Data in Material Science
The discussion reveals the challenges of training AI models in material science, highlighting the limited size of existing databases compared to the vastness of the Internet. While the inorganic crystal structure database contains over 200,000 crystals, advancements in simulations using quantum mechanics allow for millions of training points. The conversation also touches on the quality of computational versus experimental data, suggesting that the latter may not always equate to the high standards seen in Internet-scale datasets.In this clip
From this podcast
Catalyst with Shayle Kann
Can AI revolutionize materials discovery?
Related Questions