Integration of Federated Learning and AI-Generated Content: A Survey of Overview, Opportunities, Challenges, and Solutions
Ying Liu, Jianhui Yin, Weiting Zhang, Changming An, Yu Xia, Hongke Zhang
Abstract
Artificial intelligence generated content (AIGC) relies on advanced AI algorithms supported by extensive datasets and substantial computing power to generate precise and pertinent content. Federated learning (FL) enables the aggregation of large volumes of data and computing resources from various sources, all while safeguarding privacy. As a result, FL has emerged as a critical enabler in the realm of AIGC. This survey paper provides a comprehensive overview of the integration of FL and AIGC, namely federated AIGC models. First, we introduce the fundamental concepts of FL and AIGC. Next, we summarize four typical types of federated AIGC models. Subsequently, We highlight the threats to centralized federated AIGC models regarding data confidentiality, integrity, and availability and discuss the unique advantages of blockchain technology in decentralized federated AIGC models in addressing these issues. Finally, we look at potential emerging application scenarios and explore open issues and future directions for federated AIGC models.