OpenAI has opened a closed research lab in San Francisco specializing in robotics. This center is continuously testing robotic arms, training them to perform everyday tasks—from toasting bread to carefully sorting and folding clothes.
The training process uses a special controller that precisely mimics the movements of human hands. This creates a vast database of real-world actions, which is then used to train robot control algorithms. The project is not focused on creating humanoid machines, but on developing functional and affordable robots capable of operating effectively in everyday environments.
The initiative involves over a hundred operators and at least twelve engineers. The robots are controlled remotely, and all actions are recorded and stored for subsequent analysis and training of artificial intelligence models.
Furthermore, the company is already preparing to launch a second laboratory in California. The robot training approach is based on the same principle as the development of language models: systems are trained on a large array of real-world examples of human behavior, which allows for the gradual development of universal skills and increased autonomy.
As a reminder, Chinese engineers have created the smallest robot capable of learning.
To be continued…
Only registered users can leave comments