Daniel Khashabi is an assistant professor of computer science at the Whiting School of Engineering and is affiliated with the Center for Language and Speech Processing (CLSP). He is interested in building reasoning-driven modular NLP systems that are robust, transparent, and communicative, particularly those that use natural language as the communication medium. Khashabi has published more than 50 papers on natural language processing and AI in top-tier publications. His research has won the ACL 2023 Outstanding Paper Award, NAACL 2022 Best Paper Award, and the 2022 Amazon Research Award. Before joining Hopkins, he was a postdoctoral fellow at the Allen Institute for AI (2019 to 2022) and obtained a PhD from the University of Pennsylvania in 2019.


Growing Chatbots Out of Thin-Air: Opportunities and Limits of Language Models for Guiding Themselves 

Abstract: The rise of pre-trained Large Language Models (LLMs) has enabled major progress in developing chatbots. However, they still rely on a significant amount of human supervision, limiting their progress by the amount of supervision needed and its quality. In this talk, I will explore whether we can replace human supervision with LLM. Specifically:

  • Can chatbots be aligned (trained) with LLM’s own supervision?
  • Can LLMs correct their own mistakes and continually self-improve?