© ROOT-NATION.com - Use of content is permitted with a backlink.
Boston Dynamics has integrated Google DeepMind’s Gemini Robotics ER 1.6 system into its Spot quadruped robot to move it from following pre-scripted scenarios to tasks based on logical reasoning. The company demonstrated the upgrade in a video where the robot reads a handwritten to-do list and performs tasks such as organizing shoes, picking up cans, and putting clothes in the laundry basket.
In one of the episodes, Spot takes a leash and walks a dog, which emphasizes its ability to translate natural language instructions into physical actions. The system combines computer vision, speech understanding, and action planning, allowing the robot to analyze the environment and act with minimal human intervention. This means a departure from the classical approach in robotics, where precise programming is required for each task. At the same time, the demonstration also shows the gap between AI thinking and real-world execution. In one example, the robot grabs a jar from the side, which is a simple mistake that can lead to spillage, and it demonstrates the limitations of human-like understanding.
Despite the domestic focus of the demonstration, the key emphasis is on industrial inspections, where Spot is already in active use. The robot is designed to move around facilities, detect hazards, and monitor environments that are difficult or risky for humans to access. Using Gemini Robotics ER 1.6, Spot is able to autonomously detect problems, such as water accumulation, read instrumentation, and assess conditions at the facility. It can also use models such as vision language action to gain a deeper understanding of complex environments.
According to Marco da Silva, vice president and general manager of Spot at Boston Dynamics, such technological steps are an important step towards creating robots that better understand the physical world and are able to act in it. He noted that capabilities such as tool sensing and more robust logic processing will allow Spot to see, analyze, and respond to real-world challenges fully autonomously. This update is a continuation of the partnership between Boston Dynamics and Google DeepMind announced earlier this year, which aims to integrate advanced AI models into robotic systems.
Despite the improvements, the system has limitations. Currently, the model is heavily dependent on visual data and lacks the deep physical understanding that humans develop through touch and experience. Karolina Parada, head of robotics at Google DeepMind, emphasized that for robots to perform tasks reliably and safely, it is critical to align the way they perceive the world with the human. The company is working to improve the safety and quality of reasoning through a system of tests that determine the behavior of robots in real conditions. However, full autonomy still depends on achieving consistent reliability in different environments.
Boston Dynamics reports that the implementation in real-world conditions will be gradual and controlled, and customer feedback will play a key role in improving the system. The company emphasizes that the robots must achieve a level of reliability that will avoid false positives and maintain user confidence.
Read also: