Here’s a robot solving a Rubik’s cube in a tad over one second. It’s not an official record for a machine yet (that’s 2.39 seconds), but they’ll get the Guinness record before too long.
Gizmag gives more information about how the robot works, and I’ll let the geekier readers read the details there. There’s an interesting discussion in the YouTube comments about whether there’s a trick here, but I don’t think so.
You’re probably asking, “What’s the record for a human?” Well, here it is: Lucas Etter solving the cube in 4.90 seconds.
I notice the human solver had a few seconds to study the cube before the timer was started. The robot was blind until the starting gun. All the same, I think the human performance is the more impressive.
Not just impressive, but to my limited abilities, quite impossible.
The “argument from personal incredulity”. That doesn’t fly anywhere, and pretty much everyone regular here has heard that coming out of the mouths (two – one per face) of a creationist in the recent past.
For what it’s worth, my personal best was 18 seconds when I wasn’t much older than the acne-magnet in the video. 20 seconds was a frequent time. These days, I struggle to get under a minute.
I have to say, the guys in these videos are using a to-my-eyes unfamiliar method for solving the cube, whereby they laboriously turn various parts of the cube until the colours match on each side…as opposed to the normal method of peeling off all the stickers and rearranging them.
This is a strange choice as it makes the cube quite difficult to solve.
nyuk nyuk nyuk
They’re using 3-d printed cubes. The colour goes all the way through the plastic ; there are no stickers.
(Actually, I don’t know if 3-d printers are accurate enough to make a “cube” – the dimensions are quite critical, otherwise things either jam, or explode.
Seems a bit laborious.
Still, think of all the babes they’ll get!
Once the acne clears.
They are very young aren’t they. It is interesting to think that at so young an age (I seem to remember) there are talents that emerge that are almost unbelievably clever. Most of these clever talents fade with time to be replaced with other talents…and then as age increases, other talents still, until…we become talented memories.
So are Rubik’s Cubes (the commercial product) still made and are they made this way or with the stickers? I sense a “this is slightly different so should we be doing comparisons” concern …
Oh, commercially they use stickers. But the challenge isn’t about the stickers, it’s about the symmetry operations in the group.
If you play Go, do you know the “throw the Go-ban at the opponent” tesuji? It has a millennium-long (at least) history, but remains frowned upon in polite society (I’m told).
Every cube I’ve handled has required some effort in order to spin the sides. That performance looked curiously effortless.
They loosen up over time. And I think that for competition the cubes can be altered a little to be extra loose for less resistance.
… which is a minefield. Make them too loose, and the cube can explode into components in your hand.
I’m not sure there’s much you can do to effectively modify a cube that would help. Lubrication would spread from the interior to the playing surfaces, reducing grip ; slackening the springs leads to imprecise positioning and cube disintegration.
Even if you started a tournament with a brand new cube, it’d get well loosened up by the end of the weekend.
Goodness. That sounds like Armageddon. Let us not suffuse wizardry too liberally. How about taking each cube as it comes from the factory, like a cellophane wrapped deck of cards?
(1) Expensive ;
(2) The factories assemble the cubes then stick the stickers on. So that’s not a lot of help.
TBH, half a dozen random twists does the job. Pass the cube through two or three successive people doing that, then chucking the scrambled cubes into a bucket (opaque) to be pulled for each new test would be perfectly adequate. But there would be the occasional randomly “easy” or “hard” one. Which is precisely why the actual results in competition discard the fastest and slowest times from a set of five, then compute the average (by which I take them to mean the “mean”).
I’m so glad they’ve worked it out scientifically. Otherwise where would we be as a nation, as a species? Randomize 5 times and drop the low and high values. You can’t beat that.
Some of the 2×2, 4×4 or non-square versions tended to be a bit stiff, and we’d sometimes use graphite shavings to get things moving. Not that I’ve touched one of those things for about 25 years…
Watching the first video, is it not bordering on cheating to dig grooves into the actual cube so that the robot can grip it better? They don’t do the same thing for robot jockeys after all.
It seems a bit dodgy to me(although I admit I’ve only had a cursory read through the Rubik’s-Cube-Robot-Solution-Time-Trial-World-Record-Bid rulebook.). How much alteration to the cube should be permissible?
