TL;DR:
- IBM and Hugging Face partner to release geospatial AI foundation model.
- The model, built from NASA’s satellite data, is the largest of its kind on Hugging Face.
- Open-source availability accelerates climate and Earth science innovations.
- The collaboration aims to democratize AI access and foster knowledge sharing.
- The model demonstrates significant improvement with half the labeled data.
- Potential applications include deforestation tracking, crop yield prediction, and greenhouse gas monitoring.
- NASA declares 2023 as the Year of Open Science to promote data sharing.
- The geospatial model will be available through IBM Environmental Intelligence Suite.
Main AI News:
In a groundbreaking partnership, tech giant IBM (NYSE: IBM) and open-source AI platform Hugging Face have joined forces to make IBM’s watsonx.ai geospatial foundation model openly accessible on Hugging Face. Notably, this will be the largest geospatial foundation model available on the platform and marks the first-ever open-source AI foundation model created in collaboration with NASA.
Climate science, a domain where environmental conditions evolve rapidly, faces a significant challenge in accessing the latest data. Despite the abundance of data—NASA estimates a staggering 250,000 terabytes of data from new missions by 2024—scientists and researchers encounter difficulties in analyzing these vast datasets. To address this, IBM initiated the development of an AI foundation model for geospatial data earlier this year, under a Space Act Agreement with NASA. Now, by offering this geospatial foundation model on Hugging Face, which is renowned for hosting various transformer models, the effort aims to democratize access and application of AI, fostering new innovations in climate and Earth science.
“The essential role of open-source technologies to accelerate critical areas of discovery such as climate change has never been clearer,” emphasized Sriram Raghavan, Vice President of IBM Research AI. He further highlighted the collaboration’s strength in combining IBM’s expertise in creating flexible, reusable AI systems with NASA’s extensive repository of Earth-satellite data.
Jeff Boudier, Head of Product and Growth at Hugging Face, echoed this sentiment, emphasizing the scientific nature of AI and the importance of information sharing and collaboration in its advancement. Open-source AI and the release of models and datasets, according to Boudier, are fundamental to the continuous progress of AI, ensuring it benefits the maximum number of people.
NASA’s Chief Science Data Officer, Kevin Murphy, expressed confidence in the potential of foundation models to revolutionize observational data analysis, contributing to a better understanding of our planet. By open-sourcing such models, NASA aims to amplify their impact and foster widespread benefits.
The model, a result of joint efforts by IBM and NASA, underwent training using Harmonized Landsat Sentinel-2 satellite data (HLS) over a year, covering the continental United States. It was further fine-tuned using labeled data for flood and burn scar mapping, achieving a 15 percent improvement over state-of-the-art techniques with only half the labeled data. The versatility of the model allows for additional fine-tuning, enabling its application in tasks such as deforestation tracking, crop yield prediction, and greenhouse gas detection and monitoring. IBM and NASA researchers are actively collaborating with Clark University to adapt the model for various applications, including time-series segmentation and similarity research.
This collaboration follows IBM’s previous announcement of partnering with NASA to develop an AI model that accelerates satellite image analysis and supports scientific discovery. It also aligns with NASA’s decade-long Open-Source Science Initiative, aiming to create a more accessible, inclusive, and collaborative scientific community. In sync with the spirit of open science, NASA, along with the White House and other federal agencies, declared 2023 as the Year of Open Science, celebrating the benefits and successes derived from the open sharing of data, information, and knowledge.
Leveraging IBM’s foundation model technology, this initiative aligns with the company’s broader mission to create and train AI models for diverse tasks, enabling knowledge transfer across different scenarios. Additionally, IBM recently introduced watsonx, an AI and data platform empowering enterprises to scale and expedite the impact of advanced AI using reliable data. The geospatial model, as part of IBM watsonx, will also be available in a commercial version through the IBM Environmental Intelligence Suite (EIS) later this year.
Conclusion:
The partnership between IBM and Hugging Face to open source the largest geospatial AI foundation model represents a significant development in the market. This collaboration demonstrates the growing importance of open-source technologies in accelerating critical areas of discovery, such as climate science. With the availability of this model on Hugging Face, the broader scientific community can now leverage AI to analyze vast amounts of geospatial data, facilitating new innovations and solutions that can improve our planet’s environmental understanding. This move also showcases the potential for AI applications in various fields, encouraging enterprises to explore AI-driven solutions for their specific business needs. As organizations increasingly recognize the value of open-source AI and data sharing, it paves the way for greater collaboration, knowledge dissemination, and positive market impacts.
