BRIEFLY.
AI Model in Orbit
1 min read
Briefly Editorial Team

AI Model in Orbit

TL;DR

  • AI model Prithvi tested in orbit
  • Model showed high efficiency in flood zone recognition

Why it matters

AI model Prithvi can solve a wide range of applied tasks, from disaster monitoring to crop yield forecasting

Technical Details

A consortium of researchers from the University of Adelaide, the European Space Agency (ESA), Thales Alenia Space, and the Australian SmartSat Cooperative Research Centre has successfully tested an open geospatial AI model Prithvi in orbit. The neural network, developed jointly by NASA and IBM, became the first large model in history to be launched directly into space.

Background and Context

The training was based on the harmonized archives of Landsat and Sentinel-2 - a global array of images over a decade, obtained from NASA and ESA satellites. The ground version of the model has already proven itself in assessing the scale of the catastrophic flood in Valencia in the fall of 2024.

Industry Impact

The deployment of such systems directly in orbit solves a key problem of modern remote sensing - the lack of bandwidth of communication channels. Modern satellites collect gigabytes of raw data, but their transmission to Earth for further analysis takes precious time, critical in forest fires or floods.