According to the UK's "Nature" website on July 31st, Google has released an "AlphaEarth" artificial intelligence (AI) model, which is like a "virtual satellite" and can integrate massive heterogeneous observation data (including satellite images, radar data, etc.) to track changes in the Earth's land and shallow sea areas. The model has received high praise from researchers, but they have also called on Google to provide more testing details and reminded relevant researchers that they still need to verify the map data generated by the tool themselves. Google claims that this model can be used for geographic mapping "anywhere, anytime". Satellites only obtain single attribute information such as color and cloud cover, while AlphaEarth processes this information into numerical representations called "embedding vectors", allowing researchers to easily retrieve features such as tree species, buildings, and groundwater. Researchers can use this model to find the optimal location for clean energy projects, study the impact of climate change on ecosystems, monitor deforestation, and more. The processed data is encapsulated into squares with a side length of 10 meters, covering the land and coastal areas of the Earth, allowing AlphaEarth to present a more complete, consistent, and detailed picture of Earth's evolution. Google also uploaded a peer-reviewed paper on the arXiv preprint platform. The article states that the model outperforms similar AI models in terms of accuracy and data density, while significantly reducing computation time and energy consumption. Google stated that its "Earth Engine" platform had previously hosted independent datasets on topics such as agriculture, deforestation, and weather. AlphaEarth integrated these datasets together to generate a unified product for scientists to explore. (New Society)