UT Austin researchers pioneer machine unlearning to sanitize generative AI models

Researchers at The University of Texas at Austin have developed a method called "machine unlearning" to remove copyright-protected and violent content from generative AI models. This approach allows for the active blocking and removal of undesirable content without starting from scratch. Traditionally, the process required manually removing data and retraining the model. The new method is significant for ensuring responsible use and making generative AI models commercially viable. The research focuses on image-based generative AI models and incorporates human teams …

ai models austin blocking copyright data generative generative ai models machine process researchers scratch texas university university of texas university of texas at austin unlearning