“New Viruses”... With Artificial Intelligence! - Sarmad
In a scientific achievement that reflects the rapidly advancing capabilities of artificial intelligence in the field of genetic engineering, researchers have succeeded in designing new viruses that do not exist in nature, with some of them reproducing at speeds surpassing their natural counterparts. While scientists believe this development could open wide horizons for advancing medicine and biology, it has simultaneously raised questions and warnings about potential risks, particularly the possibility of misusing these technologies to develop biological weapons, according to The Telegraph.
A team from the Arc Institute in Palo Alto, California, used AI to train a computer model to recognize natural patterns in DNA, before leveraging that knowledge to design novel viruses not found in nature.
Although the resulting viruses, known as “phages,” do not infect humans or animals, the achievement has sparked concerns that the same technology could be exploited in the future to design more dangerous pathogens or to develop biological weapons.
The researchers relied on a program called “Evo,” a model that operates similarly to ChatGPT, having been trained on vast amounts of genetic sequence data from animals, plants, viruses, and microbes.
Because generative AI models rely on analyzing patterns and predicting word sequences in text, scientists applied the same principle to analyze the genetic code, predict new genetic sequences, and generate them.
During the training process, the Evo program scanned approximately nine trillion nucleotides—the fundamental building blocks of DNA—and used this data to generate new genetic codes capable of performing specific functions.
Scientists chose to start with viruses because they are far less complex than the human genome, which represents the complete set of genetic instructions found within cells.
Samuel King, a graduate student at Stanford University and one of the study’s authors, told The New York Times, “It seemed like the next logical step.”
After the Evo program learned the genome of the Phi X-174 virus—which was selected because it cannot infect humans or animals—researchers used it to produce new versions of this virus.
Of the nearly 300 viral genomes designed by AI, only 16 proved to be viable.
When tested in the lab, researchers discovered that some of these viruses reproduced faster than the original Phi X-174 virus, exiting host cells more quickly after replication, leaving dead cells behind.
Professor Mark Guillemin from the Synthetic Biology Laboratory at Pompeu Fabra University in Spain described the study as “a highly significant turning point,” because it “gives us the opportunity to dream of promising possibilities to address the greatest challenges facing humanity.”
He added, “For the first time in history, we have begun designing biology using computers.”
Researchers emphasized that they did not provide the AI model with any data related to viruses that infect humans, and they also excluded similar viruses that infect animals, plants, and fungi, in an effort to mitigate potential risks.
Nevertheless, experts have warned that this achievement raises fundamental questions regarding safety and biosecurity.
In a companion scientific article to the study, Dr. Thomas Inglesby and Dr. Moritz Hanke from the Johns Hopkins Center for Health Security stated that the findings raise “urgent questions concerning biosafety and biosecurity.”
They clarified that the debate is no longer about whether AI-driven viral genome design technologies will become a reality, but rather about how to ensure their use without “causing serious harm,” emphasizing that “there should be no pursuit of designing viruses that can cause disease.”
