There is no reasonable definition of consciousness that even the largest neural net fits. I don't understand why anyone believes that a neural net is conscious to the slightest degree.

No matter what I tell GPT-3, it will never remember one part of it. It will never learn my name, or anything about me, or any new facts from me, because GPT-3 has no mutable state at all.

A program that cannot change its state is not conscious.

Do I believe a computer program could become conscious? Yes, absolutely it's possible. I just don't think we're at all close right now. New breakthroughs will be needed.