The AI agent Claude Opus 4.7 has outperformed humans in controlling a robot dog by a head — with speeds 20 times faster.
The world of crypto technology and artificial intelligence continues to amaze. This time, the Anthropic team presented updated results from the Project Fetch experiment. Its essence is to test whether a neural network can effectively program and control a four-legged robot. The results are impressive: the Claude Opus 4.7 model completed the entire cycle of tasks 20 times faster than the best teams of human engineers working with previous versions of AI.
Full Autonomy: From Sensors to Recognition
In August 2024, employees with no experience in robotics participated in the experiment. At that time, AI only helped them find solutions faster. Now, Claude Opus 4.7 acted almost autonomously, under minimal researcher supervision. The neural network independently:
- connected to video sensors and LiDAR;
- wrote a program for manual robot control;
- created a trajectory monitoring system;
- configured an object recognition algorithm.
Numbers That Speak for Themselves
The comparison with humans looks devastating. The Opus 4.7 model proved to be 18 times faster than a team using older versions of AI, and 37 times faster than engineers working without chatbot assistance. Moreover, the code written by the neural network turned out to be 10 times more compact than human code. This is not just speed — it is a qualitatively different level of efficiency.
The authors emphasize that progress in robotics has become a side effect of the general scaling of language models. Anthropic did not introduce specialized algorithms for controlling hardware. This suggests that fundamental LLM models are already capable of solving complex physical tasks without additional tuning.
Limitations: Physics Still Poses a Challenge
Despite the success, Claude still struggles with precise physical manipulations. The model managed to guide the robot to the target but failed to gently push a ball to the exact required point. This requires complex real-time feedback — and here, humans still outperform AI.
Anthropic is confident that the industry is entering an era of "physical AI agents." Soon, neural networks will be able to use standard tools and equipment as effectively as they currently use software code.
Expert Opinion: This experiment is a powerful signal for the market. If AI can replace an entire team of engineers in robotics, then the crypto industry, where automation and speed are critical, could gain a new tool for developing decentralized systems and hardware solutions. But for now, the physical world remains the last bastion of human superiority.