Transformer models suffer from "hallucinations". It can be terrible at giving quotes or references. It's a known limitation with this tech that the industry is working to overcome.

I asked it 'What hapoened to Gandalf after he fell down from the bridge, fighting the Balrog? Did he die?'

The answer was a story about Gandalf beeing hurt badly, beeing rescued by some random dwarfs, and so on.

I asked ChatGPT in which book this is described, and it told me that you can read about it on both The Hobbit and The Lord of the Rings.

So it makes up fun stories. This makes me wondering how much of its explantions about physics (which I don't understand completely) are made-up.