Jinghui Chen, assistant professor in the Penn State College of Information Sciences in Technology, presented "Trustworthy Machine Learning in the Era of Generative AI" on September 5, 2024, as part of the IST Research Talks series.

Abstract

Generative AI models like ChatGPT have revolutionized data generation, information processing, and problem-solving across many fields. However, despite their impressive performance and vast potential, these models introduce significant challenges regarding trustworthiness, preventing us from reliably deploying them in real-world applications. This talk explored the emerging trustworthiness challenges in the era of generative AI and shared Chen's approaches to addressing these issues. He focused on two key areas: (1) evaluating and mitigating safety concerns in generative AI, and (2) steering the personality and behaviors of AI models. Additionally, he touched upon other important directions surrounding the trustworthiness of generative AI.

About the Speaker

Jinghui Chen is an assistant professor in the College of Information Sciences and Technology. He earned his doctoral degree in computer science from UCLA in 2021, and a bachelor of science degree in electronic engineering and information science from the University of Science and Technology of China in 2015. Chen’s research interests broadly include theory and applications in different aspects of machine learning, with a particular focus on building efficient and trustworthy machine learning models. His current research explores trustworthiness and safety issues in large language models, security and privacy issues for other emerging machine learning models, and efficient optimization strategies for training large-scale foundation models and federated learning.