Google DeepMind unveils Gemini Robotics On-Device, a powerful AI model enabling robots to function autonomously without cloud connectivity. This marks a significant advancement in on-device robotics, enabling faster, more adaptable robots for real-world applications.
Google DeepMind has taken a significant stride in on-device robotics with the launch of Gemini Robotics On-Device, a powerful AI model that empowers robots to operate autonomously without relying on cloud connectivity . This innovation represents a major leap towards deploying fast, adaptable, and versatile robots in real-world scenarios. Gemini Robotics On-Device incorporates the multimodal reasoning capabilities of Gemini 2.0 directly into robots, eliminating the need for internet access.
Designed for applications where latency is crucial and connectivity is limited or unreliable, this model enables robots to execute complex tasks directly on the device. Its strengths lie in its robust task generalization, natural language understanding, and fine motor control.Developers seeking access to this groundbreaking technology can now register through Google's trusted tester program. Furthermore, Google is releasing a comprehensive software development kit (SDK) to empower experimentation, customization, and the development of innovative applications. Notably, Gemini Robotics On-Device operates entirely on the robot itself, unlike its cloud-connected predecessor. This on-device processing allows for faster response times and enhanced reliability, especially in offline or restricted settings. The model is capable of performing a range of tasks without prior knowledge and can adapt to new tasks with minimal data, requiring just 50 to 100 demonstrations. Trained on Google's ALOHA robot, Gemini Robotics On-Device has been successfully adapted to various platforms, including Apptronik's Apollo humanoid and the bi-arm Franka FR3. It demonstrates remarkable dexterity in handling intricate actions such as unzipping bags and folding clothes with smooth, low-latency inference. This release marks the first time a DeepMind robotics model is open for fine-tuning by developers. This capability allows users to further customize the model's behavior by tele-operating the robot to complete a task a few times, enabling the model to autonomously perform that task subsequently.While the on-device version excels in areas like speed, privacy, and offline operation, it currently lacks the built-in semantic safety reasoning present in the hybrid model. Google emphasizes the importance of developers replicating the internal safety stack, including integrating the Gemini Live API and connecting to low-level safety-critical controllers. Safety remains a paramount concern for Google, and they are actively working with selected developers to assess real-world safety risks associated with the technology. Ultimately, Google aims through this release to empower the robotics community to develop adaptive, secure, and capable robots that can thrive in diverse environments, pushing the boundaries of what's possible in robotics
AI Robotics On-Device Gemini Deepmind Autonomous Cloud Connectivity Generative AI Task Generalization
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Google DeepMind’s CEO Thinks AI Will Make Humans Less SelfishDemis Hassabis says that systems as smart as humans are almost here, and we’ll need to radically change how we think and behave.
Read more »
Google DeepMind is sharing its AI forecasts with the National Weather ServiceWill Shanklin has been writing about gadgets, tech and their impact on humanity since 2011. Before joining Engadget, he spent five years creating and leading the mobile technology section for New Atlas. His work has also appeared on SlashGear, TechRadar, Digital Trends, AppleInsider, Android Central, HuffPost and others.
Read more »
Google DeepMind’s optimized AI model runs directly on robotsGoogle DeepMind is launching an on-device Gemini Robotics model that can run on robots without an internet connection.
Read more »
Gemini can now watch Google Drive videos for youGoogle’s latest Workspace feature allows users to ask Gemini questions about videos in their Drive without having to watch it themselves.
Read more »
Galaxy S26 might get a surprise AI assistant instead of Google GeminiSamsung and Google are close partners, but the Galaxy S26 might get a surprise AI assistant instead of Gemini.
Read more »
Google’s Gemini AI is coming to the sidebar in Docs, Drive, Gmail, and moreGoogle’s AI assistant will now work across its productivity apps, and Gemini can help you do everything from write emails to save events to your calendar.
Read more »