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

The integration of artificial intelligence into bioscience requires a shift from viewing data and modeling as separate, rigid disciplines to a problem-first, multidisciplinary approach. While scaling laws suggest that increased compute and data improve performance, success depends on curating high-quality, relevant information rather than simply increasing volume. Current breakthroughs like AlphaFold provide foundational structural insights, yet they represent only a partial solution, leaving complex dynamics, protein function, and design as critical, open challenges. True progress in translational medicine demands moving beyond static structure prediction toward models that capture the distribution and motion of proteins. Achieving this requires a balance between artisanal, physics-informed modeling and scalable, generalist architectures, ensuring that AI tools are not just powerful "black boxes" but reliable, interpretable systems capable of driving real-world clinical impact.