I think the difficulty is measured and reassured by how many close, put not quite solutions there are. Also brute forcing these would be extremely difficult. The amount of combinations of locations for each piece is 16! = 2.1*10^13 and that is without factoring in the different rotations. There are definitely some packing algorithms that could solve it, but I think as a regular human, just going through all the potentially possible options until you find it is the most consistent way. Knowing the corner pieces, what sides can match with what over sides drastically cut down on the amount of combinations you have to weed through.
im not sure about these kinda puzzles, but to my knowledge the difficulty is defined by the layout that it is cut in, and the important part is to find the right order of the corners. in the second puzzle its harder, cause the cut layout uses straight lines mostly in the middle, instead of the notches, and nods.
@@timperman9883 That's is the right amount of locations, not including rotations. There are 8 different rotations. 16! * 8! means 843,606,888,284,160,000 (8.4*10^17) different possibilities.
@@rondenverliwanag6784 I think you're right, in the "hard" variant you cannot determine in advance which pieces will be edge pieces since every piece has a flat edge. And in the "medium" variant only 2 pieces have no flat edges. I would assume an "easy" variant has 4 pieces with no flat edges which would make them obvious center pieces like a normal jigsaw puzzle.
In the Medium puzzle there were pieces that clearly had to be arranged in the middle because they had no flat side. In the hard one every piece had at least one flat side, so you couldn't figure out where the pieces would go. I think this makes it harter because there are way more possibilities.
I believe you are absolutely right. I noticed the same thing and came to the comments section to speak about it. Also I would have rated the second puzzle a difficulty of 5/5.
But, you can start to think like this "So, if all pieces have one flat side, when i start to solve, every time i want to make sure that i made moves that allows the remainings spaces to be 3 connectors and 1 flat side". So, when you are putting the puzzle together and see that one part it's not possible to put 3 connects and 1 flat, you change the layout. I think this make a little bit easier to solve this kind of puzzle
The number of pieces with flat sides doesn't have to be the only difference between the different difficulty approaches. Apart from possibly more non-unique solutions in the easier difficulties, they could be designed in a way that partial solutions fail fast or slow. Failing fast for the easy puzzle would mean that if you have a partial attempt that is already wrong and you try to fit in the remaining pieces you quickly encounter a gap where none of your remaining pieces fit. Failing slow would mean that you only realize on the last one or two pieces that the solution is wrong and you have to backtrack a lot more.
you believe you could identify each puzzle piece by labeling each puzzle piece by the values of its 4 sides (-2,-1,0,1,2). the puzzle as a whole would have to equal zero in each row and column. That might help in narrowing down which piece can go where
Also place pieces in 45 deg.... I waited for that kind of solution on hard there, but that was no such thing. (with that geometry of pieces, I think you can place four pieces with big holes pointed at the center piece, and then you can place fifth piece at the centre on 45 deg)
@@A11V1R15 its probably a trainned ai algorithm based on previous puzzle algorithm. They wouldnt be able to make a puzzle creator ai from scratch. That is prob the reason it still looks normal. An ai made from zero would probably create an inhuman puzzle with weird shapes
The neat thing about this is that I wont remember the solution so seeing it here doesn't really give it away like other puzzles that can only be solved once because the solution is so memorable.
If you look at the blue puzzle, every single piece has at least one flat side, whereas on the black one two pieces don't, so that makes it easier because you know those two have to be somewhere in the middle, since there are the same number of "pokes" and "holes" and therefore every one should match (you can't have holes near the sides for that to work)
Difficulty is very likely due to the different # of solutions; easier puzzle has more solvable states; where some pieces can be placed at different spots, orientation, etc. Whereas the harder puzzle likely has more connecting pieces which results in unsolvable positions; forcing the user to retry until he randomly stumbles on the correct starting pattern which allows you to solve the puzzle.
I think the difficulty is proportional to how many straight edges are there in total. If you count the straight edges (without the corner pieces) the medium has 10 and hard has 12 of them. It may make connecting them more difficult.
I hope the solutions are unique for both. But you could still assign different difficulties according to how deep you go on average before discovering it is a dead end. If you always find your last piece doesn't fit, it is a hard puzzle; if you get stuck on the 10th piece already, it is easier. That may have to do with how many (close to) identical pieces there are. I suspect the "AI" does that: brute-forcing and counting the number of possible unsolved states, or something like that.
@@landsgevaer AI is nothing more than pattern recognition, it has been shown lots of examples of "good puzzles", it has developed a model of the hallmarks of what humans call a "good puzzle" and now they can reverse the model to produce output that the model thinks are good puzzles. But seriously, this kind of puzzle does not need AI, just an algorythm that does what M said: divide the blocks with a certain number of sides that can fit in other places. The easier the puzzle, the fewer possible locatoins. No AI required. Think of minesweeper or sudoku, they just give you less information to start with, giving you more ways to mess up.
