ETH Zürich recently celebrated the launch of its AI-optimized “Alps” supercomputer with a scientific symposium focused on the future of AI in science. The event highlighted how increased computing power and a flexible architecture are opening up new possibilities for tackling some of the most challenging scientific projects.
Launched on September 14, the Alps supercomputer is housed at the Swiss National Supercomputing Centre (CSCS) and runs on a Cray Supercomputer EX from Hewlett-Packard Enterprise. The system is equipped with 10,752 Nvidia Grace Hopper superchips and, when fully expanded, is expected to reach half an exaflop in performance.
The symposium explored the latest advances in scientific AI and big data, with a focus on meeting the growing computational and data demands across various scientific fields. Attendees got a glimpse of the key areas the Alps infrastructure will address, including material discovery, the mapping of the universe, and AI-driven medical advancements.
Professor Nicola Spaldin from ETH Zürich discussed the future of materials science, suggesting that the progress we make in this field could define new eras of civilization. She pointed out that we may be nearing the end of the Silicon Age and emphasized the importance of developing new materials, such as multiferroic materials, to meet society’s needs in a sustainable way. With Alps’ combination of supercomputing power and machine learning capabilities, researchers will be able to explore complex materials science questions that could drive advancements for society.
The Alps infrastructure will also support efforts to map the universe. Jean-Paul Kneib, a professor at EPFL and the Swiss Science Delegate at the Square Kilometer Array Observatory (SKAO) Council, shared details about the SKAO project. This massive radio telescope array uses radio interferometry to collect data from hundreds of antennas to map the cosmos. Alps will help manage the enormous datasets generated by this project.
The symposium also looked at how AI can transform healthcare. ETH Zürich’s Professor Tanja Stadler, who previously led the Swiss advisory board on COVID-19, explained how her team developed AI tools for real-time tracking of COVID-19 variants using genetic sequencing data. Another presentation by Mary-Anne Hartley, a visiting professor at EPFL and assistant professor at Yale Institute for Global Health, explored the use of generative AI in healthcare decision-making. Hartley’s team trained AI models using data from medical research and clinical practice, creating the Meditron suite of open-source large language models. These models, co-designed with clinicians, will allow doctors and patients to ask questions and receive help with diagnoses, improving clinical practices in resource-limited areas.
Alps was launched by ETH Zürich and EPFL with the aim of establishing Switzerland as a global leader in the development and deployment of transparent and trustworthy AI solutions. It is one of the first supercomputers equipped with Nvidia’s GH200 chips. Researchers have already begun testing parts of Alps using the CUDA 12.3 software stack on GH200 nodes. Alps’ unified memory pool will provide new capabilities for researchers working on large-scale problems, particularly those requiring substantial memory.
