IBM and NASA collaborate to release open source AI basic model that can track carbon emissions

IBM recently teamed up with NASA to create a new geospatial-based AI model to improve carbon emissions tracking and monitor climate change. Impact, making it easier to analyze satellite image data and provide more real-time information. The open source model is currently released on the "Hugging Face" platform.

IBM introduced that this geospatial basic model is built on the basis of IBM watsonx.ai platform enterprise technology, which uses images taken by NASA’s Harmonized Landsat Sentinel-2 (HLS) satellite over a year Conduct training. With fine-tuning, this basic model can be redeployed for tasks such as tracking deforestation, predicting crop yields, or detecting and monitoring greenhouse gases

NASA and IBM collaborated to develop a fine-tuned model that uses traces of wildfires across the United States Conduct training. Based on a pre-trained base model, the model can perform analysis using 75% less labeled data than current state-of-the-art models, IBM said. This will greatly improve the ability to track and predict wildfires, and make the model itself run more efficiently

It is reported that researchers from IBM and NASA are working with Clark University to further improve the open source model to adapt to more A wide range of application areas, including time series segmentation and similarity studies

One of the biggest problems facing businesses and climate scientists is the lack of labeled data or data in accessible formats. A study published earlier this year by Microsoft and Tata Consultancy Services found that 80% of companies did not disclose operational carbon emissions targets. This is partly due to a lack of data on the entire supply chain and global trends. IBM believes AI can help simplify this process.

Although this base model is trained on a large dataset of unlabeled data, they can be fine-tuned to fit specific use cases and deployed with labeled data. Therefore, the geospatial model released by IBM can be re-tuned based on enterprise information or data for specific scientific purposes to improve analytical capabilities

Although this model application has been released to allow developers to freely shareAI model on the Hugging Face platform, but a commercial version will also be released on watsonx.ai later this year, making it easier for businesses to use it to achieve carbon emissions tracking and net zero goals.

Sriram Raghavan, Vice President of IBM AI Research Institute, suggests that by combining the foundational models of flexible and reusable AI systems with NASA's Earth satellite data repository and providing them to Hugging Face, we can leverage the power of collaboration to implement faster and more impactful solutions for protecting our planet's environment.

Hugging Face's head of product and growth Jeff Boudier noted: "AI remains a science-driven field, and only through information sharing and collaboration can science To make progress. That's why open source AI and the open release of models and datasets are critical to the continued progress of AI, a technology that will benefit more people."

Kevin Murphy, NASA's Chief Scientific Data Officer, stated that "this foundational model has the potential to revolutionize the analysis of observational data. By providing open-source access to this model globally, we aim to amplify its impact."

The above is the detailed content of IBM and NASA collaborate to release open source AI basic model that can track carbon emissions. For more information, please follow other related articles on the PHP Chinese website!