LLMs have their limits.
Meta wants your posts for AI training. Elon Musk and Jack Dorsey, too.
In a preprint study, researchers found that training a language model with human feedback teaches the model to generate incorrect responses that trick humans.
