AI reduces quantum chemistry complexity from cubic to near-linear
新智元 · wechat · 2026-08-26
Caltech team led by Anima Anandkumar introduced Kohn-Sham FNO, replacing the most expensive iterative step in DFT calculations with a Fourier Neural Operator (FNO), reducing complexity from $O(N^3)$ to $O(N \log N)$.
Key Advantages:
- Chain-of-Thought Iteration: Instead of direct prediction, it embeds into the physics iteration loop for self-correction, avoiding error explosion when extrapolating to large molecules common in traditional AI models.
- Extreme Efficiency: Simulated 82,500 valence electrons in a magnesium dislocation on a single NVIDIA B300, achieving a scaling exponent of 1.03 (near-linear) compared to 3.37 (cubic) in a 2019 study using 7,800 V100s.
- Generalization: Trained on only 8,504 structures, it handles both molecules and solids, showing significantly lower error than direct prediction models on drug-sized molecules.
Anandkumar has founded Accelerated Understanding to extend such simulations to 4D space.
Related event: Caltech AI Breaks 60-Year Quantum Chemistry Scaling Barrier(3 posts)→
More from Research
- WorldMemArena Accepted at NeurIPS: Memory Storage ≠ Memory Use — xwang_lk · 2026-10-16
- Mech Interp Veteran Rethinks Alignment, Misgeneralization, CoT Oversight — akbirthko · 2026-10-16
- WEDGE paper generates training environments conditioned on navigation agent failures — _akhaliq · 2026-10-16
- RExBench: best coding agent completes only 33% of autonomous research extensions — najoungkim · 2026-10-16
- Google's CLASHarness benchmarks multi-agent conflicts in computer use, finds frontier VLMs struggle — google · 2026-10-16
- Tsinghua team proves vanilla flow-matching sampling can't exceed constant-depth expressive power — SQZ · 2026-10-16