The development of these models will rely on a diverse range of datasets, including the largest collection of standardized single-cell datasets from the Chan Zuckerberg CELL by GENE software tool, resources from CZ Science research institutes, and large-scale imaging datasets from the Chan Zuckerberg Institute for Advanced Biological Imaging. The system will train AI large language models (LLMs) on human cells, which are expected to outperform human experts and provide faster insights into complex cell structures.
Key takeaways:
- The Chan Zuckerberg Initiative (CZI) has announced plans to build a state-of-the-art GPU cluster that will use artificial intelligence (AI) to develop predictive models for both healthy and diseased cells.
- The project will leverage a diverse range of datasets, including the largest collection of standardized single-cell datasets, resources from CZ Science research institutes, and large-scale imaging datasets from the Chan Zuckerberg Institute for Advanced Biological Imaging.
- The system will train AI large language models (LLMs) on human cells using over 1,000 GPUs. These models are expected to outperform human experts and provide faster insights into complex cell structures.
- Jeff MacGregor, Vice President of Communications at CZI, highlighted the potential of AI models to derive insights and conclusions that surpass the capabilities of human experts, and to accomplish tasks in weeks that might take a team of experts years to achieve.