AI Codes Life

Imagine prompting an AI like ChatGPT, not with words, but with DNA, where the AI replies back with actual coded base pairs for proteins. This is what Evo 2 does.

Evo 2 is a generative model for biology. It is similar to ChatGPT, but trained on DNA rather than language. It can create new proteins that serve the same function as existing ones, but with fewer risks or improved safety.

With generative AI model for DNA, the barrier for many biological innovations is significantly lowered.Instead of having to have humans search manually for proteins with similar effects, the AI can generate new ones that have a high chance of providing the same function. Scientists would then only need to test these proteins in a lab to identify which ones work best with the least amount of risk. Evo 2 autocompletes a gene like the way Google Docs sometimes autocompletes sentences. Using DNA trained from microbes, Evo 2 can autocomplete genes by finding patterns in the DNA, much like how google finds patterns in speech in order to complete sentences while typing.

This AI model, information, and all data associated with it, is all open source, giving scientists around the world improved DNA analysis capabilities. It could save enormous time and effort for future innovations. We could one day use this information and technology to debug DNA like how we debug software. Using AI to understand how the DNA works and eliminating the problem within.