DeepMind, the Google-owned artificial intelligence company, announced the launch of its AlphaFold 3 model, which can predict the structure of proteins, nucleic acids, and various parts of life. The model is designed to help researchers make potential discoveries in the fields of medicine, agriculture and pharmaceuticals. Development and other fields. , and others.
Previous versions of the model could only predict the structure of proteins, but the new AlphaFold 3 model is also able to predict the structure of DNA and RNA, as well as smaller biomolecules called ligands, increasing the scientific effectiveness of the model.
DeepMind claims that the prediction accuracy of the new model is 50% better than previous versions of the model.
“The AlphaFold 3 model represents an advance in using artificial intelligence to understand and model biology,” Demis Hassabis, CEO of DeepMind, said in a press release, emphasizing the importance of previous models in biological research.
As for how the model works, AlphaFold 3 includes a library of molecular structures. Researchers enter a list of molecules they want to combine, and then the model creates a 3D model of the new structure using a diffusion mechanism, which is the same mechanism. Some artificial intelligence systems are used to generate images, such as stable broadcast.
DeepMind is working on real-world commercial use of its advanced model and has announced a partnership with pharmaceutical company Isomorphic Labs, which is currently using the model in its internal projects to improve understanding of disease and facilitate drug development. For researchers. Access this template for non-commercial academic purposes.
Google said it is working with experts in the scientific community and policymakers to use the model responsibly, and that it is working with experts to understand the risks of the AlphaFold 3 model before publishing it.
Although this model opens up huge potential for scientific research, Google confirms that some biosecurity experts believe that AI models, along with other technologies, could also facilitate the development of pathogens and toxins, increasing the likelihood of them spreading and causing damage. .