Xiazhong Liu
Topic: Adopting Generative AI in Critical Domains

Speaker: Dr. Xiazhong Liu, Associate Professor of Data Science at Worcester Polytechnic Institute

Date: April 25, 2025 3pm to 4:30pm

Location:

  • In-person: Discovery Park Building, E297L


  • Online via Zoom

Abstract: Generative AI is increasingly being adopted in high-stakes domains such as healthcare, drug discovery, code generation, and legal science. While these advancements offer transformative potential, they also raise serious concerns regarding trustworthiness, transparency, and security. In critical settings, even minor errors can lead to significant and sometimes irreversible consequences. In this talk, I will present our recent efforts to address these challenges, including strategies for building trustworthy language models through human feedback integration, fostering effective human-AI collaboration, and developing techniques for generative AI calibration. We will also explore the broader social implications and risks associated with deploying these models in sensitive, high-impact environments.Generative AI is increasingly being adopted in high-stakes domains such as healthcare, drug discovery, code generation, and legal science. While these advancements offer transformative potential, they also raise serious concerns regarding trustworthiness, transparency, and security. In critical settings, even minor errors can lead to significant and sometimes irreversible consequences. In this talk, I will present our recent efforts to address these challenges, including strategies for building trustworthy language models through human feedback integration, fostering effective human-AI collaboration, and developing techniques for generative AI calibration. We will also explore the broader social implications and risks associated with deploying these models in sensitive, high-impact environments.

Bio: Xiaozhong Liu is an Associate Professor of Data Science at Worcester Polytechnic Institute. His research spans knowledge management, natural language processing, and computational social science. His recent work focuses on addressing critical challenges in LLMs and knowledge graph, leveraging cutting-edge generative AI techniques to develop novel models and systems. Dr. Liu leads large-scale projects in healthcare AI and legal AI, aiming to build trustworthy and reliable LLMs for deployment in high-stakes domains. His research also examines the risks, threats, and vulnerabilities of AI systems, designing innovative attack and defense frameworks to enhance model safety and trustworthiness. He has authored over 150 publications in premier computer science conferences and leading information science journals.