Researchers have achieved a groundbreaking milestone: for the first time, artificial intelligence has designed and created live viruses that do not exist in nature. A team from Stanford University and the Arc Institute used genomically trained language models (Evo 1 and Evo 2), which had been trained on vast libraries of DNA sequences from thousands of organisms. The AI was tasked with designing a bacteriophage genome capable of infecting E. coli. The model generated thousands of variants; approximately 300 were chemically synthesized and tested in the lab, with 16 proving to be viable. Some of these synthetic phages even outperformed natural ones — for instance, one reproduced up to 65 times faster than the original template. The team notes that one variant is «evolutionarily distant» from anything found in nature, suggesting that natural evolution would have taken millions of years to reach that configuration. While the researchers excluded human pathogen genomes from their training data, experts have raised concerns about the dual-use potential of such AI tools. Potential medical applications include developing bacteriophages as alternatives to antibiotics for drug-resistant bacteria.
About the Author
