We invite general and domain science perspectives and experiences on the challenges related to energy efficiency in large-scale computing related to earth and space sciences research and applications. AI-enabled solutions will require integrating multimodal data while being cognizant of the energy demands introduced by petabytes of data and artificial intelligence and computational methods. Planetary scale simulations and AI solutions are energy-intensive. The energy footprint can be minimized by optimizing the energy value chain across the computing continuum involving algorithms, applications, methods, data management, hardware and infrastructure. Reduced and mixed precision algorithms on supported hardware offer the
potential to minimize computational and energy costs. Data reduction and feature optimization techniques are essential for data efficiency. The multimodality of the diverse data collections presents special challenges in optimizing the amount of data used in foundation models. There
is little guidance in the research community on developing a computational plan for the optimal use of the resources for pretraining and inferencing frontier AI models using multimodal scientific data. Partnerships and collaborations across international agencies, academia and
industry at the working level are essential for synergizing our efforts toward AI solutions.
The session will discuss advances, early results and best practices related to energy efficiency at all scales across the computing continuum. We invite submissions related (but not limited) to the following topics on energy efficiency in computing: applications, algorithms, data management, and hardware solutions, including domain-specific hardware, workflows, and integrated approaches. The session also encourages discussion on partnerships, collaborations and broad community involvement toward energy-efficient solutions.
potential to minimize computational and energy costs. Data reduction and feature optimization techniques are essential for data efficiency. The multimodality of the diverse data collections presents special challenges in optimizing the amount of data used in foundation models. There
is little guidance in the research community on developing a computational plan for the optimal use of the resources for pretraining and inferencing frontier AI models using multimodal scientific data. Partnerships and collaborations across international agencies, academia and
industry at the working level are essential for synergizing our efforts toward AI solutions.
The session will discuss advances, early results and best practices related to energy efficiency at all scales across the computing continuum. We invite submissions related (but not limited) to the following topics on energy efficiency in computing: applications, algorithms, data management, and hardware solutions, including domain-specific hardware, workflows, and integrated approaches. The session also encourages discussion on partnerships, collaborations and broad community involvement toward energy-efficient solutions.