AI is dreaming up new materials. Are they any good?
16 hours ago
- #AI in materials science
- #computational chemistry
- #crystalline materials
- Google DeepMind used AI to discover 2.2 million new crystalline materials, including potential battery and graphene-like compounds.
- Critics argue some AI-generated materials are unoriginal, unfeasible, or impractical, citing examples with rare radioactive elements.
- Meta's AI suggested over 100 materials for carbon capture, but computational chemists questioned their viability.
- AI tools like GNoME and MatterGen aim to accelerate materials discovery by predicting stable crystal structures faster than traditional methods.
- The A-Lab project used AI and robotics to synthesize new inorganic compounds, but its results faced scrutiny over mischaracterization and disorder in crystal structures.
- Disorder in crystal structures poses challenges for AI predictions, as many ordered structures may not be stable or feasible in real-world conditions.
- Despite criticisms, AI holds promise for materials science, but collaboration with experimental chemists and addressing current limitations is crucial.