Page 1 of 3
Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 2:35 pm
by Michel
https://www.nature.com/articles/s42256-019-0070-z
I have not quite understood yet what they do exactly.
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 2:43 pm
by Michel
Ok the use a NN to provide an estimated solution time. The NN is trained to match the estimated solution time to the one estimated from a 1-ply search. In chess this an old idea (match the heuristic evaluation function to the result of a 1-ply search).
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 3:53 pm
by Michel
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 5:24 pm
by Daniel Shawul
Can it beat this guy though who solves 17x17x17 cube in about 2 hours ?
https://www.youtube.com/watch?v=7ChuKKL2PpU
I would like to understand what makes it a unique challenge. From what I understood from the abstract
- single goal state
- solved in reverse with root node being that single goal state (i.e. solved state)
- finds shortest path 60% of the time
Please add more if you read the full paper.
Daniel
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 5:32 pm
by Daniel Shawul
Dumb reporter from
jakarta post
In a world first, researchers at the University of California have developed a computer algorithm that can solve a Rubik's Cube without a neural network, machine learning techniques, "specific domain knowledge," or human assistance.
And this
By successfully being able to solve a Rubik's Cube without initially being trained on previous information, the DeepCubeA algorithm represents the gradual shift in machines from making carefully-directed computations to making those which appear to resemble human-like reasoning and decision-making.
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 5:32 pm
by Daniel Anulliero
Daniel Shawul wrote: ↑Thu Jul 18, 2019 5:24 pm
Can it beat this guy though who solves 17x17x17 cube in about 2 hours ?
https://www.youtube.com/watch?v=7ChuKKL2PpU
I would like to understand what makes it a unique challenge. From what I understood from the abstract
- single goal state
- solved in reverse with root node being that single goal state (i.e. solved state)
- finds shortest path 60% of the time
Please add more if you read the full paper.
Daniel
Fake , reversed video
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 5:37 pm
by Daniel Shawul
Failed for it
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 5:37 pm
by Michel
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 6:39 pm
by mar
I've seen this video, probably some ad hoc solution, but funny:
https://www.youtube.com/watch?v=f9smvQ5fc7Q
The basic idea as I understand it is to solve inner faces first, then edges and the rest is solved just like a classic 3x3 cube.
(I admit that I never managed to solve a rubik's cube myself
Re: Self-taught AI solves Rubik's cube
Posted: Thu Jul 18, 2019 9:06 pm
by Michel
mar wrote: ↑Thu Jul 18, 2019 6:39 pm
I've seen this video, probably some ad hoc solution, but funny:
https://www.youtube.com/watch?v=f9smvQ5fc7Q
The basic idea as I understand it is to solve inner faces first, then edges and the rest is solved just like a classic 3x3 cube.
(I admit that I never managed to solve a rubik's cube myself
Long ago when the cube first came out I bought one and managed to find an algorithm to solve it. Unfortunately I lost a girl friend in the process...
