On the 30th (local time), Google's artificial intelligence (AI) organization, DeepMind, unveiled a new AI model, "AlphaEarth Foundations," which analyzes vast Earth observation data and easily represents it in high-quality maps.
The AI model introduced this time helps scientists and researchers map and monitor the Earth more accurately and efficiently by analyzing Earth observation data. DeepMind noted, "'AlphaEarth Foundations' operates like a virtual satellite" and explained, "It integrates petabyte-scale Earth observation data—such as satellite images, radar, 3D (three-dimensional) lidar, and climate simulations—into an 'embedding' form that computer systems can easily process." An embedding refers to converting unstructured data, such as text and images, into numerical formats that machines can understand.
The collected information expresses surface information of the Earth with a resolution of 10x10m. DeepMind identified the fact that "AlphaEarth Foundations" uses 16 times less storage space than existing AI systems as a differentiating factor. This is because it can drastically reduce both the expense and time of analyzing large-scale Earth observation data.
DeepMind explained that this enables scientists to quickly produce high-quality, customized maps whenever needed for research on food security, deforestation, urban expansion, and water resources. A Google representative stated, "There is no longer a need to wait for a specific satellite to pass its orbit when checking crop health or tracking forest destruction," and added, "Scientists will now generate detailed and consistent maps whenever they need them."