AI Reads DNA Like Never Before, Now Open to Everyone

A new AI just cracked one of biology's toughest puzzles, and the team behind it is giving it away for free.

Evo 2 is an artificial intelligence system that learned to read and understand DNA by analyzing genetic sequences from bacteria, plants, animals, and humans. After training on 8.8 trillion bases of genetic code, it can now spot important patterns in our genome that even expert scientists struggle to find.

This might sound technical, but it's a huge deal. Human DNA is incredibly messy and complicated, unlike the neat, organized genes found in bacteria. Our genes are interrupted by junk sequences, controlled by scattered instructions, and buried in what scientists once called "junk DNA." Finding the important bits has always been like searching for needles in a massive haystack.

The team trained two versions of Evo 2, with the larger one using 40 billion parameters to understand patterns across all domains of life. The logic was elegant: if a genetic feature matters enough to survive across millions of years of evolution, it will show up repeatedly in different species, and the AI will learn to recognize it.

What makes this breakthrough special is that nobody told Evo 2 what to look for. It figured out how to spot important genetic features on its own, including regulatory switches and splice sites that control how genes work. This means it might even discover patterns that scientists don't know exist yet.

The researchers made everything completely open source: the AI models, training code, and the entire dataset. Any scientist anywhere can now use this tool to accelerate their research, whether they're studying rare diseases, developing new treatments, or understanding how life evolved.

Why This Inspires

In an era where major AI breakthroughs often get locked behind corporate paywalls, this team chose a different path. They spent enormous computing resources and years of work building something powerful, then handed it to the world.

The decision to make Evo 2 open source means researchers at small universities and underfunded labs now have access to the same tools as wealthy institutions. A graduate student in Kenya can use the same AI as a team at Harvard. Someone studying a rare genetic disease that affects only hundreds of people worldwide can now tap into insights from trillions of DNA sequences.

The team deliberately excluded human viruses from the training data, showing they thought carefully about safety while still maximizing the tool's usefulness for good. They built something powerful and trusted humanity to use it wisely.

This is science at its best: brilliant minds solving hard problems, then lifting everyone else up with them.