Convergence of AI, Simulations and HPC
summary
Introduces a Fourier Neural Operator that learns function-to-function maps using global convolution in the Fourier domain, trains on low-resolution (e.g., 64×64) simulations and zero-shot evaluates at much higher resolution, and reports about a 1000× speedup over GPU-based solvers while retaining accuracy on turbulent fluids and multiphase flows; presents OrbNet, a graph neural net built on molecular orbitals that matches gold-standard quantum energies at roughly three orders of magnitude faster and generalizes from <40-atom training molecules to molecules ~10× larger, and notes integration into NVIDIA SimNet for DGX/exascale workflows.
AI-written summary · about
chapters (5)
- 0:00AI golden age impact
- 1:41AI in scientific simulations
- 7:43Neural operator learning
- 19:45Quantum energy calculations
- 23:57Future of AI for science
related talks
24mKeynote: Next-generation frameworks for Large-scale AI
20mFoundations for AI+Science
23mFrom Gaming to the Metaverse
11mSimAI: Unifying Architecture Design and Performance Tuning for Large-Scale Large...
11mDeploying Physics Models at the Edge: Quantized AI for Weather & Fluid Simulation
04m4 real-world AI deployments at enterprise scale
44mFast Neural Networks… a no brainer!
40mFueling the AI Revolution with Gaming