@@vinny142 Yeah I doubt this is actual AI, way too expensive to setup for just a puzzle I think. AI as a term is all over the place though, so of course it's used, as it sells.
Assuming the pieces all fit together they way they look like they should, the first puzzle has 29 solutions and the second has 20 (not counting rotations and reflections of the whole puzzle).
Wow. Higher difficulty is clearly done by having pieces with smooth edges also on the inside, not just around the border. That way you need more combinations to determine the border pieces
Idk if this would work since I haven't bought the puzzle myself. However going down this line of thinking has led me to find some logic in finding the solution. There is 2 constants in this puzzle, one: you will have 4 corner pieces, two: you will have the puzzle completely filled with no excess gaps. With that said here's how I would logically go about it. On a sheet a paper I would mark down all four corners as the letters "A, B, C, and D" from there I would establish what peices could align with the different sized pegs on the corner pieces and line against the wall. I would mark these as "a1, a2, a3..., b1, b2, b3... etc." (Additionally I would make note of which side I checked for on the corner piece.) Lastly, I would compare each matrix and see which one had the least available options, "if corner A had a1 and a2, while corner C had c1, c2, c3, c4, c5, etc. then I would chose a peice from Corner A", from there I would mark the new peice down as E and do the same process. Additionally I would make note of which peice I used and 'remove' it from the available choices. Although you would need to make multiple attempts this could help point you in the right direction
The difference is the pieces in the hard one all have one flat side, so it's harder to figure out where they do in relation to the outside border. The analytical solution is still part bruteforcing it, but what comes to mind is creating a matrix of the pieces in relation to their connectors and realizing possible solutions based on that
@@carterbrown8447 there is a puzzle that you think you need to solve it 2d but it's a 3d structure that builds into a giant house the puzzle itself isn't hard if you knew what you are making but there is a trick to it.
I am from Egypt, i am really happy that finally i could see on my favorite channel a puzzle from my own country solved here, i am really happy that the puzzling community is growing in Egypt
You could try drawing a graph of the whole puzzle, to earlier discover what combinations don't work and have some more idea about which tiles need to be tested earlier
First, I wouldn't go beyond laying out two random corners. If they can't connect at all, they are diametrically placed, in which case, so are the other two. One should also grow arms from these corners and look at how they connect our not connect in the corners if flipped or not. By now you clearly have done impossible configs and some very possible ones. From here, flipping the last pieces together is for toddlers.
I noticed that the medium difficulty puzzle had 4 middle pieces, the difficult one had all edges and corners, increasing the possible locations of the pieces.
The Medium puzzle had 2 pieces with one flat side in the middle. The hard one had four pieces with one flaat side in the middle, therefor there are NO certain middle pieces. I only can imagen, that the easy puzzle has 4 certain middle pieces...
The medium puzzle has corners, edges, and center pieces, so it is significantly more limited on what pieces can go where (the four pieces without a flat edge have to go in the middle) But the Hard puzzle only has corners and "edges," so the four pieces that do go in the center look the same as the ones on the outside, making it a lot less trivial figuring out which pieces go where. There are simply more potential arrangements of the pieces, which means that the solution has a lower likelihood of being found. This puzzle feels like there isn't a real trick to it. It's just brute force, which means that the hard puzzle has a higher difficulty because there are many, many more permutations that need to be tested.
The middle was easy to figure out. I thought halfway into the hard one “Where’s the sides going if there has to be 4 flat faceing eachother?” Which got me to know there were gonna be in a straight line. This would be my approach.
I really doubt they actually use a NN for this. Standard back-tracking can generate these far more easily and faster. And BTW, what would be training data for this NN anyway?
Right? I was kinda excited for a second and than smelled big BS. Now I wonder if you could do web based AI puzzles that optimize towards being hard for humans while having some restriction in complexity (e.g. number of pieces). Jigsaw sucks on PC though, needs to be another kind of puzzle.
12:08 second puzzle has the max number of straight edge pieces that can fit in either the sides or the middle. The medium one has 2 middle pieces with straight edges and the hard puzzle has 4 pieces with straight edges. So, literally, any piece besides the corners, can fit anywhere, whereas the medium difficulty puzzle has 2 pieces which can only be in the middle. That's it really. They could have gone further and have a puzzle with more than 4 corner pieces to mess with the player, but maybe in that case it does not have a unique solution. Not sure.
I believe in order to solve such puzzle you need to start with correct corner configuration. The are only certain ways the corners are arranged so if someone got them wrong to start with he will never get to the final state.
