The robotic hand from OpenAI has learned how to solve a Rubik's cube


Another interesting organization project.

We've seen so many weird iterations of Rubik's cubes so far. One was solved by various robots, other they dissolved themselves levitating in the airand yet others allow you to monitor your health. However, we have never encountered a cube before which a robot equipped with a robotic hand modeled on the hand of a human would try to solve. Something like this has been shown to the world by the OpenAI organization.

OpenAI is a non-profit organization founded, among others, by Elon Musk. It is devoted to research on artificial intelligence and its development, and the last robot is a manifestation of this sacrifice. It is a robot that had to figure out how to solve the Rubik's cube.

The robot's software is a pair of neural networks that work together to observe the ankle and control hand movements. They were given the task of making each wall of the cube contain a single color. The networks had to find the way to achieve this goal by training their skills in computer simulation. Thanks to simulation, artificial intelligence has learned how hand movements and gravity affect Rubik's cube.

In fact, neural networks performed many computer simulations simultaneously. This training method allowed them to gain years of experience over a much shorter period. Here's how its effects look in practice:

As you can see, the robot is quite good at solving Rubik's cube, although it is clear that so far his dexterity does not match the masters in solving the ankle with one hand. That said, the machine probably would not accomplish its task so well if its creators did not interfere in the training process, introducing some difficulties, for example in the form of a rubber glove hindering making movements.

According to OpenAI, work on this type of robot is of great importance because it can lead to the creation of multifunctional robots that would match the dexterity of people. Well, we hope that these machines will not revolt against us in the future. We know well how this scenario can end.

Source: OpenAI