Why AlphaFold Didn't Solve Protein Folding — Pushmeet Kohli, Google DeepMind & Sal Candido, Biohub
October 10, 2026
AI Summary
5 min readThe "protein folding problem" hasn't been solved — not in the way most people think. When AlphaFold 2 was announced, the public narrative was that a grand challenge of biology had been conquered. But as Pushmeet Kohli (Google DeepMind) and Sal Candido (Chan Zuckerberg Biohub) explain in this conversation, that framing is misleading. What AlphaFold actually did was replicate static structures that scientists had already deposited in the Protein Data Bank. It did not resolve how proteins move, how they function, or how they behave in different cellular contexts. "We don't know what is the actual true ground state that proteins take," Kohli says. The episode is a careful unpacking of what it means to build models for biology — and why the hardest problems remain wide open.
The Bitter Lesson, Recast for Data
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Latent Space: The AI Engineer Podcast
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 Why AlphaFold Didn't Solve Protein Folding — Pushmeet Kohli & Sal Candido
- 2 (00:04) **The Bitter Lesson for Data** - Opening framing: does scaling data always win, or is there a "better lesson" for data?
- 3 (01:45) **Why "Junk" Data Beats Clean Data** - Sal explains why low-quality metagenomic sequences improve protein language models
- 4 (03:54) **The Community Solution to Data Bias** - Pushmeet introduces the need for community-driven data generation
- 5 (04:53) **Pushmeet's Take: The Bitter Lesson is About Religious Thinking** - Reframes Sutton's bitter lesson as a warning against identity-based solution choices
- 6 (08:23) **Actionable Advice: Think Multidisciplinary, Fail Fast** - How to approach problem selection when you have constraints
- 7 (09:54) **Artisanal Models vs. Scalable Strategies** - The evolution from AlphaFold 2's handcrafted features to general-purpose approaches
+ Full timestamped outline available in the app
Show Notes
From the Bitter Lesson of AI scaling to the unsolved mysteries of protein folding, Google DeepMind’s Pushmeet Kohli and Biohub’s Sal Candido are rethinking what it takes to build AI that truly understands biology. In this special panel moderated by Brandon Anderson, they explore why AlphaFold’s breakthrough was only the beginning, why scaling compute and data alone won’t solve biology, and how the next generation of AI models could transform our understanding of proteins, cells, and human disease.
We go deep on the future of AI-driven biology: finding scaling laws in biological data, the tradeoffs between scientific intuition and general-purpose architectures, why protein structure prediction is far from solved, and what it would take to build predictive models of living systems. Pushmeet reflects on the lessons behind AlphaFold, the limits of human interpretability, and whether future frontier models could understand other AI systems better than we can. Sal explains why protein language models may already contain scientific knowledge we haven’t unlocked, how biological modeling must move beyond individual proteins, and why achieving Biohub’s mission to cure all disease requires thinking in terms of 10x breakthroughs rather than incremental improvements.
We discuss:
* The Bitter Lesson for biology: why scaling compute and data isn’t enough
* Why finding the right scaling law matters more than blindly increasing model size
* How low-quality metagenomic data can improve protein language models
* Why AI researchers optimize for available data instead of the most important scientific problems
* Lessons from DeepMind on balancing modeling, data generation, and scientific expertise
* Why building a virtual cell requires fundamentally different datasets
* AlphaFold’s handcrafted architecture and the role of scientific intuition
* Why good data matters more than simply having more data
* Inductive biases, scaling laws, and the future of specialized AI architectures
* Why we aren’t in a post-Transformer world, but architectures are evolving
* Why AlphaFold didn’t actually solve all of protein folding
* Protein dynamics, disorder, and the limitations of static structure prediction
* How cryo-EM micrographs could unlock richer biological representations
* Moving from models of individual proteins to whole biological systems
* Feynman’s famous principle and why AI can now create things we don’t understand
* The hidden biological knowledge