Latent Space: The AI Engineer Podcast

🔬Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)

1h 32min

Eric Nguyen (CEO, Radical Numerics) discusses the urgent need for AI-powered biosecurity to catch up with AI-powered DNA design, framing it as a critical arms race. He shares how their Omni model achieves state-of-the-art results in human genomics, especially for non-coding disease variants, and introduces the first-ever chain-of-thought optimization in DNA models. The episode covers the company's dual mandate to advance both beneficial biological design and the defensive tools needed to prevent misuse.

Summarized by Podsumo

🎧 Listen 🎙️ Ask about this episode

✨ Key Takeaways

1

The defensive side in biosecurity is far behind the offensive design side, creating an arms race dynamic; Radical Numerics aims to close this gap by applying the same AI models to both.
2

Omni achieves state-of-the-art results on human genomics tasks, particularly in non-coding regions, outperforming both specialized models and super-ensembles like CADD.
3

A DNA language model has, for the first time, demonstrated chain-of-thought optimization: it can iteratively design better RNA aptamers by learning from progressively higher-fitness sequences.
4

Current biosecurity surveillance relies heavily on sequence matching, but AI-generated sequences can be functionally equivalent yet genetically distinct, bypassing those detections.
5

The company operates with a dual mandate: advancing DNA design for beneficial use while building frontier AI-powered defensive tools, seeing both as necessary and synergistic.

💬 Notable Quotes

"We felt it was important as a lab that a team that was both building the design capabilities is actually also best suited for building the defense capabilities because they're basically the same models."

— Eric Nguyen
"A model that is good at generating turns out is also very good at discriminating or predicting if a sequence is pathogenic or not."

— Eric Nguyen
"Why would you generate DNA? ...I spent six months going around saying, 'hey, if I generate DNA, would you find that useful?' ...almost every scientist at Stanford I talked to thought it was a stupid idea."

— Eric Nguyen