Meta✴ and Booz Allen have announced the joint development of the Space Llama artificial intelligence system for the International Space Station. The system is optimized for launch on relatively weak hardware. It is a modified large language model of Llama 3.2. On the ISS, it will help astronauts conduct complex scientific experiments, and will also handle the primary processing of raw data from the station’s sensors and devices.
HPE Spaceborne Computer-2 on board the ISS. Image source: NASA
Back in January 2024, new x86 servers from HPE were delivered to the ISS: Edgeline EL4000 with one GPU and HPE DL360 Gen10. To accumulate data, a 130 TB Kioxia SSD unit was added to the platform. Later, in August, Booz Allen Hamilton announced that it had launched a large language model on this equipment – artificial intelligence that understood natural language input for communicating with people.
The main task of the AI on board is to provide recommendations to astronauts without the need for constant data exchange with Earth. In addition, Space Llama is engaged in the primary processing of data from onboard sensors, sending ready-made analytics to Earth instead of the original raw streams. This allows for significant savings in traffic, which is always limited on the ISS.
As reported by Meta✴ and Booz Allen, the Space Llama AI is designed to support experiments in the scientific laboratory on the station – ISS National Laboratory (International Space Station National Laboratory).
Meta✴ specifies that astronauts can use Space Llama to extract data from technical reference documents. In addition, the system’s implementation of Llama 3.2 allows for the processing of multimodal files thanks to built-in computer vision functions. The AI does not require instructions from Earth to operate, meaning it can perform calculations even with limited network bandwidth.
«This initiative is designed to support ISS National Laboratory researchers in a wide range of scientific projects and represents an important step towards the exploration of the Moon and Mars, the implementation of modern satellite and drone capabilities, and the creation of next-generation autonomous systems,” explained Meta✴ employees.