Comment by josu
2 days ago
The combination of the most probable tokens doesn't necessarily have to be in the training set, thus creating something completely novel.
2 days ago
The combination of the most probable tokens doesn't necessarily have to be in the training set, thus creating something completely novel.
If you shuffle a deck of cards, you likely have a novel sequence of cards. I'm not sure how that is relevant?
I was responding to this "merely produces syntactically likely combinations of the training data".
That is called a hallucination.
Yes, like the Jacobian counterexample.
That LLM was attached to a Python and MATLAB evaluator. It generated plausible things to feed into a real computer.
1 reply →
Only if it's incorrect.