I don't think there is any analytical way to solve these puzzles. I believe they are equivalent to a well known computer science problem called "3-SAT", which computers have to brute force (sort of-you can do slightly better than brute force but not much).
The difficulty difference stems from the number of pieces with a flat (edge) side. In the medium puzzle 14/16 had a flat side, making it a guessing game as to which pieces belonged in the center of the puzzle. In the hard puzzle, 16/16 had a flat side making it even more difficult to discern which pieces belonged in the center of the puzzle. By that logic, I would imagine that the easy puzzle would then have 4 clearly distinguishable center pieces that did not contain a flat side.
Medium puzzle has one flat connection (no small connection or big connection, but a straight cut), hard puzzle has 2 flat connections. I think that more flat connections makes it more complex to solve has a flat side could be a border or go in the middle.
The solved puzzle has its whole space filled. Thats why the sides with big opening has to be filled with the thicker extensions. So I would start with connecting just this type of connection.
I would say complexity is drive by how early one reach impossibility, the earlier the easier, the more arrangement the more complex. Does it make sense? (As the more possibilities the more time you spend going ahead to reach a dead end) and I can see how a computer can calculate this , unrolling it like a chess play looking at every single combination.
If there is only 1 solution, and the easy one has more solutions; I think that is how they give it an other lvl rating. Can you confirm, this idea? The hard puzzle only have edge-pieces (1 flat border) and corner pieces (2 flat borders). With more pieces with 1 flat border, the possible configurations will be much higher, so the try and error method takes longer. Is there a fast method to see or find the solution, except try and error? I don't know. I fear there isn't such a method. The trick you see at 6:01 is very usefull, and can make your search easier. The method of backtracking is a maths method you use in labyrints but also in this kind of puzzle solving. Keep your brain working hard.
For the blue puzzle at 1 hour and 12 mins in, I had the idea of putting the 4 corner pieces in the very center and building outward. I am excited to see the solution
I suspect that the "hard" puzzle has only one solution (not counting rotations and flipping all pieces over), whereas the medium puzzle might have multiple solutions. Thus you might stumble on to a correct soluton sooner with the "medium" puzzle. Also, it occurs to me that writing a program to solve the puzzle wouldn't be too difficult. Every piece has one of 5 possible edges: flat, single bump-out, single bump-in, double bump-out, and double bump-in. First, identify each piece and assign to it the 4 types of edges that it actually has. (You'd probably want to label the pieces with letters A thru P or something.) As you discovered, every bump-out and bump-in must be adjacent to its opposite. And all flat edges must either be on the edge of the puzzle or adjacent to another flat edge. After coding these constraints, it would then be a matter of a recursive search with backtracking to (basically trial and error without repetition) to discover the solution.
First I was skeptical, noticing how there's no clever trick in the puzzles, and thought the devs were just trying to get attention with the term "AI". But I was wrong, the puzzles turned out to be fine and they match the promised difficulty.
difficulty seems to be based on permutations possible for possible solutions at higher piece counts. Pieces seem to have a flow to them, might be a way to approach that way.
A lot of interesting ideas and descriptions on what makes the harder puzzle more difficult. Maybe a good way to go about it would be like you did with the first one. Have all the pieces set but one then keep switch pieces one by one for the other potential until you get the perfect fit.
Thank you for going back to your simpler video format. I stopped watching for months so I'm not sure when you started moving away from the more elaborate intros. But I appreciate it for sure. It felt like you were copying Chris Ramsay, and your personality was so different. It's great to see you going back to simpler format with a much more laidback and relaxing personality. How the channel used to be.
the black puzzle had 10 pieces with one flat side and the blue one had 12, all center pieces on blue side had a flat side wich make more difficult to solve by corner first, sides second and middle end
One thing I've noticed with these two puzzles is that they each have nine little gaps maybe that's a signature thing for this companies puzzles or maybe that's just how yours came out I don't know but it might be a neat little guide
Humans can already create impossibly difficult puzzles. The trick to find the balance between difficult, fun, and creative which is something AIs like this are a long way away from being able to tackle.
It looks like the differance between them is the harder one has many more ways to "almost" be solved, but is actually not possible in that configuration. I think it's specifically designed to not be sloved by logic that a human can acheive with out external tools like measuring device and strong knowlage of mathermatics (if even then it if feasable to work out quicker than by just trying again and again). You would basically have to reverse engineer the algorythem by only seeing the output!
I would have attacked the hard puzzle by trying to build it, and if it fails, swap two corners and try again. After that, try turning one corner bottom to top and try again. After that try swapping two corners again and rebuild. Basically you have to find correct corners and top/bottom orientation for every corner before you can solve the puzzle. Because all the parts fit without extra space the corners pretty much force the positions of all the pieces. I would assume hard is made harder than having more similar pieces (e.g. having inner wide, inner double, outer wide side in clockwise order but having different 4th side for multiple pieces). That way, the possibility of using wrong part is higher and you have to brute force the solution.