The simple robot couldn’t handle the cube without the grooves. But there is an element of not handling friction like a human would, that makes it a bit on the odd side as far as comparisons go.
I think you may be forgetting the transcendental in all of this. Remember, these are rotations about an Platonic center of action.
Sub
How can solving attempts really be compared? Surely not all starting points are equally disordered.
(Don’t know if this is the trick mentioned in the YT comments, but one way people have faked solving the cube is to record themselves undoing a solved cube and then playing the footage backwards.)
“Surely not all starting points are equally disordered.”
My thoughts exactly.
They’re not. But they use computer-generated sets of scramble moves, which are of comparable difficulty. There are details at https://www.worldcubeassociation.org/regulations/#article-4-scrambling
I don’t know the maths exactly, but I’d suspect that deciding the optimal solution from a blind read of a cube would be a pretty hard problem. And deciding if a random sequence of moves to scramble a cube would actually be the optimum solution from that scrambled state isn’t going to be easy.
But still, you wouldn’t actually have to realize that you’ve got a less disordered starting point in order to reap the benefits.
I’ve never been able to “read” a cube to determine a solution from scratch. I only have the systematic solution methods, which require reading the cube as you go along through the solution.
There is a competition option where the competitor is given a cube to examine (a card or something is jammed into the cube to stop it rotating). They “read” it and plan their solution, then put it down on the (timing) mat, oriented however they want it. Then the are blindfolded. Then they’re timed on solving the cube.
Clearly, some people can plan the full solution from scratch. Not me, and I’ve never met such an animal.
I have to plan it as I go too. I can see it would be possible to plan it from scratch, but that’d take a huge amount of practice.
I still do it every year or so to make sure I still can, but it usually takes me at least five minutes these days, and I was never competition speed back in the day.
I didn’t get to try it for the first time until I was about thirteen, and the cube belonged to my younger brother. It wasn’t something girls were supposed to do.
As with all these things. (I’m just watching some piece of bullshit “hidden history” programme, with the presenter giving up on getting through a cave entrance that my grandfather could have got through, carrying his walker-frame in one hand and a camera in the other. Practice.
Related… the minimum number of steps needed to complete any cube, however disordered, was calculated a while back I think – was it twelve? Something like that. Maybe it was twenty? It was either twelve or twenty(or it might have been fourteen). Anyway, they calculated it and called it God’s Number(which seems a bit portentous considering it’s concerned with a plastic toy from the 80s).
The optimal solution declined through the years from “mid-20s” to 23, to 22, then 20 in 2010. There are several positions known to take 20 moves to solve, so that’s as good as it’s going to get. There may be many (potentially an infinite number) algorithms which result in solution at 20 moves.
I still don’t understand. I can imagine a one move solution if the cube is scrambled by only rotating one side 90 degrees. How can you say a cube requires at least 20 moves to be solved irrespective of its starting state?
It’s a worst case scenario. A particular algorithm will take AT MOST 20 moves to solve.
Ah. I was looking at Saul’s term “minimum”, which Aidan didn’t correct.
And I suppose if they use cubes that aren’t randomly scrambled, but scrambled to computer-generated specifications, then comparison makes more sense.
Oops – meant maximum.
They obviously had some minimal standard of mixing. I forget the details. The papers are available.
IIRC, at first glance there are 8! arrangements of corner pieces, 12! arrangements of edges, and 6! arrangements of the centre faces. However, there are conservation/ symmetry laws that knock those numbers down somewhat. I’d guess they pick a random example from that fairly extensive data set, and count from there. You’d probably need to require a certain low level of overall symmetry too. Your example of a single 90deg turn on one face would still have a 4-fold axis of symmetry, for example.
The breakthrough “Cambridge Solution” (1981) was down to 45 moves, when previous conventional methods were in the low 3 digits. It used a completely different methodology (none of this one layer at a time bizness). so that to the superficial observer the cube looks quite scrambled even 6 or 7 moves before the end. It’s been improved over time and been gotten down to 20.
The ideal algorithm has been referred to as “God’s algorithm”.
One more little step toward human obsolescence.
