When AI Cannibalizes Its Data | NPR's Short Wave

Today, we’re featuring an episode from NPR’s science podcast Short Wave. In it, host Regina G. Barber talks to computer scientist Ilia Shumailov about maybe the buzziest topic around: AI. I’m sure you know AI models like OpenAI's ChatGPT are trained on millions of examples of human-written text. Nowadays, a lot of content on the Internet is written by these generative AI models. That means that AI models trained now may consume their own synthetic content and suffer the consequences. What's the harm? Find out with this episode of Short Wave . Follow NPR's Short Wave podcast on Apple Podcasts and Spotify for more episodes like this, featuring new discoveries, everyday mysteries and the science behind the headlines. Listen to every episode of Short Wave sponsor-free and support our work at NPR by signing up for Short Wave+ at plus.npr.org/shortwave . Hosted on Acast. See acast.com/privacy for more information.