🔬 An Oscar, Two Asteroids, and the Algorithm in Your sklearn: John Platt on AI for Science
September 23, 2026
AI Summary
5 min readJohn Platt, a Google Fellow and the mind behind textbook algorithms like SMO for SVMs (still running under the hood in scikit-learn), walks through his team's work on AI for science. He has discovered two asteroids, holds an Oscar for the elastic simulation code that became a core tool at Pixar, and studied under John Hopfield and Richard Feynman. The conversation centers on a system called Aria (which stands for something like "AI for Research and Analysis") and the broader question of how general-purpose LLMs are changing what it means to do science.
The scorable task framework
The core idea behind Aria is that many scientific problems can be mapped into what the team calls a scorable task: you want a piece of code that maximizes some score. That sounds like ordinary machine learning, but the trick is that the "inner loop" is not random mutation or brute-force search — it is an LLM (Gemini) that has enormous prior knowledge about the world. Instead of randomly tweaking code, the system can propose sensible mutations because it understands what it is doing.
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 (00:00) **Predictive vs. Descriptive Models** - John Platt introduces a core distinction for scientific modeling.
- 2 (03:05) **Introducing the AI for Scientific Discovery (AIRS) Framework** - Platt explains the core "scorable task" mapping that underlies the AIRS system.
- 3 (09:58) **How AIRS Differs from General Agentic Coding** - Platt details the search mechanism and parallelization strategy.
- 4 (15:27) **The Human Role: Defining the Score and Avoiding "Reward Hacking"** - The hardest part of the problem is defining the right scoring function, which is a human task.
- 5 (17:37) **Why This Works: From Random Mutation to Informed Gradients** - Platt explains why LLM-based code mutation is a breakthrough over 1970s evolutionary coding.
- 6 (19:31) **Phase Transition with Newer Models** - The effectiveness of AIRS has seen a dramatic improvement with newer versions of Gemini.
- 7 (22:21) **The Core Challenge: Multiple Hypothesis Testing and Overfitting** - Platt addresses the risk of overfitting when an agent explores a huge state space.
+ Full timestamped outline available in the app
Show Notes
How often do you get to talk to a guest who has both an Academy Award and who invented textbook machine learning algorithms? John Platt has an Oscar, two textbook algorithms, two named asteroids, and an Erdos-Bacon number of 6. This was easily the most fun bio of all the guests we’ve read to date. And the result was an epic and fun chat covering Google’s Empirical Research Assistance (ERA), how AI can help battle climate change, and tons of great stories about the co-evolution of science and AI.
John’s colleague Dave Bacon likes to tease John that his career has been defined by being twenty years early to the next big thing. This may be convolutional neural networks (some credit him with coining the term), fusion research, quantum computing. John and Google have been working on solving some of humanity’s hardest problems with AI and computation for well over a decade now. Recently John and his team set their sights on using AI to solve any scientific problem that can be written down as a score.
Google’s Empirical Research Assistance (ERA)
John’s team has taken on many hard scientific problems over the years. In solving these, they noticed a pattern, many scientific problems can be reduced to what John calls a “scoreable task”. Once you have the score function, the goal is to find some code that maximizes the score. The hard part is in formulating the score, but once you have the score finding the maximizer can still be quite a lot of effort.
John’s team set out to automate solutions to this general problem. This came out of the idea of an “auto-Kaggle” AI, which can solve any Kaggle problem you can throw at it. Kaggle is owned by Google, so all the data was ready and easily available to them!
The result is Google’s Empirical Research Assistance or ERA (paper, github, blog). ERA is surprisingly simple conceptually. Gemini (or your LLM of choice) keeps a running tree of past experiments (notebooks) and where they’re going. It’s a close cousin of Monte Carlo Tree Search: at each iteration the