Looking at those puzzles, I know it's common to start from the corners, but thinking about it, if you know you have to have it fit a certain way, you might want to start from the center by finding the four middle pieces and figure out all the ways they can fit together and work outward instead of the typical jigsaw puzzle style of working from the corners and edges inward. ---- Also I'm seeing that method would only work on the medium, and most likely easy, puzzles as they have clear center pieces. The Hard puzzle has some inner pieces that could pass as edge pieces, which is why it was innately harder.
I can’t understand why one is so much more difficult than the other when they appear to be so similar. That’s the real question. Pretty amazing!! Nice to see you. Hope all is good Mr.Puzzle 😀☕️
I'm pretty sure what makes the Hard puzzle more difficult is the fact that most of the pieces have flats sides on them. This makes them harder to places on the inside of the puzzle.
The hard puzzle is really difficult to solve. Absolutely love the concept of the puzzle. The uniqueness makes every puzzle very personal. I was wondering while watching this video... will Chris Ramsey solve the hard puzzle faster than you?
It looks like the medium puzzle has some 4 sided pieces unlike the hard puzzle which only has 3 sided and corner so the 4 sided are easier to identify as middle pieces
I think your first problem is that it took you way too long to vary your corner locations. Second, logically the pieces that have the double female ends on opposite sides should be somewhere in the middle as that is where it offers the most possible combinations of connections. And that is only true because you have more flat sides than there are border pieces. If you had counted the flat pieces instead of the connectors you would have seen there was 4 more which would dictate that they need to be on the inside. That combined with the uniqueness of the two pieces should suggest that at the minimum they should be connected to each other on the flat sides since it removes 2 possibilities from the total combinations because they are symmetric. It's the same on the medium puzzle. The unique pieces are on tbe inside
Maybe the difference in the difficulty comes from the corner pieces. Perhaps in the easy one there are solutions possible for more than one corner configuration, but in the hard one maybe there is only one solution. But that's just speculations.
It says one of a kind, but there is obviously a finite number of puzzles to be made using 16 pieces and five possible edge types, especially considering it seems that a maximum of 2 identical edge types can be used on any single piece. Any math wizards able to figure out the number of possible puzzles one could create with those constraints? Also, if they chose to make one of the pieces with four flat edges, would that make the puzzle easier or harder?
The differences between medium and hard are all about the line of logic. The so called medium puzzle has all its clues presented to you on the outset, meaning if you start at any logical starting point (i.e. corners or particularly unique pieces) you’ll find the solution by luck or by following your own thinking. The hard one is actually very smart. It breaks its own loop of logic into more of a spiral, starting somewhere in the middle and looping around into something akin to an infinity symbol, tucking under its self in a manner of speaking. It’s easiest to see with the interior straight lines, using them as a guideline where the lines of logic “overlap” Essentially, the medium one is a closed loop of thinking, guiding you around in a more traditional puzzle way. The hard one requires more meditating on the possible ways to make a puzzle, more than solve it
Honestly these seem quite simply made by bruteforcing Not a Neural network but just taking a 4×4 grid and then carving out a random arrangement of pieces. This would still create a feasible amount of unique puzzles due to the 3^24(?) possible connectors (4^24 on hard). The reason I think this approach is very likely is due to the way the hard vs medium difficulty is made. On hard it just adds the flat corner as a requirement for every piece. That is *not* something the algorithm would do on its own, but something the algorithm would be written to do. On a larger setboard you might even find artifacts of this approach if done inperfectly (like all the flat sides on the hard puzzle being in a line), but on this setboard I couldn't identify such a problem. But overall a really neat puzzle. Just having a bit of an annoyance with the marketing saying it was made by a neural network instead of just randomized (in general there's just a shitton of misinformation about AI). Also of course great they put a good bit of the profit to good causes.
@@Mr.Puzzle I don't know, just the way you said it in the video seemed to imply that because the hour long session was on the previous day that it didn't count the same.
I sincerely doubt this has anything to do with AI. It looks algorithmically generated, that doesn't imply "AI" of any kind. A cursory glance, difficulty is probably related to the number of similar pieces (the "hard" puzzle has quite a few nearly identical pieces).
