https://dev.mpost.io/ai-researchers-have-taught-large-language-models-to-lie-less/