오픈소스 AI, 국제우주정거장으로 — Llama 3.2 배포
무슨 발표인가
- Booz Allen, Llama 3.2를 ISS National Lab 배포
- 인터넷 없이 오프라인 추론 가능
- 보안·비용·적응성 이점
국제우주정거장(ISS)
원문 (영어)
Our open source AI model, Llama, is going to space. In a standout case of American innovation, Booz Allen has collaborated with Meta to deploy a fine-tuned version of Llama 3.2 aboard the International Space Station (ISS) National Laboratory, where astronauts in low orbit conduct scientific research not possible on Earth.
Why Llama? As an open source platform, Llama offers greater flexibility and can be downloaded and then deployed on machines without internet access . There’s no need to transfer data through an AI company for processing – helping to secure this highly sensitive environment – or even connect to computers on Earth, which is important since the internet is unavailable on the ISS.
Additionally, Llama’s lower costs and adaptability, made possible by its publicly available “model weights” (the numerical values that determine how Llama makes decisions), are helpful in a setting like the ISS’s National Laboratory, where researchers must be able to respond to new conditions quickly and keep projects on budget.
Transforming Research Processes with Llama The tech stack used for Space Llama brings together Booz Allen’s own A2E2™ (AI for Edge Environments), Hewlett Packard Enterprise’s (HPE) Spaceborne Computer-2, NVIDIA accelerated computing, enabling powerful AI to operate directly on the ISS, and Meta’s Llama vision AI capabilities.
Using a compact, energy-efficient system similar to those used in satellites, the team sped up AI tasks from minutes to just over a second using models fine-tuned with NVIDIA CUDA software and its acceleration libraries, including cuDNN and cuBLAS .
원문: Meta Newsroom — "Space Llama: Meta’s Open Source AI Model Is Heading Into Orbit" (2025-12-16) 공식 원문: https://about.fb.com/news/2025/04/space-llama-metas-open-source-ai-model-heading-into-orbit/