Google’s DeepMind is taking a huge leap by introducing two AI models–“Gemini Robotics” and “Gemini Robotics ER”–which are a Gemini 2.0-based model designed for robotics and the physical world. With these new models, the tech giant is aiming to enhance robotic dexterity, adaptability, and responsiveness.
Google has partnered with Texas-based robotics firm Apptronik to develop the next generation of humanoid robots powered by Gemini 2.0. Apptronik, known for its collaboration with industry giants such as Nvidia and NASA, recently secured $350 million in funding, with Google joining the investment round last month.
Google’s DeepMind Expands Gemini
According to the announcement, Gemini Robotics has some extraordinary capabilities that are beyond traditional generative AI which only excels in text and image outputs.
The video linked to the blog shows Apptronik’s AI-based robots performing everyday actions; plugging devices into power strips, filling lunchboxes, handling plastic vegetables, and zipping up bags. These all actions were executed in response to spoken commands.
According to Google, effective AI-based robotics must embody three core attributes:
-
- Generality: The ability to adapt to varied environments and tasks.
- Interactivity: A capacity for real-time comprehension and reaction to instructions.
- Dexterity: Precision in handling objects, mirroring human-like motor skills.
Gemini Robotics-ER is specifically designed as a foundational model for roboticists, providing tools to train custom AI solutions. The system is accessible to Apptronik and select partners, including Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools.
However, Google is not alone in this race. OpenAI recently invested in Physical Intelligence, a startup focused on integrating general-purpose AI into robotics. Tesla, too, has ventured into the field with its Optimus humanoid robot.
Sundar Pichai, Google’s CEO, explained the importance of robotics as a proving ground for AI’s real-world applications. In a post on X, he noted that multimodal AI would enable robots to “make changes on the fly + adapt to their surroundings.”
Source: https://deepmind.google/discover/blog/gemini-robotics-brings-ai-into-the-physical-world/
Latest Stories:
