Exploring the fundamental connections between artificial intelligence and physical systems, bridging the gap between computational intelligence and the laws of nature.
Our AI & Physics research program investigates the deep connections between artificial intelligence and physical systems. We believe that understanding these fundamental relationships is key to developing more robust, efficient, and interpretable AI systems.
By drawing inspiration from physical principles such as energy conservation, symmetry, and emergent behavior, we develop novel approaches to machine learning that are both theoretically sound and practically effective.
Developing a physics-inspired framework that treats neural network inference as a path integral problem, enabling new approaches to optimization and uncertainty quantification.
Creating AI systems that respect energy conservation principles, leading to more stable and predictable behavior in complex environments.
Incorporating physical symmetries into neural architectures to improve generalization and reduce the need for large training datasets.
Studying how complex behaviors emerge from simple interactions in AI systems, similar to phase transitions in physical systems.
Exploring quantum computing principles to develop new algorithms and architectures for classical AI systems.
Applying thermodynamic principles to understand and improve the efficiency of machine learning algorithms and hardware.
Our physics-inspired AI approaches are particularly well-suited for scientific applications, where interpretability and adherence to physical laws are crucial. We're working on applications in materials science, drug discovery, and climate modeling.
Energy-aware AI systems can lead to more efficient and stable robotic control systems, particularly in dynamic environments where energy conservation is important.
We're developing AI tools that can accelerate computational physics simulations while maintaining physical accuracy and interpretability.
Coming August 2025 — arXiv preprint
A physics-inspired framework linking deep-network inference, energy bounds, and cardinality-cascade pruning.
Interested in collaborating on AI & Physics research? We're always looking for researchers, students, and industry partners who share our vision of bridging AI and physical systems.
Contact: [email protected]