ESA to build digital Chat assistant powered by EO data

ESA to build digital Chat assistant powered by EO data
by Erica Marchand
Paris, France (SPX) Mar 26, 2024

The European Space Agency (ESA), in collaboration with technology partners, is embarking on an ambitious project to develop artificial intelligence (AI) applications designed to transform the way we retrieve information from Earth observation data. This initiative aims to create a digital assistant capable of producing scientifically accurate responses based on verified data, answering complex questions about environmental and geographical phenomena.

One of the project's highlights is the I*STAR platform, funded by the ESA InCubed programme, which leverages AI to automate satellite data acquisition planning in response to natural disasters like earthquakes and volcano eruptions. Similarly, the SaferPlaces tool, also supported by InCubed, utilizes AI to generate flood maps by integrating satellite data with on-the-ground measurements, proving instrumental in disaster response efforts.

The advancement of AI technologies, exemplified by tools like ChatGPT, has underscored the potential of integrating these innovations with Earth observation data. ESA's vision involves creating a text-based query system that employs natural language processing to interpret human inquiries and deliver responses using Earth observation foundation models.

Foundation models represent a significant shift from traditional machine learning approaches by training on large datasets without extensive human labeling. These models, adaptable and potent, form the core of ESA's digital assistant, with ongoing projects like PhilEO aiming to tailor them specifically for Earth observation tasks. PhilEO, which focuses on recognizing environmental features without human intervention, is a testament to the potential of these models in enhancing Earth observation applications.

Moreover, ESA's efforts extend to developing the user interface for this digital assistant, ensuring it can process natural language queries and provide answers through a combination of text and images. Initiatives like the Digital Twin of Earth project have already demonstrated the feasibility of such a multimodal approach.

Giuseppe Borghi, Head of ESA F-lab, emphasizes the importance of these developments, stating that the combination of advanced AI with Earth observation data holds the promise of a revolutionary digital assistant. With projects like PhilEO and the upcoming exploration of natural language processing for Earth observation, ESA is poised to make significant strides in making this vision a reality.