The First A.I. Designed Viruses Are Here
from Pretrained Hosted by Pierce Freeman & Richard Diehl Martinez
A model generated 285 candidate genomes, a lab synthesized them, and 16 of them turned out to work. Sophia Kivelson (Stanford, biomedical data science) breaks down the Arc Institute paper everyone read as "AI designs viruses" - what actually happened, why it can't touch a human cell, and why the only RL environment for this stuff is a wet lab with a two-month rollout. Then the part that made us nervous: some of these phages work fine while barely resembling the original, and nearly every biosecurity screen we have is looking for resemblance. Plus benchtop DNA printers with blocklist firmware, why the rules stop at the lab door and not the laptop, and bacteria as tiny programmable factories.
Embed this episode
Chapters
- 0:58 Intro
- 2:30 What a Virus Is and Why PhiX174 Is Harmless
- 6:51 Phage Therapy, Antibiotic Resistance, and the Toy-Model Framing
- 10:55 Evo: A GPT-2-Level Genome Language Model
- 18:34 StripedHyena and What Genome Models Borrow from LLM Research
- 22:58 RL, Perturbational Data, and the Missing Post-Training Loop
- 29:24 Biosecurity by Omission and 16 Working Phages
- 35:21 Validating Genomes Without RLHF: The Wet Lab as an RL Environment
- 38:41 Gain-of-Function Research and the Regulatory Gap
- 45:06 Pandemic Risk and the Full Biorisk Stack
- 48:53 Synthesis Screening Against Generatively Designed Genomes
- 52:08 Open Weights, Vetted Access, and Who Decides
- 59:19 Programmable Bacteria and Drug-Discovery Economics
- 1:04:27 Outro
- 1:05:15 Post Show Banter
No transcript yet.