Visual forensics to detect fake text

Researchers at the SEAS and IBM Research developed a better way to help people detect AI-generated text. In a world of Deep Fakes and far too human natural language AI, researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) and IBM Research asked: Is there a better way to help people detect AI-generated text? That question led Sebastian Gehrmann, a PhD candidate at SEAS, and Hendrik Strobelt, a researcher at IBM, to develop a statistical method, along with an open access interactive tool, to detect AI-generated text. Natural-language generators are trained on tens of millions of online texts and mimic human language by predicting the words that most often come after one another. For example, the words "have" "am" and "was" are statically most likely to come after the word "I".