DeepMind to Utilize AI Models for Enhancing Physical Robots
Google DeepMind Launches New AI Models for Robotics
On Wednesday, Google DeepMind introduced two cutting-edge AI models designed specifically for robotics, powered by their latest technology, Gemini 2.0. According to Google, this version is their "most capable" AI yet, marking a significant step forward in how AI can interact with the physical world.
Introduction of Gemini AI Models
The two new AI models are named Gemini Robotics and Gemini Robotics-ER (Extended Reasoning). While earlier AI technologies focused on generating text and images, these models aim to control robots with physical actions. This new approach represents a move from passive outputs to active engagement in real-world tasks.
Partnership with Apptronik
Google has formed a partnership with Apptronik, a robotics developer based in Texas. This collaboration aims to create the next generation of humanoid robots utilizing Gemini 2.0 technology. Apptronik has previously collaborated with prominent organizations like Nvidia and NASA, further solidifying its reputation in the robotics field. Recently, Google participated in a $350 million funding round for Apptronik, signaling a strong commitment to advancing humanoid robotics.
Demonstration of Capabilities
In a series of demonstration videos, Google showcased the capabilities of Apptronik robots equipped with the new AI models. These robots were observed performing various tasks such as plugging items into power strips, filling lunchboxes, manipulating plastic vegetables, and zipping up bags, all while responding to verbal commands. However, Google has not yet announced a timeline for when these robotic technologies will be available on the market.
Essential Qualities of AI in Robotics
For AI models to be effective in the realm of robotics, Google highlighted three crucial qualities:
- Generality: The AI models must be adaptable to diverse situations and tasks.
- Interactivity: They should quickly understand and respond to instructions and changes in their environment.
- Dexterity: The robots need to replicate the intricate movements humans can perform with their hands and fingers, allowing them to manipulate objects skillfully.
Gemini Robotics-ER serves as a foundation for roboticists to train their models. It is accessible not only to Apptronik but also to trusted testers which include several notable companies like Agile Robots, Boston Dynamics, and Enchanted Tools.
Growing Interest in AI for Robotics
Google is not the only company venturing into AI for robotics. In November, OpenAI invested in a startup named Physical Intelligence, which aims to integrate general-purpose AI into physical tasks through advanced models and algorithms. Additionally, OpenAI recently appointed the former head of Meta’s augmented reality initiative to lead their robotics division, indicating a significant investment in this area.
Tesla is also making strides in humanoid robotics with the introduction of their Optimus robot, which further emphasizes the rapid evolution of this industry.
Vision for the Future
Sundar Pichai, CEO of Google, mentioned that the company views robotics as an essential testing platform for implementing AI advancements in the physical realm. The robots will leverage Google’s multi-modal AI capabilities to make adjustments in real-time and adapt to their surroundings, enhancing their functionality.
The developments showcased by Google DeepMind and their collaborative efforts with Apptronik signify an exciting new chapter in the field of robotics, where AI can play a pivotal role in enhancing the capabilities of humanoid robots. As research and innovation continue to advance, the potential applications for this technology are vast and promising.