I'm really enjoying your detailed feedback and analysis of these puzzles and how you would approach it! Thanks a lot! Keep on puzzling! 🙏❤️
I think the difficulty is measured and reassured by how many close, put not quite solutions there are. Also brute forcing these would be extremely difficult. The amount of combinations of locations for each piece is 16! = 2.1*10^13 and that is without factoring in the different rotations. There are definitely some packing algorithms that could solve it, but I think as a regular human, just going through all the potentially possible options until you find it is the most consistent way. Knowing the corner pieces, what sides can match with what over sides drastically cut down on the amount of combinations you have to weed through.
im not sure about these kinda puzzles, but to my knowledge the difficulty is defined by the layout that it is cut in, and the important part is to find the right order of the corners. in the second puzzle its harder, cause the cut layout uses straight lines mostly in the middle, instead of the notches, and nods.
I think the difficulty is realized by the number of pieces that has one flat side in the inner part of the puzzle... Maybe.
@@timperman9883 That's is the right amount of locations, not including rotations. There are 8 different rotations. 16! * 8! means 843,606,888,284,160,000 (8.4*10^17) different possibilities.
@@rondenverliwanag6784 I think you're right, in the "hard" variant you cannot determine in advance which pieces will be edge pieces since every piece has a flat edge. And in the "medium" variant only 2 pieces have no flat edges. I would assume an "easy" variant has 4 pieces with no flat edges which would make them obvious center pieces like a normal jigsaw puzzle.
I like how you warn from getting spoiled the solution of a one of a kind puzzle lol
In the Medium puzzle there were pieces that clearly had to be arranged in the middle because they had no flat side. In the hard one every piece had at least one flat side, so you couldn't figure out where the pieces would go. I think this makes it harter because there are way more possibilities.
I believe you are absolutely right. I noticed the same thing and came to the comments section to speak about it. Also I would have rated the second puzzle a difficulty of 5/5.
The two pieces on the middle left side have flat sides on them.
I expect the medium one just has more possible solutions than the hard one
But, you can start to think like this "So, if all pieces have one flat side, when i start to solve, every time i want to make sure that i made moves that allows the remainings spaces to be 3 connectors and 1 flat side". So, when you are putting the puzzle together and see that one part it's not possible to put 3 connects and 1 flat, you change the layout. I think this make a little bit easier to solve this kind of puzzle
The number of pieces with flat sides doesn't have to be the only difference between the different difficulty approaches. Apart from possibly more non-unique solutions in the easier difficulties, they could be designed in a way that partial solutions fail fast or slow. Failing fast for the easy puzzle would mean that if you have a partial attempt that is already wrong and you try to fit in the remaining pieces you quickly encounter a gap where none of your remaining pieces fit. Failing slow would mean that you only realize on the last one or two pieces that the solution is wrong and you have to backtrack a lot more.
you believe you could identify each puzzle piece by labeling each puzzle piece by the values of its 4 sides (-2,-1,0,1,2). the puzzle as a whole would have to equal zero in each row and column. That might help in narrowing down which piece can go where
almost like sudoku lol
@@DaftFader what?
@@DaftFader I'd equate it more to picross myself
@@KhoiV every row/column/sector in sudoku adds up to 45. Some of the advanced sudoku puzzles heavily rely on this for the solution.
Man, i can't believe that I independently came up with the same method as you, lol.
I fear the day the day when the AI realizes it can cut out edges of the pieces without connecting to other parts, just to distract people.
Also place pieces in 45 deg.... I waited for that kind of solution on hard there, but that was no such thing. (with that geometry of pieces, I think you can place four pieces with big holes pointed at the center piece, and then you can place fifth piece at the centre on 45 deg)
It doesn't looks like AI, looks like a regular algorithm, AI things have a tendency to look way more organic than this puzzle
@@A11V1R15 its probably a trainned ai algorithm based on previous puzzle algorithm. They wouldnt be able to make a puzzle creator ai from scratch. That is prob the reason it still looks normal. An ai made from zero would probably create an inhuman puzzle with weird shapes
Love how you deliberated on the spoiler break then split the difference by putting a super short spoiler break anyway. That was very funny.
The neat thing about this is that I wont remember the solution so seeing it here doesn't really give it away like other puzzles that can only be solved once because the solution is so memorable.
the main reason you won't remember the solution is because it's one of a kind
If you look at the blue puzzle, every single piece has at least one flat side, whereas on the black one two pieces don't, so that makes it easier because you know those two have to be somewhere in the middle, since there are the same number of "pokes" and "holes" and therefore every one should match (you can't have holes near the sides for that to work)
Difficulty is very likely due to the different # of solutions; easier puzzle has more solvable states; where some pieces can be placed at different spots, orientation, etc. Whereas the harder puzzle likely has more connecting pieces which results in unsolvable positions; forcing the user to retry until he randomly stumbles on the correct starting pattern which allows you to solve the puzzle.
I think the difficulty is proportional to how many straight edges are there in total. If you count the straight edges (without the corner pieces) the medium has 10 and hard has 12 of them.
