Archive — US — techcrunch
Google DeepMind unveils new AI models for controlling robots
Google DeepMind has unveiled Gemini Robotics, advanced AI models for robot control, enabling machines to interact with objects and navigate environments. Demo videos showed robots performing various tasks by responding to voice commands. Gemini Robotics generalizes behavior across different hardware, showcasing effective performance in untrained environments. A simplified model, Gemini Robotics-ER, is available for researchers.
Key points
- Google DeepMind has launched Gemini Robotics, an AI model for controlling robots.
- The robots can perform tasks such as folding paper and handling objects based on voice commands.
- Gemini Robotics is designed to generalize behaviors across various robotics hardware.
- In tests, the robots demonstrated the ability to operate well in environments outside their training data.
- A smaller version, Gemini Robotics-ER, is offered for researchers to develop their own robotics control models.
- DeepMind introduced the Asimov benchmark to assess risks associated with AI-driven robots.
Why it matters
- The announcement aligns with growing trends in AI and robotics, focusing on automation and human-machine collaboration.
- By 2025, there is an increasing emphasis on AI models that improve generalization capabilities, as seen in various tech advancements.
- The development of benchmarks like Asimov reflects a need for ethical considerations and risk assessments in AI applications.
- Similar advancements in AI have been noted in other sectors, pushing the boundaries of machine learning in customer service and manufacturing.
DeepMind's launch of Gemini Robotics marks a significant leap in AI capabilities for robotics, emphasizing the importance of adaptability and real-world application, which will likely shape the future of automation by 2025.