Roundtables: A Conversation With the Creator of AI-Designed Viruses
A blueprint for a virus is a strange kind of creative act, and King's work makes me wonder whether we're ready for machines that propose life forms faster than we can understand them. The model didn't build anything, just spelled out genetic code, but the step from code to a living thing is getting shorter. I keep thinking about how much of biology is already digital, and how thin the line really is between a sequence on a screen and a particle that can infect a cell.
I'd want to know what the AI was trained on, and whether anyone has tried to synthesize its proposals. That's the test that matters. If a lab can take one of these blueprints and get a working virus out of it, we've crossed a line that most people haven't even started thinking about. The conversation around AI safety tends to focus on chatbots and autonomous weapons, but a generative model that proposes viable genomes feels more immediate to me.
This connects to the bigger question of AI as a scientific partner. We usually think of it as a calculator or a text generator, but here it's acting as a designer of biological machinery. I'm not sure the safety framework exists for that. I'm excited by the possibility, but I'd want a lot more scrutiny before anyone grows one of these things. The wonder and the worry are tangled together, and I think that's the honest place to stand.