No Priors: Artificial Intelligence | Technology | Startups

AI for Atoms: How Periodic Labs is Revolutionizing Materials Engineering with Co-Founder Liam Fedus

29 min

Liam Fedus, co-founder of Periodic Labs and a key figure in ChatGPT's development, discusses his new venture applying AI to materials science and the physical world. Periodic Labs aims to accelerate scientific discovery and engineering by creating an "AI Foundation Lab for Atoms," leveraging advanced AI models to orchestrate experiments and generate high-quality data in a closed-loop system. Fedus highlights the multidisciplinary nature of this challenge and the potential for AI to revolutionize physical development, akin to the agricultural revolution.

Summarized by Podsumo

🎙️ Ask about this episode

✨ Key Takeaways

1

Periodic Labs focuses on applying advanced AI to materials science and the physical world, aiming to accelerate scientific discovery and engineering by connecting AI systems to reality.
2

The company's AI architecture uses language models as an orchestration layer to direct experiments and integrate specialized neural networks designed for atomic systems, leveraging existing foundation models for general understanding.
3

Success hinges on generating high-quality experimental data within an interactive closed-loop system, where AI learns from experiments to inform subsequent ones, overcoming inconsistencies in literature data.
4

Fedus envisions AI literally generating matter, with profound implications for industries like semiconductors, aerospace, and energy, dramatically increasing the pace of physical development.
5

The project is inherently multidisciplinary, combining top AI researchers with physicists and chemists, and aims to bring the mindset of scaling laws and large-scale experiments from the digital to the physical sciences.

💬 Notable Quotes

"Science ultimately isn't sitting in a room thinking really hard. You have to conduct experiments, you have to learn from them, you have to interface with reality."
"I think just the inevitability of connecting these systems to the physical world."
"I think one fallacy is thinking about intelligence as a scalar. We've consistently seen these systems have a very odd spikiness."