Trustworthiness and Energy Efficiency in AI for Science

This panel will convene leading experts to address critical challenges in developing reliable and efficient AI systems for scientific applications. Panelists will examine methodological innovations required for validating complex scientific models under uncertainty constraints, alongside architectural optimizations that reduce computational footprint without sacrificing performance. Discussion will focus on emerging hardware and software frameworks, energy-aware training paradigms, and domain-specific benchmarking necessary for deploying AI systems that scientists can both trust and deploy responsibly within constrained computing environments.