Throughout their article they repeatedly show how GPT-4.0 is generally much more competent than GPT-3.5 at everything including science and they show it's not y

Throughout their article they repeatedly show how GPT-4.0 is generally much more competent than GPT-3.5 at everything including science and they show it's not yet to the level of a trained human scientist. Their conclusion is "Until true AI comes along, viewing GPT-4 as your partner in IA [intelligence assistance] can be mutually beneficial." I have a hard time understanding what they mean by "true AI" but I assume they mean one who can be respected on equal terms as a human Ph.D. researcher. To me that's a weird terminology. It reminds me of the gaming trope where 'everyone less skilled than me is a noob ("not yet true AI") and everyone more skilled than me has no life (maybe "too dangerous AI").'