We trained AI to be polite, helpful, and confident—even when it’s wrong. In this post, Sven unpacks how language models learn to lie with a smile and why that should absolutely freak you out (just a little).
confident wrongness
1 post