Can Google’s new research assistant AI give to scientists to ‘superpowers’?

Google has unveiled an experimental artificial intelligence system that uses advanced arguments to “synthesize the vast amounts of literature, generate novel hypotheses, and suggest detailed research plans”, according to its press release. In Google, Alan Kartikesalingam says, “(AI) is to give the viewers to the superpowers.”

The device, which has no official name yet, makes Gemini’s Gemini on large language models. When a researcher asks a question or specifies a goal – to find a new drug, says – the tool comes with initial ideas within 15 minutes. Many Gemini agents then “argue” these hypotheses with each other, rank them and improve them in the next hours and days, Vivek Natarajan in Google.

During this process, agents can discover scientific literature, use databases and use devices such as Google’s alphabet system to predict protein composition. “They constantly refine ideas, they debate ideas, they criticize ideas,” Natarajan says.

Google has already made the system available for some research groups, which have issued short letters describing its use. The teams that tried are excited about its ability, and these examples suggest that the AI ​​will be helpful to synthesize co-scientific conclusions. However, it is a matter of debate whether examples support the claim that the AI ​​novel can generate hypothesis.

For example, Google says that a team potentially used the system to find “new” methods to treat liver fibrosis. However, drugs proposed by AI have been first studied for this purpose. “All identified drugs are installed to be well antifibrotic,” says Steven o’reli In the UK Biotech Company Elcomics. “Nothing is new here.”

While this possible use of treatment is not new, team members Gary peltz Stanford University School of Medicine in California states that two of the three drugs selected by AI co-scientists showed promise in trials on human liver orgonoids, while neither both were personally chosen-to support their choice Despite being more evidence for. Peltz says Google gave him a little money to cover the costs of tests.

In another paper, Jose pendes The Imperial College London and his colleagues have explained how co-scientists proposed a hypothesis matching an unpublished discovery. She and her team study mobile genetic elements – DNA bits that can move through various means between bacteria. Some mobile genetic elements hijack the bacteriophage virus. These virus have a shell with DNA plus a tail that binds specific bacteria and injects DNA in it. Therefore, if an element can be found in the shell of a phase virus, it receives a free ride for another bacterium.

A type of mobile genetic element makes its own shells. This type is particularly broad, which surprised the Pendes and its team, as any type of phase viruses can only infect a narrow range of bacteria. Answer, he recently discovered, it is that these shells can hook with a tail of different stages, allowing mobile elements to reach a wide range of bacteria.

While the discovery was still unpublished, the team asked AI co-scientists to convince the puzzle-and its number one suggestion was stealing the tail of different stages.

“We were surprised,” says Penade. “I sent an email to Google, which has been said, you have access to my computer. Is that right? Because otherwise I can’t believe what I am reading here. ,

However, the team published a paper in 2023 – which was fed to the system – how this family of mobile genetic elements was “Bacteriophage steals the tail to spread into nature”. At that time, the researchers thought that the elements were limited to obtaining the tail from the steps that infect the same cell. Only later they came to know that the elements can also lift the tail floating around the outside cells.

So an explanation of how the AI ​​co-scientist has come up with the correct answer, it missed the obvious limit that stopped humans from receiving it.

It is clear that it was fed everything necessary to find the answer rather than a completely new idea. “Everything was already published, but in various bits,” Penade says. “The system was capable of keeping everything together.”

The team already tried other AI systems in the market, which none of which came with an answer, they say. In fact, some managed it even when describing the answer and fed the paper. “The system suggests things that you never thought of,” says pencas that have not received any funding from Google. “I think it will be a game-changing.”

Will it be really clear over the game-changing time. The track record of Google comes when it comes to claiming the AI ​​tool to help scientists to mix. Its alphabet system is living for publicity, winning the team behind the Nobel Prize last year.

In 2023, however, the company announced that Around 40 “new materials” Its ganom was synthesized with the help of AI. Nevertheless, according to 2024 analysis Robert Palgrav University College London, One of the synthesized material was not really new,

Despite his findings, Palgrav feels that AI can help scientists. “In general, I think AI has a large amount to contribute to science if it is applied in collaboration with experts in the respective fields,” they say.