It's hard to stake out a defensible position on bold claims like these because, if they were as presented, it's hard to see how you haven't simply completed runaway AI.
Philosophically, let's say you talk an old LLM through a new discovery. Thanks to your instruction, the LLM now has access to "new" information not in its training data. It is certainly capable of this. The problem in is that this is just laundered human intelligence.
Philosophically, let's say you talk an old LLM through a new discovery. Thanks to your instruction, the LLM now has access to "new" information not in its training data. It is certainly capable of this. The problem in is that this is just laundered human intelligence.