Skip to main content
← Back to sources

Why There Is No AlphaFold Equivalent for Materials Science — Challenges in AI for Materials Discovery

Published 2026-03-25Foundation ModelsLow

Summary

Latent Space published a podcast episode featuring Heather Kulik, an MIT professor working at the intersection of AI and materials science, discussing why the materials discovery domain has not yet produced a breakthrough equivalent to AlphaFold in protein structure prediction. The conversation explores the fundamental differences between biological protein folding and materials science that make the latter a harder problem for AI — including the vastly larger combinatorial design space, the het

Alignment: Neutral
ai-for-sciencematerials-discoveryalphafoldfoundation-modelscomputational-chemistrymachine-learningscientific-computinglatent-spaceheather-kulikai-research
Why There Is No AlphaFold Equivalent for Materials Science — Challenges in AI for Materials Discovery — Intelligence — Agentic Developer Tools Radar · Signal