The AI model Claude Opus 4.7 has surpassed humans in controlling a robot dog: speed 20 times higher
Anthropic has made a significant step forward in integrating artificial intelligence with physical devices. As part of the updated Project Fetch experiment, the Claude Opus 4.7 model demonstrated impressive results, completing tasks for setting up and controlling a robotic dog 20 times faster than teams of human engineers.
Autonomous operation without human intervention
Unlike previous testing phases, where AI acted only as an assistant, the new version of the model operated almost entirely autonomously. Under minimal researcher supervision, Claude Opus 4.7 independently completed a set of tasks: connecting to video sensors and LiDAR, writing a program for manual control, creating a robot path monitoring system, and configuring an object recognition algorithm.
Key performance indicators are impressive. The model proved to be 18 times faster than a team using previous AI versions and 37 times faster than humans working without chatbot assistance. Moreover, the code written by the neural network turned out to be 10 times more compact than that of human teams, indicating higher algorithm efficiency.
Limits of capabilities and prospects
However, despite clear progress, Claude still faces limitations in precise physical actions. The model successfully guided the robot to its target but could not handle the task of gently pushing a ball to the right spot. This highlights that complex real-time feedback remains an area where humans still hold an advantage.
Interestingly, progress in robotics has become a side effect of general scaling of language models, rather than the result of implementing specialized algorithms for controlling physical devices. This suggests that the potential of such systems may be significantly broader than previously assumed.
Anthropic believes the industry is entering an era of "physical AI agents," where neural networks will be able to use standard tools and equipment as effectively as they currently work with software code. This paves the way for creating truly universal assistants capable of solving both intellectual and physical tasks.
Expert opinion. The results of the Project Fetch experiment are not just a demonstration of technical progress but a signal that the boundary between the digital and physical worlds is blurring faster than many assume. In the coming years, we may see AI agents capable not only of analyzing data but also of controlling complex mechanisms in real time, which will fundamentally change approaches to automation in industry and logistics.