The OpenAI robot learns dexterity through virtual training

The advantage of humans over machines is their manual dexterity. This state of affairs may not last very long, and all because of the software that virtually increases the coordination of the robotic arm.

The algorithm is created using the Dactyl system. The program documents how the robot manipulates, for example, a cube and then remembers the position of its fingers each time the object is moved. Each time the data is updated, the machine gains more and more dexterity and turns the cube faster and in a more “human” way.

We are dealing with a revolutionary invention in several respects. Dactyl can operate in an environment in which it is influenced by many stimuli, both visual and auditory. In addition, it has a much higher so-called the number of degrees of freedom compared to typical robotic arms (ratio 24: 7 in favor of Dactyl). It is also worth mentioning that the software works not only virtually, but also in the real world. Moreover, the machine is able to operate on more than one object at the same time, it is capable of multitasking.

The OpenAI company responsible for the development of the project does not intend to rest on its laurels. The next stage of their work will be to create a robot that develops in a similar way as in the case of children. The machine will explore the world by itself by interacting with the things around it. Such a futuristic vision seems tempting, but the control over the machines will remain a problem. Do we have to remind about the Skynet system once again?

[Źródło: blog.openai.com; grafika: OpenAI]