Toward AI as Co-Scientists: From Physics-Grounded Learning to Strong and Trustworthy AI
Date: Nov. 12, 2026
Speaker: Dr. Zongrui Pei
Affiliation: The University of Texas at Arlington
Abstract:
Artificial intelligence (AI) agents based on large language models (LLMs) for various
scientific applications have been developed, accelerating research. Nonetheless, these
AI tools are used mainly for simple tasks and short reasoning chains with heavy human
involvement. They are still weak at handling complex research tasks. Here, we use
a few examples, such as materials discovery and advanced computing, to discuss how
LLMs and AI agents can help mitigate barriers to research and to further explore opportunities
with world models. Building on insights from these examples, we extend our discussion
to co-scientists who are AI agents that accelerate and automate research, and we outline
pathways toward strong AI, or artificial general intelligence (AGI), from four perspectives.