Argonne National Laboratory, a prominent multi-disciplinary science and engineering research centre operated by the University of Chicago for the U.S. Department of Energy, is making strategic use of its available supercomputing infrastructure. The laboratory is reallocating spare capacity from its powerful supercomputers to establish a private artificial intelligence (AI) inference service. This internal initiative, reportedly referred to as 'ChatDoE', is designed to provide secure and dedicated AI capabilities for scientific research within the Department of Energy's extensive network.
The creation of an in-house AI inference service signifies a growing trend among major research institutions to build bespoke AI environments. By utilising existing, under-utilised supercomputing power, Argonne can avoid the reliance on commercial public cloud AI services, which often come with concerns regarding data security, privacy, and potentially high operational costs. This approach allows for greater control over the AI models and the data they process, which is particularly crucial for sensitive scientific data and proprietary research.
AI inference refers to the process of using a trained AI model to make predictions or decisions based on new data. In a scientific context, this could involve accelerating material discovery, analysing vast datasets from experiments, simulating complex physical phenomena, or even aiding in drug development. Having a dedicated, high-performance inference service accessible only to authorised users within the Department of Energy could significantly boost the pace and scope of research projects.
The move by Argonne underscores the increasing integration of AI into fundamental scientific research across various disciplines. Supercomputers, traditionally used for complex simulations and data processing, are now being re-purposed to meet the demanding computational requirements of advanced AI models. This dual-use strategy maximises the return on investment for these expensive national assets, ensuring their continued relevance in the rapidly evolving landscape of scientific computing.
While this development is specific to a US institution, it reflects a broader global recognition of the need for robust, secure, and dedicated AI infrastructure. Governments and research bodies worldwide, including in the UK, are exploring similar strategies to harness AI's potential while mitigating risks associated with data handling and external dependencies. The implications for the scientific community are profound, suggesting a future where bespoke AI services become an integral part of national research capabilities.