The Reversal Curse: Why a Language Model That Knows “A Is B” Can’t Tell You “B Is A”

# The Reversal Curse: Why AI Gets Confused When You Flip a Question Around AI language models like ChatGPT can learn a fact in one direction but struggle when you ask it backwards—for example, a model might know "Elon Musk founded Tesla" but fail to answer "Who founded Tesla?" even though it's the same information. This happens because these models memorize facts in the specific way they encounter them during training, rather than truly understanding the underlying relationship. It's a reminder that despite appearing intelligent, AI systems don't reason the way humans do and can have surprisingly narrow knowledge gaps.