Google DeepMind Unveils Gemini Robotics 2 for Humanoid Control

Google DeepMind has announced the launch of Gemini Robotics 2, a sophisticated AI capable of full-body control of humanoid robots. This development marks a significant leap from its predecessor, which focused primarily on upper-body motions, with the new model controlling movements from feet to fingertips, according to The Verge.

The AI model empowers robots to perform complex tasks like walking, crouching, and using hands for manipulative tasks such as screwing in lightbulbs or tying bags. Wired reports that these capabilities result from integrating vision-language models, enabling robots to interpret and act on visual data.

The upgrade aims to advance the potential of physical AGI, where robots could carry out a wide variety of tasks akin to humans. Despite these advancements, there remain hurdles, particularly in terms of the speed and complexity of movements, which Google DeepMind acknowledges as areas needing improvement.

Gemini Robotics 2 also supports collaboration among multiple robot types, facilitating cooperative task execution. This functionality is demonstrated by a scenario where Apollo 2 and other robots work together to organize tools in a garage, as noted by The Verge.

However, with these capabilities come significant safety concerns. The deployment of AI in physical spaces could lead to unpredictable and potentially hazardous outcomes, highlighting the need for robust safety measures, reports Wired.

To mitigate risks, Google employs a multi-layered safety strategy, which includes the introduction of the ASIMOV-Agentic benchmark to evaluate the safety of commands executed by robots. This initiative aims to ensure the safe interaction of AI within varied environments.

The broader goal, as expressed by Google DeepMind's leadership, is to develop a universal operating system for robots, similar to Android for mobile devices. This would facilitate widespread adoption and versatility across different robotic applications.

Google's commitment to advancing robotics through AI reflects its belief in AI's potential beyond the digital world into practical, everyday applications. The successful implementation of Gemini Robotics 2 could significantly influence how robots are integrated into societal frameworks.