DeepMind Announces Plans to Utilize AI Models for Physical Robots
Google DeepMind Introduces Advanced AI Models for Robotics
New AI Models: Gemini Robotics and Gemini Robotics-ER
On Wednesday, Google DeepMind unveiled two new AI models specifically designed for robotics, powered by Gemini 2.0—the company’s most advanced AI system to date. Unlike traditional generative AI which specializes in processing text and images, Gemini Robotics takes a bold step into the physical realm by issuing commands that enable robots to perform various tasks.
Partnership with Apptronik
Google has announced a collaboration with Apptronik, a robotics firm based in Texas, to develop next-generation humanoid robots utilizing Gemini 2.0. Apptronik has previously partnered with prominent organizations like Nvidia and NASA. This partnership is particularly significant as Google contributed to Apptronik’s recent funding round of $350 million, reflecting a growing interest in advanced robotic solutions.
Demonstrations of Robotic Capabilities
In recent demonstration videos, Apptronik robots equipped with the new AI capabilities showcased their talents by plugging devices into power strips, packing lunchboxes, and manipulating plastic vegetables—all in response to spoken instructions. While these demonstrations highlight the potential of the technology, Google has not yet provided a timeline for commercial release.
Essential Qualities of Effective Robotic AI
For AI in robotics to be genuinely helpful and effective, Google identifies three key qualities that such models must possess:
- Generalization: The ability to adapt to various scenarios, allowing robots to be versatile in their functions.
- Interactivity: A quick and efficient response to user commands and changes in the environment.
- Dexterity: The ability to perform tasks typically handled by human hands, requiring a nuanced approach to object manipulation.
Gemini Robotics-ER for Developers
The Gemini Robotics-ER model serves as a foundational tool for roboticists looking to train customized models. Apptronik and a select group of "trusted testers," including well-known companies like Boston Dynamics, have access to this advanced platform, enabling them to push the boundaries of what robotics can achieve.
Competitive Landscape in AI Robotics
Google is not alone in the quest to integrate AI with robotics. In November, OpenAI made headlines by investing in a startup called Physical Intelligence, which is focused on implementing general-purpose AI into the physical world. This initiative aims to develop extensive AI models and algorithms for robots.
OpenAI has also made strategic hires, including bringing on the former leader of Meta’s augmented reality project to spearhead their efforts in robotics and consumer hardware. Additionally, Tesla has entered the rapidly evolving humanoid robotics industry with their Optimus robot, reflecting broad interest in this emerging field.
Google’s Vision for Robotics
Sundar Pichai, Google’s CEO, emphasized the importance of robotics as a testing ground for AI applications in the real world. According to Pichai, Google’s robotics will utilize multimodal AI models, making them capable of adapting to their surroundings and altering their behavior dynamically.
As robotics technology advances, companies are racing to harness AI’s capabilities to create machines that can assist in everyday tasks, from household chores to complex industrial functions. The future of this field looks promising, with major players like Google, OpenAI, and Tesla leading the way in innovation and development.
