DeepMind Plans to Utilize AI Models for Driving Physical Robots
Google Unveils Advanced AI Models for Robotics
Introduction to Gemini 2.0
On Wednesday, Google DeepMind announced two innovative artificial intelligence (AI) models for robotics, powered by Gemini 2.0, which the company describes as its most advanced AI system to date. These new models, Gemini Robotics and Gemini Robotics-ER (Extended Reasoning), aim to extend the capabilities of AI beyond text and images into controllable physical actions, significantly advancing the field of robotics.
Collaboration with Apptronik
To further develop this technology, Google has partnered with Apptronik, a Texas-based robotics firm known for its collaboration with prominent organizations such as Nvidia and NASA. Recently, Google participated in Apptronik’s funding round, which raised $350 million. This partnership is focused on creating the next generation of humanoid robots, utilizing the advanced capabilities of Gemini 2.0.
Demonstrations of Robot Functionality
In demonstration videos, Google showcased Apptronik’s robots utilizing the new AI models. The robots performed various tasks, including:
- Plugging in devices to power strips
- Filling up lunchboxes
- Moving plastic vegetables
- Zipping up bags
These actions were carried out in response to spoken commands, illustrating the AI’s ability to interact with and adapt to its environment efficiently. Google has not yet revealed a timeline for when these capabilities will be commercially available.
Key Features of the AI Models
According to Google, for AI models designed for robotics to be truly effective, they need to possess three essential qualities:
General Adaptability: This means the robots need the ability to handle various situations effectively.
Interactive Response: Robots must be able to quickly understand and respond to instructions or changes in their surroundings.
- Dexterity: They should mimic the fine motor skills and hand-like functions that humans typically perform.
Gemini Robotics-ER for Developers
Gemini Robotics-ER serves as a foundational tool for roboticists to train their own models. This AI model is accessible to Apptronik and several trusted testers, including companies like Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools. This effort aims to create a diverse ecosystem of robotics applications.
The Larger AI and Robotics Landscape
Google is not the only player in the AI and robotics space. In November, OpenAI made a significant investment in a startup called Physical Intelligence, which focuses on integrating general-purpose AI into robotics. OpenAI’s engagement in this sector reflects a broader trend of established tech companies venturing into robotic technology development.
Moreover, Tesla is also making strides in humanoid robotics with its Optimus robot, showcasing the industry’s rapid evolution.
Google’s Vision
Sundar Pichai, the CEO of Google, expressed that the company views robotics as an essential platform for applying AI advancements in real-world scenarios. The robots developed through their collaboration with Apptronik will utilize Google’s multimodal AI models, enabling them to adapt and modify their actions based on real-time sensory information.
In summary, Google DeepMind’s launch of the Gemini 2.0 AI models and its partnership with Apptronik marks a significant step forward in the integration of AI with physical robotics. The emphasis on adaptability, interactivity, and dexterity reflects the increasing sophistication of robotic technologies and their potential future applications in everyday life.
