3 Minutes
Imagine a computer in a research lab not just analyzing data, but inventing new forms of life. That’s the reality reported by teams at Stanford and the Arc Institute, where researchers used artificial intelligence to design viruses that can kill bacteria — and those designs worked in the lab.
This marks the first time scientists have used AI to generate entire viral genomes from scratch. No copying of known viruses, no small edits to existing sequences — instead, the system produced wholly new genetic blueprints that translated into living biological agents.
The AI, named Evo, functions similarly to large language models but was trained on about 2 million viral genomes rather than books or web pages. When asked to design variants of a simple bacteriophage called phiX174, Evo produced 302 original genome sequences. Sixteen of those designs were synthesized and shown to infect E. coli successfully in laboratory tests.
Brian Hie, who leads the lab, described the process as watching digital code turn into biology — a moment that felt equal parts exhilarating and unnerving. The ability to move from in-silico design to real-world activity highlights how quickly computational tools are reshaping synthetic biology.
The potential benefits are significant. AI-designed viruses could be tailored to target antibiotic-resistant bacteria, which cause hundreds of thousands of deaths annually. They could also be adapted as precision delivery systems in gene therapy, ferrying corrective genes to specific cells. In short, this technology could accelerate treatments that are difficult or impossible with current antibiotics and vectors.
But there are serious safety and ethics concerns. J. Craig Venter, a pioneer in synthetic genomics, warned that this approach accelerates the trial-and-error process and raises the risk of misuse. Although Evo was trained only on bacteriophages that don’t infect humans, the underlying methodology could theoretically be retrained on more harmful targets. That creates a dual-use dilemma: a tool with powerful medical promise that could also enable dangerous biological engineering if misapplied.
Experts note we are still some way from AI creating full synthetic cells, which would require millions of genetic bases rather than the thousands in simple phages. Still, companies such as Ginkgo Bioworks are developing integrated pipelines that could automate the cycle from AI design to biological creation, potentially reducing human oversight in the loop.
This breakthrough is both breathtaking and disquieting. We may be at the point where life becomes programmable and the boundary between digital design and living organisms blurs. The pressing question is less about whether this will change everything — it clearly will — and more about whether regulation, governance, and safety controls will keep pace.
In related AI hardware and tools news, Huawei has introduced its Atlas 950 and 960 SuperPoDs to challenge Nvidia’s dominance, and Tencent released a free 3D AI tool that expands creative capabilities for developers and creators.
Source: gizmochina
Comments