From Models to Systems: The Next Frontier of Generative AI

Abstract

Generative AI is rapidly evolving from isolated models to integrated, system-level solutions. This seminar presents a unifying perspective that connects probabilistic foundations with modern deep generative approaches, including variational autoencoders, diffusion models, and attention-based architectures for complex data. It highlights the importance of joint modeling, where generative and predictive components are learned together to improve performance, robustness, and data efficiency. Building on these ideas, the talk introduces Generative AI Systems (GenAISys), a paradigm for composing multiple models and tools into scalable, reliable, and explainable frameworks. This systems-level view is positioned as a key step toward next-generation AI applications, including agentic systems and scientific discovery.

For more info, please follow this link.