Protein AI: Alex Rives Scaled Data, Not Just Models
Biohub's Alex Rives explains how his ESMC protein AI leveraged billions of metagenomic sequences to build a protein "world model," showing the power of scaling data over human intuition.
🔬 The Bitter Lesson is Coming for Proteins - Alex Rives, BioHub
With R.J. Haneki, Brandon, Alex Rives · Sunday, May 31, 2026
This episode features Alex Rives from Biohub discussing the application of the 'bitter lesson' (scaling laws and empirical evidence) to protein biology through the ESMC language model. Rives explains how massive datasets, particularly metagenomics, enabled ESMC to build a comprehensive 'world model' of proteins, facilitating the design of novel therapeutics like antibodies and offering deep insights into protein function through mechanistic interpretability. The discussion also covers Biohub's ambitious Virtual Biology Initiative, aiming to integrate advanced experimental techniques with AI to create predictive digital representations of biological systems, from molecules to cells, to accelerate scientific discovery and combat disease.
Key takeaways
- The "Bitter Lesson" applies to biology: Alex Rives, a key figure at Biohub, champions the principle that empirical scaling with massive data often outperforms human intuition or handcrafted rules in building powerful AI, even for complex protein biology. Read more →
- ESMC's Blind Insight: The ESMC protein language model, trained only to predict amino acids from sequences, developed a "world model" of proteins so accurate it intrinsically understood complex biological concepts like the 'nucleophilic elbow'—without any prior biological input. Read more →
2 articles from this episode
Your Model Knows More Than You Taught It: ESMC & SAEs
Biohub's Alex Rives reveals how ESMC protein AI, using Sparse Autoencoders, autonomously discovered deep biological principles like the 'nucleophilic elbow' from sequence data alone.
More Latent Space episodes
- Inside the Team That Killed Mandatory Code Review — Quinn Slack, AMPSep 13 · 6 articles
- The Inference Frontier: from 100 to 10,000 tokens per second — Sean Lie, Cerebras CTOSep 6 · 6 articles
- Trillion Token Context. No, Really — Anima Anandkumar & Benedikt Jenik, Accelerated UnderstandingSep 6 · 5 articles
- 🔬 Why Transformers Hit a Wall the Moment Physics Shows Up — Anima Anandkumar, CaltechAug 30 · 5 articles