DeepMind Announces Plans to Utilize AI Models for Enhancing Physical Robots
Google DeepMind Unveils New AI Models for Robotics
On Wednesday, Google DeepMind introduced two innovative AI models specifically designed for robotics, powered by their advanced Gemini 2.0 framework. This latest version is marketed as the company’s most robust AI iteration to date.
Introduction of Gemini Robotics Models
The new AI models, named Gemini Robotics and Gemini Robotics-ER (Extended Reasoning), expand the functionalities of generative AI beyond typical outputs like text and images. These models focus on controlling robots with physical actions, making them capable of executing real-world tasks.
Partnership with Apptronik
To enhance their robotic capabilities, Google is collaborating with Apptronik, a Texas-based robotics developer. This partnership aims to design the next generation of humanoid robots utilizing the Gemini 2.0 platform. Apptronik has previously worked with well-known organizations like Nvidia and NASA, showcasing its capability and reputation in the robotics sector. Recently, Google also participated in Apptronik’s $350 million funding initiative, indicating a significant commitment to advancing robotic technology.
Demonstration of Capabilities
In showcase videos, Google displayed Apptronik robots operating using the new AI models. These robots demonstrated impressive capabilities, such as:
- Plugging items into power outlets
- Filling lunchboxes
- Manipulating plastic vegetables
- Closing bags
These actions were executed in response to verbal commands, proving the effectiveness of the AI’s real-time response system. However, Google did not specify when the technology might be available for widespread use.
Essential Qualities for Robotic AI
Google outlined three key qualities that AI models for robotics must possess:
- Generalization: The ability to adapt to a variety of situations.
- Interactivity: Quick understanding and response to instructions or environmental changes.
- Dexterity: The capability to perform tasks that require fine motor skills, much like how humans use their hands.
Gemini Robotics-ER for Developers
Gemini Robotics-ER serves as a framework for roboticists to build and train their specific models. This tool is accessible to Apptronik as well as trusted partners such as Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools.
The Competitive Landscape in AI Robotics
Google is not alone in exploring the intersection of AI and robotics. In a recent announcement, OpenAI made a significant investment in Physical Intelligence, a startup focused on developing large-scale AI models and algorithms for practical robotic applications. This startup emphasizes adapting general-purpose AI into the physical realm.
In addition, OpenAI has recently recruited a leader from Meta’s augmented reality initiative to spearhead its robotics and consumer hardware projects. Competitor Tesla is also venturing into the humanoid robotics space with their Optimus robot, further intensifying competition in this field.
Future Perspectives from Google
Google CEO Sundar Pichai emphasized the importance of robotics as a testbed for realizing AI advancements in tangible applications. He stated that the robots would leverage Google’s multimodal AI models, allowing them to adapt and adjust to varying environments seamlessly.
In this rapidly evolving landscape of AI-powered robotics, companies like Google are setting the stage for transformative changes in how humans and machines interact in everyday life. With ongoing research and development, the future of robotics looks promising.
