Initiated in March 2024, the Aurora GPT project was advocated and detailed in the report on recent DOE workshops on Artificial Intelligence (AI) for Science, Energy, and Security.

How it Works

The AuroraGPT project is leveraging DOE supercomputing resources to  develop and enhance understanding of powerful foundation models (FMs) [1] such as large language models (LLMs), for science — as outlined in a series of DOE workshops on Artificial Intelligence (AI) for Science, Energy, and Security [2].

By creating FMs for science—while developing underlying capabilities, tools and workflows, data resources, and other processes and artifacts—Argonne Argonne aims to significantly improve how science is conducted, by fostering a deeper integration of AI capabilities into research workflows. To this end, Argonne’s AuroraGPT project is creating and evaluating a series of increasingly powerful FMs, each with more parameters and/or trained on more data than those that precede it, designed to assist researchers in making more informed and efficient discoveries. The AuroraGPT research program focuses on producing this sequence of models while ensuring that each provides both a scientifically useful capability and knowledge concerning scientific and computational performance to guide the design of the next model in the sequence.

The project team comprises eight internal working groups led by Argonne scientists, including overall project planning and direction, data and training, evaluation and safety, inference, model architecture and performance, post- and pre-training, as well as distribution and communication teams.

This undertaking also requires strategic partnerships with like-minded projects around the world, leveraging Argonne partnership arrangements with world-class institutions around the world.