AI-Powered Robot Hand Solves Rubik's Cube In Minutes
Infineon developed a robot in 2016 aimed at solving the Rubik's cube - which it did, in under a second. And later, in March 2018, MIT engineers made another system capable of completing the same task in a minuscule 0.38 seconds. Impressively, both robots could outcompete the world record set by a human to an unbelievable extent.
Although these superhuman speeds make Dactyl - the humanoid robotic hand developed by OpenAI last year - and its four-minute average time less grand at face value, it provides an insight into the degree to which artificial intelligence has come along.
By using computer simulations that would take a human over 10,000 years to do, the robot was able to consistently solve cubes single-handedly, and, unlike the robot created by Infineon, it is versatile and not confined to a single, narrow task. In fact, to achieve such spectacular sub-second results, artificial intelligence doesn't even need to be used as it only entails making a more precise, and quicker robot.
What's more, in a video demonstration of Dactyl, the company highlighted how the robot can deal with other unexpected disruptions to the process, such as poking the cube with a stuffed giraffe toy or covering the cube with a black sheet. The success rate was affected, however, but researchers say that the robot was still able to persist through these disturbances and manage to complete the puzzle.
Furthermore, it would be impossible to use other robots to approach tasks outside of what they were designed for, which is what OpenAI's goal is for Dactyl - to create a general-purpose robot; as close as possible to a human hand which performs so many different tasks. To accomplish this, it is trained using software which tries to imitate the brain learning to use our hands for the first time as children.
The possibility of this technology being pursued to the kinds of robots you only expect to see in science fiction movies is thrilling. Someday robots may be a part of our daily lives, such as in our homes or other chaotic environments, without putting us at risk but rather to benefit us.
Nonetheless, Prof Ken Goldberg from UC Berkeley mentions that even though the demonstration of Dactyl illuminates the real progress which OpenAI has made, its a little ambiguous.
So, yes, there is still a lot of progress to be made before we start seeing AI-powered robots in the real world, but the 24-join robotic hand that is Dactyl is part of OpenAI's vision for the future in which robots can integrate into society by learning new tasks and adapting to changing environments.