With This 1 Move, Google May Have Solved a Huge AI Content Problem
Lichao Sun, professor of computer science and engineering, and research colleagues have released a study into a technique that prevents AI systems from unfairly or illegally “scraping” musical tracks to train their data-hungry AI algorithms, Inc. reports. The HarmonyCloak tool adds data to music recordings that effectively tricks an AI algorithm into thinking it’s already heard a particular track. The algorithm decides there’s nothing new there to learn from, and so it moves on without incorporating the music into its training data.