An AI algorithm has taught itself how to solve Rubik’s Cube. Sort of. By building an algorithm that approximates how close the current cube configuration is to the solution, the AI was able to craft its own reward system and outperform most human methods of solving the cube. Why it may be important – similar approaches may be generalized from this technique for problems where it is hard to evaluate progress toward a goal, such as protein folding problems that are important in genetics and medical research.