Observed Signal · Sep 22, 2026 · Podcast Episode · Source: AINews swyx · Impact: 2/5 · Sentiment: Positive
John Platt Discusses AI for Science and Google's ERA System
In a podcast episode from latent.space, John Platt, a Google scientist and Academy Award winner, discusses his work on applying AI to scientific problems. He introduces Google's Empirical Research Assistance (ERA), a system that uses language models like Gemini to automate the search for solutions to 'scoreable' scientific tasks. ERA has contributed to papers on climate change, including reducing contrail warming and improving fire detection via FireSat. Platt emphasizes the importance of domain expertise and warns against overfitting and reward hacking. The conversation also touches on his background in physics, including time with Richard Feynman, and his views on the future of AI and science.
Discusses AI technology (ERA) for scientific discovery, relevant to AI in ad tech and broader tech, but not directly about advertising or marketing.
Key Takeaways & Evidence Grounding
- Google's ERA uses LLMs like Gemini to automate solving scientific problems.
- ERA has led to at least ten papers, including climate change research.
- John Platt discussed reducing contrail warming via flight level adjustments.
- The FireSat project uses satellites to detect fires early.
- Platt advises starting with linear regression or SVM before complex models.
Connected Companies & Entities
1 Entity mappedOntology Mapping & Concepts
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.