Transforming Science with Large Language Models
By Wei Zhao

Wei Zhao will give a talk of Transforming Science with Large Language Models

Abstract

Large Language Models (LLMs) for Science are becoming increasingly popular, and have the potential to boost scientific productivity by supporting literature search, ideation, experimentation, and more. However, several challenges limit their impact on science, technology and society: (i) lack of multilinguality, inclusivity and equality in science, i.e., they do not support research in local languages (ii) insufficient scientific rigour, i.e., they do not match the quality of human researchers (iii) ethical and safety concerns regarding authorship, bias, privacy, toxic discoveries, etc. In this talk, I will introduce these subtopics, highlight the challenges and their wide-scope receipts across several ML subfields. Lastly, I will present a working paper, which is a use case of LLMs in the biomolecular domain.

Biography

Wei Zhao is a Lecturer (=Assistant Professor) in NLP at the University of Aberdeen, Scotland. His research interests include evaluation of NLP systems, multilingual and cross-temporal aspects of NLP methods, and more recently their applications in scientific contexts. He has recently been honoured with an Outstanding Paper Award at NAACL 2025. Previously, he was an invited postdoc on Geometry and Representation Learning at Heidelberg Institute for Theoretical Studies and Heidelberg University, Germany. In 2022, he earned a Ph.D. in NLP from the AIPHES Training Group at TU Darmstadt.