing all false claims from text it uses, and since AI uses internet, that is impossible.
Local models and AI notebooks can act according to scientific literature, but AI using internet will have a hard time telling what is true on internet, even if you limit it to only most correct sites. There is a point where some answers only exist on sites of very questionable accuracy, such as Quora, and AI in the past has quoted Quora as fact without knowing it is just random people giving answers. Same with reddit. Basically, unless you ban AI from using these sites, AI wont stop hallucinating because these sites are full of contradictive answers and many answers are retarded.