It may make connecting them more difficult.
I hope the solutions are unique for both.
But you could still assign different difficulties according to how deep you go on average before discovering it is a dead end. If you always find your last piece doesn't fit, it is a hard puzzle; if you get stuck on the 10th piece already, it is easier. That may have to do with how many (close to) identical pieces there are.
I suspect the "AI" does that: brute-forcing and counting the number of possible unsolved states, or something like that.
@@landsgevaer AI is nothing more than pattern recognition, it has been shown lots of examples of "good puzzles", it has developed a model of the hallmarks of what humans call a "good puzzle" and now they can reverse the model to produce output that the model thinks are good puzzles.
But seriously, this kind of puzzle does not need AI, just an algorythm that does what M said: divide the blocks with a certain number of sides that can fit in other places. The easier the puzzle, the fewer possible locatoins. No AI required.
Think of minesweeper or sudoku, they just give you less information to start with, giving you more ways to mess up.
@@vinny142 Yeah I doubt this is actual AI, way too expensive to setup for just a puzzle I think. AI as a term is all over the place though, so of course it's used, as it sells.
Assuming the pieces all fit together they way they look like they should, the first puzzle has 29 solutions and the second has 20 (not counting rotations and reflections of the whole puzzle).
if there were puzzles like this in JRPGs, i'd never finish any of those games.
Wow. Higher difficulty is clearly done by having pieces with smooth edges also on the inside, not just around the border. That way you need more combinations to determine the border pieces
Idk if this would work since I haven't bought the puzzle myself. However going down this line of thinking has led me to find some logic in finding the solution. There is 2 constants in this puzzle, one: you will have 4 corner pieces, two: you will have the puzzle completely filled with no excess gaps. With that said here's how I would logically go about it. On a sheet a paper I would mark down all four corners as the letters "A, B, C, and D" from there I would establish what peices could align with the different sized pegs on the corner pieces and line against the wall. I would mark these as "a1, a2, a3..., b1, b2, b3... etc." (Additionally I would make note of which side I checked for on the corner piece.) Lastly, I would compare each matrix and see which one had the least available options, "if corner A had a1 and a2, while corner C had c1, c2, c3, c4, c5, etc. then I would chose a peice from Corner A", from there I would mark the new peice down as E and do the same process. Additionally I would make note of which peice I used and 'remove' it from the available choices. Although you would need to make multiple attempts this could help point you in the right direction
More of those detailed descriptions! Thanks!
there's something very calming about watching someone solve a puzzle.
That's why solving puzzles is part of ergotherapy. It's calming and rewarding.
Increasing the number of flat edges makes it harder to determine which ones are actually inner or outer pieces, thus increases the difficulty.
The difference is the pieces in the hard one all have one flat side, so it's harder to figure out where they do in relation to the outside border. The analytical solution is still part bruteforcing it, but what comes to mind is creating a matrix of the pieces in relation to their connectors and realizing possible solutions based on that
Imagine asking for "Impossible" and it just takes away one piece 😂
Or adds one
@@carterbrown8447 there is a puzzle that you think you need to solve it 2d but it's a 3d structure that builds into a giant house the puzzle itself isn't hard if you knew what you are making but there is a trick to it.
Difference in difficulty is in the number of red herrings (fake flat pieces)
I am from Egypt, i am really happy that finally i could see on my favorite channel a puzzle from my own country solved here, i am really happy that the puzzling community is growing in Egypt
You could try drawing a graph of the whole puzzle, to earlier discover what combinations don't work and have some more idea about which tiles need to be tested earlier
Love your work keep it up. Excellent puzzles👍
Thanks!
First, I wouldn't go beyond laying out two random corners. If they can't connect at all, they are diametrically placed, in which case, so are the other two.
One should also grow arms from these corners and look at how they connect our not connect in the corners if flipped or not.
By now you clearly have done impossible configs and some very possible ones.
From here, flipping the last pieces together is for toddlers.
I noticed that the medium difficulty puzzle had 4 middle pieces, the difficult one had all edges and corners, increasing the possible locations of the pieces.
Lol @ including the spoiler warning break on a one-of-a-kind puzzle
The Medium puzzle had 2 pieces with one flat side in the middle. The hard one had four pieces with one flaat side in the middle, therefor there are NO certain middle pieces.
I only can imagen, that the easy puzzle has 4 certain middle pieces...
The medium puzzle has corners, edges, and center pieces, so it is significantly more limited on what pieces can go where (the four pieces without a flat edge have to go in the middle)
But the Hard puzzle only has corners and "edges," so the four pieces that do go in the center look the same as the ones on the outside, making it a lot less trivial figuring out which pieces go where. There are simply more potential arrangements of the pieces, which means that the solution has a lower likelihood of being found.
