Stay informed with free updates
Simply sign up for the Artificial Intelligence myFT Digest, delivered straight to your inbox.
Google DeepMind and BioNTech launch AI lab to help researchers plan scientific experiments and predict outcomes more accurately as companies race to discover specialized applications for energy and data-intensive artificial intelligence models Building an assistant.
Google’s head of AI, Sir Demis Hassabis, has developed a specialized AI model to act as a research assistant to help scientists collaborate across disciplines and make unexpected connections more easily. He is leading the company’s efforts to
At a recent Nobel Foundation event, he said that “a revolution is occurring” in biology as a result of AI software.
“We’re like research assistants, working on scientific large-scale language models that maybe help predict… the results of experiments,” Hassabis said.
Over the next few years, he said, the tools DeepMind is building will be able to propose and design experiments based on a given hypothesis, and provide scientists with probabilistic information about the potential success or failure of a proposed experiment. He said that he could give his opinion.
Meanwhile, German pharmaceutical company BioNTech and its London-based AI subsidiary InstaDeep announced on Tuesday that they are building a specialized AI known as Laila with “in-depth knowledge of biology” built on Meta’s open source Llama 3.1 model. He announced that he had designed an assistant.
In a live demonstration, research scientist Arnu Pretorius will show how AI agents can automate routine scientific tasks in experimental biology, such as analyzing and segmenting DNA sequences and visualizing experiment results. I did.
Scientists at the BioNTech Institute in Mainz also demonstrated how Lyra connects to experimental equipment and monitors ongoing experiments and tasks performed by the robot, and during a live demonstration an assistant was able to monitor the BioNTech machine. A mechanical failure was detected.
“We do not believe that full AI automation will be achieved in the near future. We do not believe that we will achieve full AI automation in the near future. InstaDeep CEO Karim Beguir told the Financial Times.
InstaDeep also unveiled an AI model that helps identify and discover new targets to tackle cancer, at the company’s first technology launch since BioNTech acquired InstaDeep for up to £500 million in 2023.
Beguir acknowledges that rival companies like DeepMind could also develop AI assistants, but having InstaDeep’s technology “under the same roof” as BioNTech’s biological expertise is , described it as a “catalyst” for AI adoption and “unique” in the pharmaceutical sector.
The new scientific assistants come as technology companies spend billions of dollars on AI models and products, believing the technology can transform industries from health care to energy to education.
The wave of AI innovation in science has so far focused on predicting new and useful drug candidates. However, the bottleneck in bringing new treatments to market remains conducting real-world experiments, which are the gold standard for scientific research.
The goal of AI research assistants is to simplify this process by planning experiments more effectively, for example by selecting the most promising among a set of possible experiments.
Companies like Google and Microsoft are using large-scale language models (software that can generate text, code, images, and even DNA and molecular sequences based on large training datasets) to accelerate scientific progress. is being adapted.
In 2022, DeepMind designed an AI system known as AlphaFold that can predict the shape of almost any known protein. This could solve a 50-year-old scientific challenge and significantly reduce the time needed for biological discoveries.
Nobel Prize-winning geneticist Paul Nurse said at a Nobel event in March that members of his lab “always use AlphaFold” in biochemistry experiments, adding that the AI model’s output “is always correct. “It’s not necessarily true, but it’s quite correct,” he added. It’s a great tool. ”
Recommended
Hassabis has since spun off the research into an AI pharmaceutical spinoff known as Isomorphic Labs, a group whose advisory board includes Nurse, and agreed to a partnership worth up to $3 billion with Eli Lilly and Novartis. did.
Microsoft’s AI4Science Research division also leverages LLM to accelerate scientific discovery. Its director, Chris Bishop, said at this year’s research forum that one of the notable properties of the LLM is that it “can act as an effective inference engine,” which is particularly useful in science. said.
Bishop said the team, in collaboration with the Global Health Drug Discovery Institute, used LLM to discover new molecules to more effectively treat tuberculosis.
Video: Content creators take on AI | FT Tech
Source link
