Some microbiologists have taken 10 years to solve a problem. The AI ​​has solved it in two days

Professor José R. Penadés and his Imperial College team in London had been working on a way of demonstrating how certain superbacteria are immune to antibiotics. After all that time they had finally achieved a conclusion in this regard. They recently dedicated to give AI a chance to solve the same problem. They got a surprise.

Ten years they, 48 hours the AI. The scientists used a new Google AI tool called co-scientist to provide a brief prompt asking him about the essential problem they had been investigating for ten years. The AI ​​reached the same conclusion as them, but it took only 48 hours to reach it.

And the AI ​​did not know the response in advance. One of the important details here is that AI could not have found out the solution looking for it on the Internet. Penadés and her team had not published the study, so Co-Scientist could not find the answer because it was not public domain.

“I need to digest it”. The Penadés Professor He indicated in BBC How he learned that AI had reached the same conclusion in just two days. “I was shopping with someone, and I said,” Please leave me only one hour, I need to digest this. ” no.

I could have saved years. Even considering the AI ​​solution as a hypothesis when they started working on the problem, having it would have saved years of work, Penadés explained.

But it was the AI ​​went further. In fact, Google tool had done more than replicating its research. “It was not only that the hypothesis he provided was correct. He provided another four, and all of them made sense. And in one of the cases, he never occurred to us, and now we are working on it.”

A for superbacteria. The study tried to reveal how some superbacteria are created that are resistant to antibiotics. The hypothesis proposed that superbacterial can form a tail from different viruses, and that allows them to spread between species. It is as if they had the “keys” to move from one place to another and “enter” different guests. That is the same hypothesis that Co-Scientist arrived in 48 hours.

“This will change science”. For the Penadés professor “this will change science, sure.” Tools like this, says, “give the opportunity to play a great game. I feel as if I was finally playing a Champions League match with this.”

And there I don’t know the thing: Evo 2 arrives. This discovery binds to launch these days of Evo 2the most large biology -oriented model so far. A group of researchers from the universities of Berkeley, Stanford, the ARC Institute, the UCSF and NVIDIA have trained an AI model with 9.3 billion nucleotides (from which DNA chains are created) from 128,000 Complete genomes. Thanks to this model it is possible to precisely predict all types of genetic mutations.

Image | Wikimedia

In Xataka | Some engineers have simulated 500 million years of evolution with an AI. Now we have a fluorescent protein