Generative AI and LLMs empower the most promising AI products so far.
We focus on understanding and solving very fundamental issues of generative AI and LLMs, including inference, optimization, and data-centric AI.
I have been working on theories (science of AI) and theory-driven AI algorithms for a long time.
My research covers very fundamental issues on optimization, generalization, representation, and data-centric AI.
In most of my theoretical works, physics inspires novel insights and tools for understanding and improving deep learning.
Many interesting things emerge from deep learning, ranging from small models to large models (e.g., GPT).
However, we still lack theoretical understanding and theory-driven advancements to towards the science of AI.
I believe that fundamental AI research is the key to maintaining a long summer of AI, especially in the era of large models.
“We are trying to prove ourselves wrong as quickly as possible, because only in that way can we find progress.” ― Richard P. Feynman