DeepMind Announces Plans to Utilize AI Models for Physical Robots
Google DeepMind Launches New AI Models for Robotics
Introduction to Google’s New AI Models
On a recent Wednesday, Google DeepMind introduced two innovative AI models specifically designed for robotics, marking a significant advancement in its Gemini 2.0 framework. The company describes these models as the most advanced AI it has created to date. These models signify a shift from purely virtual AI applications to physical actions that can control robotic systems.
Collaboration with Apptronik
Google has announced a partnership with Apptronik, a robotics firm based in Texas, to create the next generation of humanoid robots utilizing Gemini 2.0 technology. Apptronik has a rich history of collaboration with major technological companies, including Nvidia and NASA. This partnership, which recently culminated in Apptronik’s funding round of $350 million—partially supported by Google—aims to enhance the capabilities of humanoid robots.
Demonstrations of Robotic Capabilities
In a series of demonstration videos, Google showcased Apptronik’s robots equipped with its new AI models. These robots were seen performing everyday tasks like plugging devices into a power strip, organizing a lunchbox, manipulating plastic vegetables, and zipping up a bag—all in response to spoken commands. However, Google has yet to disclose when these advanced robotic technologies will be broadly available to consumers.
Essential Qualities of AI Robots
According to Google, for robots powered by AI to be truly helpful, they must possess three key qualities:
- General Adaptability: The ability to adapt to various situations and environments.
- Interactivity: Quickly understanding and responding to instructions or changes around them.
- Dexterity: The capability to manipulate objects with precision akin to human hands.
These attributes are critical for enabling robots to perform tasks that are essential in everyday life.
Details on Gemini Robotics-ER
The Gemini Robotics-ER model is specially designed for robotic developers. It serves as a foundational tool for creating custom AI models. This model will not only be available to Apptronik but also to a select group of trusted testers, including renowned firms like Agile Robots, Boston Dynamics, and Agility Robotics. This collaborative endeavor is expected to pave the way for advanced robotic solutions in various sectors.
Industry Context and Competitors
Google is not the only company venturing into AI-driven robotics. In November, OpenAI invested in a startup called Physical Intelligence, which is focused on integrating general-purpose AI into robotics systems. Furthermore, OpenAI has also started to build its own robotics division, having recently recruited the former leader of Meta’s augmented reality initiatives to drive its efforts.
Tesla has also entered the humanoid robotics market with its own Optimus robot, highlighting the competitive landscape in this rapidly evolving field.
Insights from Google’s Leadership
Sundar Pichai, the CEO of Google, expressed on his social media platform that the company views robotics as a promising avenue for testing and implementing advancements in AI within real-world scenarios. He emphasized that these robots would employ multimodal AI models, allowing them to adapt and make real-time adjustments based on their surroundings.
As these technologies develop, the potential for AI-enhanced robotics could significantly impact various industries, from manufacturing to home assistance, shaping the future of how we interact with machines. In this evolving narrative, it’s clear that Google and its partners are at the forefront, striving to create intelligent, agile, and interactive robots that can seamlessly integrate into our daily lives.
