I am fairly confident LLMs are plateauing or will do so soon. I am also fairly confident that there are other kinds of models, either already invented or yet to be conceived, that will go further than LLMs.

I don't think this is controversial at all.