DeepMind Announces Plans to Utilize AI Models for Physical Robots
Google DeepMind Introduces New AI Models for Robotics
Introduction to Gemini 2.0
On a recent Wednesday, Google DeepMind unveiled two innovative AI models designed specifically for robotics. These models operate on the Gemini 2.0 platform, which Google proudly claims is its "most capable" AI technology to date. By developing these models, Google aims to extend the capabilities of artificial intelligence from generating text and images to directing physical actions through robots.
Partnership with Apptronik
Google has teamed up with Apptronik, a robotics company based in Texas, to advance the construction of next-generation humanoid robots using Gemini 2.0 technology. Apptronik has previously worked with well-established entities like Nvidia and NASA, and has received significant financial backing, including a recent $350 million funding round that included Google as a participant.
Demonstrations of Robotic Abilities
In demonstration videos released by Google, robots equipped with Gemini’s new AI models performed various tasks. These included plugging devices into power outlets, packing lunchboxes, relocating plastic vegetables, and zipping up bags, all in response to vocal commands. Although the demonstration showcased impressive functionality, Google has not yet shared a timeline for when these robotic systems will be available in the market.
Key Qualities for Robotic AI
Google has identified three essential qualities that AI models must possess to be effective in the field of robotics:
- General Adaptability: The AI should easily adjust to various situations.
- Interactive Responsiveness: The system must understand spoken commands and react quickly to any changes in its environment.
- Dexterity: The robots should perform complex tasks similar to how humans use their hands and fingers to manipulate objects carefully.
These qualities will help ensure that the robots can be truly helpful and useful in everyday tasks.
Gemini Robotics-ER: An Advanced Tool for Researchers
In a more specialized offering, Google also announced Gemini Robotics-ER (Extended Reasoning), designed for roboticists who wish to train their own custom models. This tool will be accessible to Apptronik, alongside select "trusted testers" such as Agile Robots, Agility Robots, Boston Dynamics, and Enchanted Tools. This feature could pave the way for a new wave of robotic innovations tailored to various applications.
The Broader Landscape of AI in Robotics
Google is not the only tech giant pursuing advancements in robotics powered by AI. OpenAI, for example, has invested in a startup called Physical Intelligence, which aims to introduce general-purpose AI into the physical world through advanced algorithms and models.
Furthermore, OpenAI has made headlines by hiring a former head from Meta’s augmented reality division to lead their efforts in robotics and consumer hardware. Tesla, too, is venturing into humanoid robotics with its own project, the Optimus robot, showcasing how major tech companies are turning their attention to this fast-growing industry.
Google’s Vision for Robotics and AI
In a post shared on X, Google CEO Sundar Pichai expressed that the company views robotics as an effective platform for translating advancements in AI into the physical realm. The intention is to utilize Google’s multimodal AI models to allow these robots to adapt to their surroundings in real-time, enhancing their usefulness in everyday applications.
With the ongoing advancements in AI technology and robotics, the future promises exciting developments for how machines can interact with the world around them.
