Why AlphaFold Didn't Solve Protein Folding — Pushmeet Kohli, Google DeepMind & Sal Candido, Biohub

AlphaFold only replicates static PDB structures, not true protein dynamics — the folding problem remains far from solved.

• Bitter Lesson reframed: Sal Candido says scaling laws aren't inherent everywhere; much of the real work is discovering where more compute/data actually improves outcomes, turning science into an engineering problem.
• Noisy data can help: Training protein language models on low-quality, often non-real metagenomic sequences still boosts performance on real protein design and understanding tasks.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.