DeepMind and Biohub discuss limits of AlphaFold and AI-driven biology

Google DeepMind's Pushmeet Kohli and Biohub's Sal Candido discussed in a panel moderated by Brandon Anderson why AlphaFold has not solved protein folding or dynamics.

The panel covered whether scaling compute and data is enough for AI-driven biology and the limits of static structure prediction.

According to a Latent Space post promoting the discussion, Kohli and Candido argue that scaling compute and data alone is insufficient for biology. They also say protein language models may encode structure, function and evolution that has not yet been unlocked, and that better biological data and virtual cells could enable 10–100x faster drug discovery. These are the speakers' claims, not reported results.