The many roles that AI can play in accelerating scientific discovery
Invited-In-person · Invited · Withdrawn
Abstract
Artificial intelligence is rapidly transforming scientific discovery across multiple domains, with the AI industry investing over $350B in data center construction and targeting artificial general intelligence (AGI) by 2030. This transformation is driven by breakthrough advances in reinforcement learning, test-time compute scaling, and foundation models that are reshaping how we approach complex scientific problems.
AI's roles in accelerating discovery span six critical areas: advanced properties inference and inverse design for materials and chemistry; AI-robotics integration for autonomous laboratories; AI-based surrogates that enable 1000x faster high-performance computing simulations; software engineering automation including code translation and quantum compilation; prediction and control of complex engineered systems from reactors to power grids; and foundation models for scientific knowledge tasks including hypothesis formation and theory synthesis.
The Department of Energy's initiative in AI demonstrates these capabilities through novel tools, as well as specialized infrastructure that can industrialize AI development for scientific applications, providing the scale and specialized toolchains necessary to compete globally in AI-accelerated discovery. From cutting fusion energy R&D timelines to enabling real-time astronomical discoveries, AI is poised to compress research cycles from years to weeks across energy, materials, biology, and physics domains.
AI's roles in accelerating discovery span six critical areas: advanced properties inference and inverse design for materials and chemistry; AI-robotics integration for autonomous laboratories; AI-based surrogates that enable 1000x faster high-performance computing simulations; software engineering automation including code translation and quantum compilation; prediction and control of complex engineered systems from reactors to power grids; and foundation models for scientific knowledge tasks including hypothesis formation and theory synthesis.
The Department of Energy's initiative in AI demonstrates these capabilities through novel tools, as well as specialized infrastructure that can industrialize AI development for scientific applications, providing the scale and specialized toolchains necessary to compete globally in AI-accelerated discovery. From cutting fusion energy R&D timelines to enabling real-time astronomical discoveries, AI is poised to compress research cycles from years to weeks across energy, materials, biology, and physics domains.
–
Presenters
-
Rick Stevens
- Argonne National Laboratory