Date：2025/10/8 16:00-17:20
Location：R103, CSIE
Speakers：Cheng-Chun Lee 李丞竣
Host：羅紹元教授
Abstract:
The rapid advancement of Generative AI like ChatGPT and Gemini, has highlighted the critical need for robust safety measures. As these models become more capable, ensuring they are helpful, harmless, and aligned with human values is a paramount challenge.
In this talk, I will give a high-level introduction to GenAI safety, drawing from my work on Gemini post-training. We will explore different techniques to shape model behavior and mitigate risks. I will share practical learnings, outline the open challenges the AI community still faces in building reliable and helpful systems.
Beyond the technical dive, this presentation will also chart my personal path from Taiwan to Switzerland. I will reflect on my journey through the Swiss academic landscape, sharing experiences from my master's studies at EPFL and my thesis work at ETH Zürich, and how my time in Switzerland shaped my career.
In this talk, I will give a high-level introduction to GenAI safety, drawing from my work on Gemini post-training. We will explore different techniques to shape model behavior and mitigate risks. I will share practical learnings, outline the open challenges the AI community still faces in building reliable and helpful systems.
Beyond the technical dive, this presentation will also chart my personal path from Taiwan to Switzerland. I will reflect on my journey through the Swiss academic landscape, sharing experiences from my master's studies at EPFL and my thesis work at ETH Zürich, and how my time in Switzerland shaped my career.
Biography:
Cheng-Chun is a Senior Software Engineer at Google DeepMind. He received his B.S. degree in Electrical Engineering from National Tsing Hua University in 2017, and his M.S. degree in Computer Science from École Polytechnique Fédérale de Lausanne (EPFL), Switzerland, completing his thesis at ETH Zürich. His career began with work on several language modeling projects, including BERT and LaMDA. Since ChatGPT was released, he has been working on GenAI safety and health for Gemini.