IBM and NASA have joined forces in an exciting collaboration that aims to revolutionize climate science using open-source artificial intelligence (AI). With the contemporary world witnessing rapid changes in environmental conditions, the availability of data has surged. NASA projects that by 2024, there will be a staggering 250,000 terabytes of data from upcoming missions. However, analyzing this vast amount of data remains a complex challenge for researchers and scientists.
Earlier this year, IBM and NASA teamed up through a Space Act Agreement to address this challenge by creating an AI foundation model specifically designed for analyzing geospatial data. To further democratize access to AI, IBM has partnered with Hugging Face, a renowned leader in open-source AI and a popular hub for transformer models. The ultimate goal of this collaboration is to stimulate innovations in both Earth and climate science.
Sriram Raghavan, Vice President at IBM Research AI, emphasized the importance of open-source technology in critical areas of research, particularly climate change. He believes that the collaborative effort with NASA’s Earth-satellite data, channeled through Hugging Face’s open-source AI platform, has the potential to deliver transformative solutions for the planet.
Kevin Murphy, NASA’s Chief Science Data Officer, echoed similar sentiments, highlighting the potential of foundation models in revolutionizing the analysis of observational data. By open-sourcing these models, their impact can be significantly enhanced.
The collective efforts of IBM and NASA have already yielded promising results. Over the course of a year, the model was refined using Harmonized Landsat Sentinel-2 satellite data from the US. The results have demonstrated a 15 percent increase in efficiency compared to traditional techniques, all while utilizing only half of the labeled data. With further fine-tuning, this base model holds promise in various domains, from deforestation tracking to predicting agricultural yields and monitoring greenhouse gas emissions. In fact, a collaboration with Clark University is already underway to adapt the model for niche applications such as time-series segmentation and research on similarities.
This achievement builds on IBM’s previous announcement of their collaboration with NASA, which aims to accelerate the analysis of satellite images and drive scientific discoveries. It is also aligned with NASA’s long-term Open-Source Science Initiative, which envisions a more accessible scientific community that encourages collaboration. In line with this vision, NASA, the White House, and other federal institutions have designated 2023 as the Year of Open Science, emphasizing the numerous advantages of open data and knowledge sharing.
IBM’s current model leverages the company’s foundational technology and aligns with their broader vision of developing AI models that cater to diverse tasks, seamlessly transferring information across different contexts. In June, IBM introduced watsonx, an AI-data platform designed to amplify the impact of AI on enterprises. As part of this, a commercial version of the geospatial model integrated with IBM watsonx is set to launch through the IBM Environmental Intelligence Suite (EIS) in the coming months.
In conclusion, the collaboration between IBM and NASA, in conjunction with Hugging Face, is poised to revolutionize climate science by leveraging open-source AI. By providing researchers and scientists with access to advanced AI tools, this partnership aims to drive innovation and deepen our understanding of the Earth. With the increasing availability of data and the democratization of AI, the possibilities for transformative solutions to global challenges like climate change are endless.