By 
On August 1, Booz Allen Hamilton announced the successful deployment and operation of a generative AI large language model (LLM) in space using Hewlett Packard Enterprise’s (HPE) Spaceborne Computer-2 onboard the International Space Station (ISS) National Lab. This LLM is believed to be the first one deployed in space and one day aims to help enable astronauts to use generative AI without depending on Earth-bound internet in the power and communications constrained environment of space.
“Booz Allen is thrilled to be at the cutting edge of this exciting development and is committed to pushing the boundaries of what is possible with AI and other mission-critical technologies in space,” said Chris Bogdan, executive vice president at Booz Allen and leader of the firm’s space business. “Generative AI in space is truly the new frontier and this capability unlocks the potential for on-orbit generative AI to integrate and develop mission-critical solutions and is aligned with Booz Allen’s mission to build human-led AI solutions from which our nation will thrive.”
Taking inspiration from the Wright brothers’ first four flights more than 120 years ago, Booz Allen, in coordination with HPE, successfully uploaded the LLM to the ISS National Lab as part of a forward-leaning payload experiment. Once uploaded, the team repeated the experiment with new queries four times within the command window with repeatable responses resulting in a modern-day Kitty Hawk moment.
“When milliseconds matter, on-orbit AI becomes a must-have,” said Dan Wald, principal AI solutions architect for space applications at Booz Allen. “This patent-pending proof-of-concept demonstrates Booz Allen’s ability to deploy state-of-the-art generative AI techniques, such as retrieval-augmented generation, by compressing and optimizing a containerized solution to run reliably in resource-constrained computing environments at the edge. If we can deploy generative AI in space, we can deploy it anywhere.”
Developed rapidly over eight weeks, this LLM application, which builds upon years of extensive infrastructure investments from both HPE and the ISS, can play a new and critical role in providing remote data ingestion and retrieval-augmented generation, which will help enable edge deployed personnel to efficiently retrieve relevant information, accurately interpret, and solve complex issues using natural language processing at the edge of space. This proof-of-concept plays a paramount role in the deployment of AI and it can be expanded to solve future use cases where such capabilities are needed in disconnected environments, including under extreme conditions on Earth and in space.
“This type of breakthrough result by the Booz Allen team is exactly in line with the mission and purpose of the HPE Spaceborne platform. To make what was previously unattainable not just possible but also deployable,” said Norm Follett, senior director, global marketing at Hewlett-Packard Enterprise.
HPE’s Spaceborne Computer-2 is an award-winning AI Edge focused High Performance Compute platform that provides the international scientific community access to a powerful compute platform in space and constrained environments.
Source: Booz Allen
Stay in the know with breaking news from across the IC and IC contracting landscape by becoming a paid subscriber to IC News. Your support makes our work possible.