Does make me wonder. Of course, natures been beating us at running (short distances) and swimming (almost any distance) for a long time. Robots would not seem like that much of a threat.
Cars and boats can out perform us and we seem to have thought that does not really matter.
But outthinking is something totally different from outperforming physically.
First, I will say he was fast.
But it is not the same thing. The human was able to study the cube before time was started and once he had a solution – then he was timed. I still think these contests should start the time when a scrambled cube is revealed until solved. That is a more interesting task.
The computer had to solve the cube and make the moves in the time.
The time to “solve” the cube is negligible. Typical solution paths involve a hundred or so algorithms and 50 to 60 moves. That’s negligible for computers, but as you can see, “reading time” is a significant time for humans (due to only having one optical system).
Genau! thus an interesting competition for humans is to solve from first glance.
Of course, the computer is fast so what. It’s the human times that are interesting.
And once upon a time, the argument used to be ‘computers can’t play chess’.
Now it’s ‘the computer is fast but so what’
The machines have won.
cr
What would intrigue me would be Braille Rubiking.
Bingo!
I could manage that, as long as we ignore these silly dots.
Did that back in the day, gluing on bits of sandpaper, corduroy etc. to get distinctive feels, and it worked OK. There’s a list of (other) examples here.
Reblogged this on My Selfish Gene and commented:
And you thought you were fast? Robot solves Rubik’s cube in a second +. From Why Evolution is True.
Puzzles, like this dastardly cube, vex me. I’m glad we’ve outsourced the unpleasant work to robots. 😉
That leaves little to us humans but romance.
Then possibly raising the outcome of romance.
Well, unless you consider offspring part of the romance?
I don’t know … there are already robots for certain aspects of that activity … (In fact, in Korea it supposedly has already been ruled that a sex bot is not a prostitute.)
That would depend on the degree of professionalism. I hope at least she’d be a hooker with a heart of gold connector pins?
Actually, I’ve been told it was to determine if the activity is legal – prostitution is illegal, apparently, there, but sex toys are legal, so …
That’s *physically* solving a cube! I’d imagined it would be a virtual cube. That must be a very well-made cube not to fly apart from the accelerations.
On a side note, what I’ve always found even more puzzling than the solving (which is really just a matter of combining a series of moves, though I can’t do it) – is how the cube is physically made. How do you hold the segments so they can slide past each other one way then another without falling off? That’s magic! (I know a spherical core is involved but I still can’t get my head around it).
Oh, and the computer runs Linux. Yay! (Like me and all* the world’s supercomputers).
(*To a first approximation)
cr
“what I’ve always found even more puzzling than the solving (which is really just a matter of combining a series of moves, though I can’t do it) – is how the cube is physically made.”
This!!
I don’t know which is more impressive, the robot or the guys who built it!
First it was “Deep Blue” and “Watson,” in chess and Jeopardy, now this — portents all that the Kurzweilian Singularity is nigh!
Soon there will be nothing a grouchy, old white guy can do better than a computer, and we will all know the resentment that drives Trump fans.
And now Deep Mind plays world-class go.
I love this definition of a game (which some thought to be undefinable) by Bernard Suits: the voluntary attempt to overcome unnecessary obstacles.
That’s rather good. I’d never heard it before – it’s instantly set me thinking of possible counterexamples.
Looking at the video –
Where it differs from a human solving the cube is that the six stepper motors never let go of the centre squares on each side – they just rotate them as required.
So each centre square never moves from its initial position. This is a little counter-intuitive for anyone who’s fiddled with a cube by hand.
They say the Arduino carefully controls acceleration – obviously necessary to not break the cube.
(Trying to build a robot to grip the cube like a human would be enormously more complex and – manifestly – topologically unnecessary).
cr
Not be long now before the Machines take over.lol
Interesting that there was not a girl in the bunch and no one over 17 in the Lucas Etter record setting video.
Also impressive are the people who examine a cube and then solve it blindfolded. For those records the examination time and solving time are lumped together. One guy was able to solve a very large 40?) cubes at the same for blindfolded. That to me is amazing.
I suppose these folks get similar special (spatial?) memory to chess players – blindfold chess is well known, after all.