DeepMind Announces Plans to Utilize AI Models for Physical Robotics
Google DeepMind Unveils Gemini 2.0: Pioneering AI Models for Robotics
Introduction to Gemini 2.0
On Wednesday, Google DeepMind introduced two innovative AI models tailored for robotics, both powered by the new Gemini 2.0 platform, which Google describes as its "most capable" AI system to date. This marks a significant step for Google as it transitions its advanced AI capabilities from digital formats to physical robots that can perform various tasks.
New AI Models
The two AI models unveiled are named Gemini Robotics and Gemini Robotics-ER (Extended Reasoning). Unlike traditional generative AI that focuses on creating text and images, these models have been designed specifically to issue physical action commands for robotic control. Their exceptional range enables them to adapt to different environments and tasks seamlessly.
Collaboration with Apptronik
As part of this venture, Google announced a partnership with Apptronik, a Texas-based robotics developer. This collaboration aims to develop the "next generation of humanoid robots" using Gemini 2.0. Apptronik has a notable background, having previously collaborated with organizations like Nvidia and NASA. In a recent funding initiative, Google participated in a $350 million investment round aimed at furthering Apptronik’s capabilities in this field.
Demonstration of Capabilities
Google showcased the practical applications of its AI models through demonstration videos featuring Apptronik robots. These robots successfully performed a range of tasks, including:
- Plugging devices into power strips
- Filling lunchboxes
- Moving plastic vegetables
- Zipping up bags
These tasks were performed in response to spoken commands, illustrating the potential of AI to enhance robot interactivity and utility.
Key Qualities of AI Models for Robotics
Google emphasizes three principal qualities that AI models must possess to be effective in robotics:
- Generality: The ability to adapt to a variety of situations and environments.
- Interactivity: Quick understanding and response to instructions or changes in surroundings.
- Dexterity: The skill to manage tasks like humans, such as delicately manipulating objects.
Advanced Tools for Roboticists
The Gemini Robotics-ER model is tailored for robotic developers, providing a robust foundation for creating customized AI models. It will be available to Apptronik and a select group of "trusted testers," including companies like Agile Robots, Boston Dynamics, and Enchanted Tools. This initiative is intended to propel forward the development of advanced robotics.
Broader AI and Robotics Landscape
Google is not alone in exploring the intersection of AI and robotics. In November, OpenAI invested in Physical Intelligence, a startup dedicated to integrating general-purpose AI into the physical world. This company focuses on creating large-scale AI models and algorithms for robotic applications.
Additionally, OpenAI has expanded its robotics efforts by recruiting the former lead of Meta’s Orion augmented reality initiative to guide its robotics and consumer technology projects. Companies like Tesla are also diving into the humanoid robotics realm with their Optimus robot, signaling a widespread interest in this technology.
Insights from Google’s CEO
In a recent statement, Google CEO Sundar Pichai expressed that the company views robotics as an essential area for testing AI innovations. He indicated that the robots equipped with Gemini 2.0 would leverage multimodal AI capabilities to adapt to their environments dynamically, thus enhancing their practical utility.
This new venture into robotics by Google showcases the potential synergies between advanced AI and robotic technology, paving the way for future developments in this exciting field. With Gemini 2.0 at the forefront, the possibilities for automation and interaction in everyday tasks appear to be limitless.
