Artificial intelligence has made exponential leaps in its ability to produce text that looks and sounds like human-generated writing. Research shows, however, that machine-generated text is far from undetectable.
Yes, AI detection software is available and continuously improving. But computer scientists from the University of Pennsylvania School of Engineering and Applied Science have demonstrated that we can train ourselves to recognize AI-generated text using tools that operate like simple games, sharpening our attention to the glitches and false steps to which AI is prone.
The study, led by Chris Callison-Burch, associate professor in the Department of Computer and Information Science (CIS), along with Liam Dugan, Ph.D. student in CIS, and Daphne Ippolito, recent CIS doctoral graduate, is now the subject of an interactive CNN feature.
Read CNN’s “Bot or not? How to tell when you’re reading something written by AI” to learn more about how you can train yourself to spot when the author is AI.