Self-solving Rubik’s Cube and General Purpose AI

I like mathematics, puzzles, problem solving, brain teasers etc. I knew what Rubik’s Cube was. But somehow I never saw one till I was 35 years old! Strange, but true. To be precise, I should say, I discovered Rubik’s Cube at that time. Got fascinated.

After many months of tinkering (and also with help from some cheat sheets) I finally managed to solve it and I was super delighted! However, I am still fascinated by the Cube. I know how to solve it, but don’t know how the solution works. i.e. I know the steps and the process, but I don’t know WHY the steps work. I am still searching for the answer.

Isn’t it similar to many things in life (or the life itself!)? We know how most things function, but not WHY. And it’s a beautiful mystery we try to solve, or we should try to solve.

Anyways, that was a slight deviation.

Another theme I am fascinated by, in last few years, is automation/artificial intelligence/ man vs machine. I have written few blog posts on the topic. The one was about Paul – The Robot. The second was about Garry Kasparov’s book “Deep Thinking” on Artificial Intelligence and the major breakthrough event in Chess when Google’s general purpose AI learnt rules of chess, practiced for few hours and defeated StockFish, one of the best Chess programs (works on brute force and database, not based on AI).

The purpose of writing this blog is because I encountered something interesting – combination of Rubik’s Cube and Artificial Intelligence!

I recently came across this amazing video of self-solving Rubik’s Cube machine!

 

I wondered how it works? I wanted to know if the bot really “solves” the cube or only performs a set process/runs a defined algorithm.

So I searched for a while and found this video which explains how the self-solving bot really works.

So basically the cube remembers the steps performed when it is being scrambled and then unscrambles itself using the reverse steps. It doesn’t really “solve” the cube.

In that sense the self-solving Cube only does what I or many other people do manually. I am not undermining this. This is a HUGE achievement! I am only comparing it with Chess. For many decades we had Chess playing software (which uses brute force analysis and a huge repository of games) and they kept on getting better and better. However they were still “programmed”. Then we had general purpose AI, which learnt the rules of the game and practiced and defeated the best software! So it learnt the game on its own! (“Solved” is a wrong word for Chess, because Chess cannot be “solved”, it doesn’t have a unique or finite solution).

In that sense, machines are yet to “solve” Rubik’s Cube. But I guess that would eventually happen. The general purpose AI machines would become stronger and stronger and one day, not in distant future, they would “solve” Rubik’s Cube (without knowing the algorithm or the steps).