Midweek Keynote: The New Landscape of Climate Computing

Climate change and AI are conspiring to create new markets in climate services: climate change contributes by making information, previously discounted by experience, more valuable; AI contributes by making the provision of this information scalable and affordable. The response to this new situation is redefining the climate computing landscape to meet three grand challenges: (1) to design and run more physical climate models at the scale of local impacts globally, across decades and scenarios; (2) to use their output — a complex hyperplane in a terra dimensional space — to train new types of foundation models; (3) to fine tune the models to map what is learned into the imagination of users, and then scale this across a very wide user community. The first two steps are computationally intensive, as they require access to exa-scale computing and data, but can be localized and intermittent. The third step involves structured human interaction and advanced AI capabilities, and must be continuous and distributed. This changing landscape of climate computing is reflected in initiatives like EVE, Destination Earth, and the Klima Hanse. In this talk, I detail this new landscape, the forces that are shaping it, and the ecosystem of services it entails.