Google, Gemini Robotics 2.0 공개 — 향상된 민첩성과 안전성 약속
Google/DeepMind
Google DeepMind가 Gemini Robotics 2.0을 공개했습니다. 이는 로봇 제어를 위한 새로운 모델 모음으로, 로봇이 더 복잡한 작업을 수행하고, 변화하는 환경을 분석하며, 다른 로봇과 상호작용할 수 있게 해줍니다. 핵심 구성 요소인 Gemini Robotics ER 2는 Gemini Live API를 통해 개발자에게 제공되며, 이전 버전보다 비디오 이해 능력이 크게 향상되었습니다.
Google has unveiled Gemini Robotics 2.0, a suite of models that enable robots to perform more complex tasks, analyze changing environments, and collaborate with other robots. At its core is an embodied reasoning model called Gemini Robotics ER 2, which represents a significant improvement over the previous version 1.6. This model is integrated with the Gemini Live API, allowing developers to leverage its capabilities. ER 2 is a vision-language model (VLM) capable of processing live video streams from a robot's cameras to track progress. Google claims that ER 2 can classify the completeness of video segments with nearly 60% accuracy, outperforming competing AI models. The project aims to create a general-purpose robot capable of performing any human action, sometimes referred to as 'physical AGI'.
출처: Ars Technica —
원문