This puzzle feels like there isn't a real trick to it. It's just brute force, which means that the hard puzzle has a higher difficulty because there are many, many more permutations that need to be tested.
The middle was easy to figure out. I thought halfway into the hard one “Where’s the sides going if there has to be 4 flat faceing eachother?” Which got me to know there were gonna be in a straight line. This would be my approach.
Yeah, this is a good observation which should remove some random trials. The 4 middle pieces have to be paired into flat sides touching.
I really doubt they actually use a NN for this. Standard back-tracking can generate these far more easily and faster. And BTW, what would be training data for this NN anyway?
Right? I was kinda excited for a second and than smelled big BS.
Now I wonder if you could do web based AI puzzles that optimize towards being hard for humans while having some restriction in complexity (e.g. number of pieces).
Jigsaw sucks on PC though, needs to be another kind of puzzle.
@@cougarten INB4 the new captchas are "solve this puzzle" instead of "click all squares with traffic lights"
yep, just a pseudo random generator and a A* algorythm. How many way this is possible ? 2 ? medium difficulty ship it ! nothing like an AI...
I was hoping to find their methodology on their site and came back empty handed. Very sparse website.
12:08 second puzzle has the max number of straight edge pieces that can fit in either the sides or the middle. The medium one has 2 middle pieces with straight edges and the hard puzzle has 4 pieces with straight edges. So, literally, any piece besides the corners, can fit anywhere, whereas the medium difficulty puzzle has 2 pieces which can only be in the middle. That's it really. They could have gone further and have a puzzle with more than 4 corner pieces to mess with the player, but maybe in that case it does not have a unique solution. Not sure.
I believe in order to solve such puzzle you need to start with correct corner configuration. The are only certain ways the corners are arranged so if someone got them wrong to start with he will never get to the final state.
I don't think there is any analytical way to solve these puzzles. I believe they are equivalent to a well known computer science problem called "3-SAT", which computers have to brute force (sort of-you can do slightly better than brute force but not much).
The difficulty difference stems from the number of pieces with a flat (edge) side. In the medium puzzle 14/16 had a flat side, making it a guessing game as to which pieces belonged in the center of the puzzle. In the hard puzzle, 16/16 had a flat side making it even more difficult to discern which pieces belonged in the center of the puzzle. By that logic, I would imagine that the easy puzzle would then have 4 clearly distinguishable center pieces that did not contain a flat side.
Medium puzzle has one flat connection (no small connection or big connection, but a straight cut), hard puzzle has 2 flat connections. I think that more flat connections makes it more complex to solve has a flat side could be a border or go in the middle.
The solved puzzle has its whole space filled. Thats why the sides with big opening has to be filled with the thicker extensions. So I would start with connecting just this type of connection.
I would say complexity is drive by how early one reach impossibility, the earlier the easier, the more arrangement the more complex. Does it make sense? (As the more possibilities the more time you spend going ahead to reach a dead end) and I can see how a computer can calculate this , unrolling it like a chess play looking at every single combination.
This project is so brilliant... given different shapes and incongruent designes; the possibilities for a layout are infinite
... fractals
Nice video, nice music, nice dude! Thanks, was interesting and enjoyable to watch! 🙂
Here's my detailed feedback and analysis: "Mr. Puzzle? More like Mr. Hand-model"
Thank you have a nice day.
🙌
The harder / blue puzzle.. had 2 extra flat edge pieces in the middle. - 1 factor i think makes the arranging options more difficult to narrow down.
If there is only 1 solution, and the easy one has more solutions; I think that is how they give it an other lvl rating. Can you confirm, this idea?
The hard puzzle only have edge-pieces (1 flat border) and corner pieces (2 flat borders). With more pieces with 1 flat border, the possible configurations will be much higher, so the try and error method takes longer.
Is there a fast method to see or find the solution, except try and error? I don't know. I fear there isn't such a method. The trick you see at 6:01 is very usefull, and can make your search easier.
The method of backtracking is a maths method you use in labyrints but also in this kind of puzzle solving.
Keep your brain working hard.
Human mind and power still beat AI 👍nice puzzles
If you are sure the gaps are all filled in put three corners to the side and just start with one corner.
For the blue puzzle at 1 hour and 12 mins in, I had the idea of putting the 4 corner pieces in the very center and building outward. I am excited to see the solution
Makes me wonder what the first fully automated puzzle would look like.
We will find out one day 😁
You're already living in the matrix, mate.
@@bluellamaslearnbeyondthele2456 Good one
I suspect that the "hard" puzzle has only one solution (not counting rotations and flipping all pieces over), whereas the medium puzzle might have multiple solutions. Thus you might stumble on to a correct soluton sooner with the "medium" puzzle.
