Scientific computing must adapt in an AI world

While the scientific computing community has long been a driver of computing innovation, that is no longer possible. Today, the community must embrace the new realities of an AI-dominated technology world, say Jack Dongarra and colleagues in this Policy Forum. It should integrate AI with simulation and – because energy has become a primary design constraint – it should focus on energy-efficient methods and systems. To make swift progress, the authors propose an initial “minimum viable” suite of workflow-shaped benchmarks. As part of these workflows, data and models are “intellectual gold,” say the authors; they “are investments in future scientific leverage.” Dongarra et al. propose a “moonshot” to build a new foundation that would benefit both scientific computing and AI.