He's Teaching AI to Read and Write DNA

DNA is this genetic code that makes up a lot of life on Earth. Just last year, one of the things that scientists [music] had used to generate with our models was the first AI genome from scratch. Genome is the complete set of instructions for an organism. They created what's known as a bacteria

phase. Something that did not exist in nature. We realized that this power, [music] this potential to create life itself or to manipulate life itself with AI has the potential [music] to reinvent pretty much all of biology. There's more DNA publicly online than all the text on the internet. the expressivity power of these models being applied to biology.

They're basically as strong as linear regression or logistic regression. Basically, this is the vast green space [music] that's open for mining. Basically, we don't see any tool out there publicly available on the AI side that can detect dangerous pathogens. So, we initially were like, let's just make an API and just host it. For us, this

should be pretty simple [music] to do. And then when we started engaging with biocurity experts, they basically told us like that don't do that.

Last year in a lab, researchers sat down at a computer and designed a virus using an AI model called EVO. Not a paper about a virus, but an actual working bacteria phase built and infectious. Eric Nuin is one of EVO's creators. He is also the CEO of Radical Numeric, a company building cuttingedge AI models

for biology, starting with DNA. Since its founding last summer, Radical has raised $50 million from investors like Emergence Capital and Patrick Collison. In today's conversation, Eric and I discuss his winding path [music] to founding Radical, the unique language of DNA, why he believes data isn't the bottleneck for biological models, and why the company that makes it possible