The AI model Claude outperformed humans in controlling a robot dog: 20 times faster
A new phase of Anthropic's Project Fetch experiment demonstrates an impressive breakthrough in integrating artificial intelligence with physical systems. My analysis shows: the Claude Opus 4.7 model completed tasks for configuring and controlling a robotic dog 20 times faster than teams of human engineers. This is not just a number—it is a fundamental shift in understanding how AI can interact with the real world.
Autonomy at a New Level
While in August 2024, AI only assisted employees with no robotics experience, helping them find solutions faster, the situation has now changed dramatically. Claude Opus 4.7 operated almost entirely autonomously, under minimal researcher supervision. Without specialized algorithms for controlling hardware, the neural network independently:
- Connected to video sensors and LiDAR;
- Wrote a program for manual control;
- Created a robot path monitoring system;
- Configured an object recognition algorithm.
The key metric is code efficiency. The volume of code written by the neural network turned out to be 10 times smaller than that of human teams. This means not only speed but also quality: Claude writes cleaner and more concise solutions.
Comparative Performance
The numbers speak for themselves. Opus 4.7 was 18 times faster than a team using older AI versions and 37 times faster than people working without chatbot assistance. However, it is worth noting that this progress is a side effect of the general scaling of language models, not the result of targeted work on robotics.
Boundaries of Capabilities
Despite the impressive results, Claude still faces fundamental limitations. The model successfully guided the robot to the target but could not gently push a ball to the right spot. Tasks requiring complex real-time feedback still remain for humans. This reminds me that physical intelligence is an entirely different level of complexity compared to digital intelligence.
Anthropic rightly notes that the industry is entering the era of "physical AI agents." In the near future, neural networks will be able to work with standard tools and equipment as effectively as they currently operate with software code.
My expert opinion: This experiment is an important signal for the market. Investors and developers should reconsider their strategies: autonomous physical agents are becoming a reality faster than most predicted. However, the key challenge—precision of interaction with the physical world—remains unsolved. Companies that can bridge this gap between digital and physical intelligence will gain a tremendous competitive advantage.