DeepMind Announces Plans to Utilize AI Models for Physical Robots
Google Unveils New AI Models for Robotics
On Wednesday, Google DeepMind introduced two innovative AI models designed for robotics, both operating on the highly-advanced Gemini 2.0 framework. This platform is touted by Google as their "most capable" AI system to date. The new models are called Gemini Robotics and Gemini Robotics-ER (Extended Reasoning).
AI Models Overview
Gemini Robotics marks a significant shift as it moves beyond traditional outputs like text and images, typical of many generative AI systems. Instead, it focuses on executing physical actions, which allow robots to perform various tasks in real-world environments. This capability could revolutionize how AI interacts with the physical world.
Partnership with Apptronik
To further the development of humanoid robots, Google announced a partnership with Apptronik, a robotics company based in Texas. Apptronik has previous experience collaborating with companies like Nvidia and agencies such as NASA. In a recent funding round, it was reported that Google took part in Apptronik’s $350 million investment initiative, highlighting its commitment to advancing humanoid robotics.
Demonstrations of Capabilities
In a series of demonstration videos, Google showcased Apptronik robots equipped with these cutting-edge AI models. The robots were seen performing various tasks, including:
- Plugging devices into power strips
- Filling lunchboxes
- Moving plastic vegetables
- Zipping up bags
These tasks were executed in response to spoken commands, demonstrating the robots’ ability to understand and act based on instructions.
Key Features of Gemini Robotics
According to Google, effective AI models for robotics must display three essential qualities:
- Generality: They should be versatile enough to adapt to different scenarios.
- Interactivity: They must quickly understand and respond to instructions or changes in their environment.
- Dexterity: They should be capable of performing tasks that humans usually can do with their hands, such as manipulating objects carefully.
Gemini Robotics-ER: A Resource for Developers
The Gemini Robotics-ER model is specifically designed as a foundational tool for roboticists, allowing them to train their own models. This version of the AI technology will be available not only to Apptronik but also to a select group of trusted testers, which includes other prominent robotics firms, such as Agile Robots, Agility Robots, Boston Dynamics, and Enchanted Tools.
The Robotics Landscape
Google isn’t the only tech giant focused on AI for robotics. Recently, OpenAI announced an investment in Physical Intelligence, a startup dedicated to integrating general-purpose AI into robots. The company aims to develop large-scale AI models and algorithms to facilitate this goal. Moreover, OpenAI also hired a former executive from Meta to oversee its robotics and consumer hardware initiatives.
Tesla is also actively participating in the humanoid robotics sector, working on its Optimus robot, which reflects an increasing trend among tech companies to explore advancements in robotics paired with artificial intelligence.
Comments from Google Leadership
Sundar Pichai, the CEO of Google, expressed his enthusiasm about the company’s role in this field on social media. He emphasized that robotics serves as a valuable testing ground for applying AI advancements in tangible environments. According to Pichai, the robots will leverage Google’s multimodal AI models, enabling them to adapt and make real-time adjustments to their surroundings.
As we move forward, the intersection of AI and robotics appears poised for rapid growth and development, promising exciting possibilities in how we interact with technology in daily life.
