Ames Laboratory develops DuctGPT physics-trained AI workflow — discovers rare-earth-free permanent magnets, novel materials with production costs and component sourcing built into design
The Ames Laboratory developed an AI workflow using a physics-trained model called DuctGPT to discover rare-earth-free permanent magnets. Unlike traditional AI trained on existing data, DuctGPT understands underlying physics and can invent new materials while considering production costs and component sourcing. The physics-trained approach represents substantively different AI-for-scientific-discovery methodology.
The substantive piece is the physics-trained-not-data-trained methodology distinction. Pre-DuctGPT AI-for-scientific-discovery methodology dominantly trained on existing materials data — the AI could find patterns in existing materials but couldn't invent genuinely novel materials. Physics-trained methodology operates against underlying physical principles — enabling invention of novel materials including production-cost + component-sourcing constraints.
The competitive read for H2 2026 AI-for-scientific-discovery direction is that physics-trained methodology represents substantively different research direction than data-trained methodology. The rare-earth-free permanent magnet application has substantial strategic-materials implications (China dominates rare-earth supply); successful physics-trained methodology for materials discovery could substantially affect technology supply chains.
Crescendo — The Latest AI News and Breakthroughs That Matter Most → · AI News — Latest News — Insights Powering AI-Driven Business Growth →