Leading Supplier of Linear Modules,Electric Grippers,Motion Control Systems
Intelligent Robot Hands Can Learn to Solve Rubik’s Cube with One Hand? and can Completly Teach itself Without Programming?
You might say that it is not strange for robots to solve Rubik’s cube. As early as 2016, a robot made by Infineon can recover a Rubik’s cube in less than one second, far exceeding the fastest record of mankind. A robot developed by MIT in 2018 reduced this time to 0.4 seconds.
Open AI is an artificial intelligence non-profit organization jointly established by many Silicon Valley tycoons. It is the home of manipulator dactyl.
The terrible thing about the manipulator dactyl is that, unlike those robots in front, they are made specifically for solving the magic cube, and the completion of the action is the result of manual programming intervention. The manipulator dactyl completes self-learning in the virtual environment and adds a large number of random events in the training process, so that dactyl can understand the secret of completing the task by himself in this process. Human beings are not specially programmed for the operation of manipulator. Dactyl understands everything by himself. Dactyl’s success in solving the Rubik’s cube with one hand means that the robot has the consciousness and ability of self-learning.
Before that, researchers had to randomize the parameters in the environment by manually selecting the arrangement they thought would produce a better algorithm. The current training system can do this by itself. Whenever the robot reaches a certain level of proficiency in the existing environment, the simulator will adjust its own parameters to make the training conditions more difficult.
The result is a more stable algorithm that can move with the accuracy required to rotate the cube in real life. Through the test, the researchers found that dactyl can successfully solve the magic cube without training. For example, it is wearing rubber gloves, several fingers are tied together, and a stuffed toy giraffe is poking it.
OpenAI does not explicitly program dactyl to help it crack the cube. In order to train dactyl to recognize these actions, it only sets the ultimate goal of cracking the cube for the underlying software of the robot hand, and uses modern AI deep learning technology to help it teach itself how to solve problems.
When given the real magic cube, dactyl took advantage of his training and solved the problem by himself. Even when the interference scene was temporarily added, dactyl remained unmoved and successfully completed the task.
We can see that the Intelligent Robot Hands gradually cracked the Rubik’s cube in a groping way. Although its action seems a little clumsy, it is very accurate. The action of the Intelligent Robot Han is obviously not as flexible as that of a real human hand, nor can it be compared with the amazing speed of those who can crack the magic cube in just a few seconds. But for openai, dactyl’s achievements have taken it another step towards the desirable goal of the wider AI and robotics industry.
OpenAI said it was trying to build a universal robot. Similar to the dexterity of humans and human hands, it can complete a variety of different tasks.
Dactyl is a robot hand that starts from scratch and gradually grasps the ability of self-study. Intelligent Electric Grippers can handle new tasks like human beings. Dactyl also needs software training and is currently trying to replicate millions of years of evolutionary experience in an elementary way, which has also helped us learn how to instinctively use our hands as children. Openai hopes to help humans develop humanoid robots that we can only see from science fiction in the future. Intelligent Grippers can integrate into society without endangering our safety and perform various tasks in chaotic environments such as urban streets and factory workshops.
At present, dactyl has accumulated about 100 years of training experience, while in reality, this process is only 50 hours. You can imagine dactyl learning a variety of human skills in the virtual world and accumulating tens of thousands of years of training experience, but in reality, we just spent a few months. It is not far away that intelligent machinery will replace more and more people.