Google's AI solved 10-year-old scientific problem within two days

Scientists at Imperial College London have spent 10 years of their lives trying to figure out how some superbugs become resistant to antibiotics, a significant threat that claims millions of lives every year. However, as soon as they asked this question to Google's AI tool, it provided an answer that matched their unpublished findings in just two days.

Live Science writes about it.

How Google's AI coped with the task within two days

The principal investigator sent an email to Google to check whether they had access to his research and was told they did not.

Scientists have been working on the problem of antimicrobial resistance, when bacteria, viruses, fungi, and parasites become resistant to antibiotics, making medicines ineffective. A team of scientists led by microbiologist José Penadés began to look for ways in which the cf-PICI family of superbugs acquires the ability to infect different types of bacteria.

The scientists suggested that the viruses use their "tails" to introduce the viral genome into the bacterial host cell. The experiments confirmed their hypothesis, revealing a breakthrough mechanism of horizontal gene transfer that none of the scientists had known about before.

Before sharing their findings publicly, they engaged Google's AI to collaborate with the scientists. Two days later, it provided answers, one of which was correct.

"This effectively meant that the algorithm was able to look at the available evidence, analyse the possibilities, ask questions, design experiments and propose the very same hypothesis that we arrived at through years of painstaking scientific research, but in a fraction of the time," Penadés said.

While using AI from the outset would not have eliminated the need for experiments, it would have helped them come up with a hypothesis much faster, saving years of work.

Despite this and other findings, the use of AI is still a controversial issue in science. Sometimes, the research provided by AI turns out to be unreproducible or false. Therefore, scientists are developing tools to detect AI misbehavior and establish an ethical framework to assess the accuracy of the results.

As a reminder, Google's Gemini 2.0 AI model has been used to remove image watermarks. Users online noticed that AI has no limitations on copyright infringement.

We also wrote that developers from Google DeepMind unveiled an AI model for robots. Gemini Robotics is designed to control machines that can perform complex multi-stage tasks requiring precise manipulations.