Image: Tong Xie, NCI’s 2023 HPC-AI Talent Program Recipient

Canberra, ACT – Tong Xie, a PhD researcher at the School of Photovoltaic and Renewable Energy Engineering (SPREE), UNSW Sydney, and recipient of NCI’s 2023 HPC-AI Talent Program, has made significant progress in material discovery using large language models (LLMs). His recently published research, titled “Creation of a Structured Solar Cell Material Dataset and Performance Prediction Using Large Language Models,” lays the groundwork for a new approach to materials discovery.

Leveraging LLMs for Material Science

Xie’s research focuses on utilising LLMs to accelerate material discovery by understanding the composition and properties of materials within solar cells. The vast number of scientific publications makes it challenging for researchers to stay up-to-date. LLMs offer a potential solution by efficiently processing information and extracting valuable insights. This can significantly improve the efficiency of information extraction from scientific literature and expedite the discovery of new materials. The research also establishes a benchmark for evaluating the accuracy and efficiency of LLMs, which will be instrumental in guiding future research and development in this field.

Role of NCI HPC-AI Talent Program

Xie expressed his deep gratitude to the NCI HPC-AI Talent Program, which played a critical role in his research by enabling him access to the Gadi supercomputer. Gadi’s immense computational power was essential for training and deploying the large language models used in the study. The HPC-AI Talent Program is NCI’s annual initiative to support early-career researchers by providing them with access to NCI facilities to undertake their research. Recipients are allocated computing time on the Gadi supercomputer and $10,000 for their research.

Future Directions and Continued Support from NCI

Xie’s research not only paves the way for further exploration but also sets a strategic direction for maintaining Australia's competitive edge in advanced manufacturing and material technologies. He envisions expanding the application of his AI-LLMs framework to various materials beyond solar cells, including transformative work with carbon nanotubes, supercapacitors, and so on. These functional material breakthroughs also enhance the manufacturing and application of renewable energy technologies, aligning with Australia’s Net Zero 2050 goals. Furthermore, Xie aims to deepen the integration of LLMs with material science and chemistry. The ultimate goal is to leverage LLMs to understand scientific principles, formulate hypotheses, and make scientific discoveries, functioning akin to a human scientist, Xie tells NCI. He underscored the critical role of NCI’s vast supercomputing resources in supporting the large-scale application of his framework. This integration of cutting-edge AI and computational power is a technological advancement and a strategic necessity, aligning seamlessly with Australia's existing industrial reconstruction goals. Continued support from NCI will be crucial to achieving these ambitious goals, as AI and LLMs represent a significant contribution to, and a necessary advancement for, the nation’s scientific and technological landscape.

Advice for Future Applicants of NCI HPC-AI Talent Program

Xie encouraged future NCI HPC-AI Scholarship applicants to be bold and explore innovative ideas that extend beyond the core focus of their PhD research. The NCI’s HPC resources offer a unique opportunity to explore these possibilities and make ground-breaking discoveries. He also highlighted the importance of exploring the full potential of Gadi’s capabilities and actively engaging with NCI staff to discover hidden functionalities. One way to achieve this, Xie recommends, is by attending NCI’s events for inspiration and networking opportunities.