A new artificial intelligence (AI) tool has solved a complex scientific problem in just 48 hours—a puzzle that had taken microbiologists more than a decade to unravel.
Professor José R. Penadés and his team at Imperial College London had spent years investigating why certain superbugs—antibiotic-resistant bacteria—develop immunity. Curious to test the capabilities of AI, Penadés used Google’s experimental tool, known as “co-scientist”, to see if it could replicate their findings.
The AI was given a brief prompt outlining the core issue, and within two days, it reached the same groundbreaking conclusion as the researchers—despite the fact that their study had not yet been published.
Speaking to the BBC’s Today programme, Prof. Penadés admitted he was “shocked” by the AI’s success. “I was shopping with somebody, I said, ‘please leave me alone for an hour, I need to digest this thing,’” he recalled. Concerned that the AI might have accessed his unpublished data, he emailed Google to ask if it had somehow tapped into his computer. The company assured him it had not.
Cracking the Superbug Code
The researchers had hypothesized that superbugs become resistant by acquiring “tails” from viruses, which act like “keys” that allow the bacteria to spread between different species. This theory, developed exclusively within their lab, had never been shared or published. Yet, Google’s AI not only identified this exact mechanism but also proposed four additional hypotheses, all of which, according to Penadés, “made sense.” One of these new ideas was so compelling that the team is now actively exploring it.
While the scientists spent years conducting experiments to prove their theory—work the AI cannot replace—Penadés acknowledged that having the correct hypothesis from the outset would have saved them years of trial and error.
A Game-Changer for Science
The rapid success of AI in solving scientific problems has sparked both excitement and concern. Critics fear that such tools could replace human jobs, but Penadés sees it differently. “When you think about it, it’s more that you have an extremely powerful tool,” he explained.
He believes AI will transform scientific research by accelerating discoveries and reducing the time spent on trial-and-error processes. “This will change science, definitely,” he said, comparing his experience with the AI to playing in a “Champions League match.”
While the technology cannot replace the need for human experimentation and validation, Penadés and his team are confident that AI will become an invaluable partner in scientific research, opening the door to faster breakthroughs in fields ranging from microbiology to medicine.
