@fribbledom so this is GPT trained on just texts? I.e. it is more interested in generating text that looks like text it has seen?
Instead of responding with the most relevant knowledge, it might respond with knowledge it can produce text best with?
Here, it might have steered away from the actual nanotechnology topic.
https://medium.com/@kirkouimet/nanotechnology-6b3ebbae5e80
I reckon it's also steered that direction a bit with the "wise and loving being" description.