Technology Keynote: AI Native Architectures and considerations for Generative AI

Joseph Yang, Hewlett Packard Enterprise. ABSTRACT: "Current cloud native datacenter architectures are not enough to support the demands of Artificial Intelligence, and in particular Generative AI and the large models that allow Generative AI to happen. The demands of Generative AI don’t just impact the infrastructure required to train LLMs but also the infrastructure for running day to day inference. This revolution in infrastructure demand also has sustainability implications as well. Luckily a lot of the problems in building AI Native infrastructure is not new, and have actually been encountered and solved as part of building the largest supercomputers in the world. HPE will go into, how the requirements for AI will continue to push the limits of datacenter architecture, as well as how HPE is taking lessons learnt from building the only 2 Exascale Supercomputers in the world and applying them to the world of AI."