Exploring Challenges in Applying Foundation and Generative Models in AI

Abstract

This study provides a comprehensive review on generative models, and basic components both from the perspective of unimodality and multimodality. The analysis aims to distinguish contemporary generative AI models from their predecessors.  After providing a brief historical background the study discusses the recent applications of generative AI models, commonly used techniques in AIGC, and addresses concerns surrounding trustworthiness and responsibility in the field. Finally, it explores open problems and future directions for AIGC, highlighting potential avenues for innovation and progress.