Thank you Naja.

To answer your question, no I do not think GPT5 nor GPT25 will solve these issues, since they are not a function of 'scale' nor a function of the underlying neural architecture, but a function of the paradigm itself. In short, DNNs do not have the capabilities of making reasoning at the abstract level. Full "understanding" requires symbolic knowledge that is not available in DNNs.

Long subject that cannot be fully addressed here...