Also, it occurs to me that writing a program to solve the puzzle wouldn't be too difficult. Every piece has one of 5 possible edges: flat, single bump-out, single bump-in, double bump-out, and double bump-in. First, identify each piece and assign to it the 4 types of edges that it actually has. (You'd probably want to label the pieces with letters A thru P or something.) As you discovered, every bump-out and bump-in must be adjacent to its opposite. And all flat edges must either be on the edge of the puzzle or adjacent to another flat edge. After coding these constraints, it would then be a matter of a recursive search with backtracking to (basically trial and error without repetition) to discover the solution.
Really unique and very interesting. I totally enjoyed this 😃
First I was skeptical, noticing how there's no clever trick in the puzzles, and thought the devs were just trying to get attention with the term "AI". But I was wrong, the puzzles turned out to be fine and they match the promised difficulty.
I'm still skeptical about the AI part.
For me, it looks like generated by a handcrafted algorithm. I can't see how it needed AI.
@@Mr_Yeah an AI is an algorithm, it’s probably being improved on so they can generate more difficult layouts. Kinda like a Chess Engine.
@@Alsry1 Sorry, my bad. I've mixed up AI with machine learning.
difficulty seems to be based on permutations possible for possible solutions at higher piece counts. Pieces seem to have a flow to them, might be a way to approach that way.
A lot of interesting ideas and descriptions on what makes the harder puzzle more difficult. Maybe a good way to go about it would be like you did with the first one. Have all the pieces set but one then keep switch pieces one by one for the other potential until you get the perfect fit.
Thank you for going back to your simpler video format. I stopped watching for months so I'm not sure when you started moving away from the more elaborate intros. But I appreciate it for sure. It felt like you were copying Chris Ramsay, and your personality was so different. It's great to see you going back to simpler format with a much more laidback and relaxing personality. How the channel used to be.
me and a friend of mine made a version of that puzzle in the past. The difficulty is based on how many options you have to place the pieces.
I was struggling so much on the hard puzzle...
Thanks for the solution 😄
the black puzzle had 10 pieces with one flat side and the blue one had 12, all center pieces on blue side had a flat side wich make more difficult to solve by corner first, sides second and middle end
One thing I've noticed with these two puzzles is that they each have nine little gaps maybe that's a signature thing for this companies puzzles or maybe that's just how yours came out I don't know but it might be a neat little guide
🤔
It has the gaps because of the rounded corners on the pieces.
Amazing. Love your videos
What an awesome concept!
Great puzzles. You need to try the easy one.
I would assume the easy one has 4 pieces that can only fit in the center
Brian Young. You are a genius!!
Solve each corner first, as in each 2x2 square. Find ways they could match up, brute force from there if necessary.
Algorithm-created puzzles are going to take the difficulty of puzzles to a whole new level.
Humans can already create impossibly difficult puzzles. The trick to find the balance between difficult, fun, and creative which is something AIs like this are a long way away from being able to tackle.
It looks like the differance between them is the harder one has many more ways to "almost" be solved, but is actually not possible in that configuration. I think it's specifically designed to not be sloved by logic that a human can acheive with out external tools like measuring device and strong knowlage of mathermatics (if even then it if feasable to work out quicker than by just trying again and again). You would basically have to reverse engineer the algorythem by only seeing the output!
Perhaps it'd be easier if you solved the centre first?
I would have attacked the hard puzzle by trying to build it, and if it fails, swap two corners and try again. After that, try turning one corner bottom to top and try again. After that try swapping two corners again and rebuild. Basically you have to find correct corners and top/bottom orientation for every corner before you can solve the puzzle. Because all the parts fit without extra space the corners pretty much force the positions of all the pieces.
I would assume hard is made harder than having more similar pieces (e.g. having inner wide, inner double, outer wide side in clockwise order but having different 4th side for multiple pieces). That way, the possibility of using wrong part is higher and you have to brute force the solution.
Next video: Solving the Eternity II puzzle ;)
en.wikipedia.org/wiki/Eternity_II_puzzle
Looking at those puzzles, I know it's common to start from the corners, but thinking about it, if you know you have to have it fit a certain way, you might want to start from the center by finding the four middle pieces and figure out all the ways they can fit together and work outward instead of the typical jigsaw puzzle style of working from the corners and edges inward.
----
Also I'm seeing that method would only work on the medium, and most likely easy, puzzles as they have clear center pieces. The Hard puzzle has some inner pieces that could pass as edge pieces, which is why it was innately harder.
always think that corner pieces don't have to belong in corners.
thank you very much for this beautiful video ♥
imo ugly4 and 8 cats are superior aesthetically
This puzzle was designed to be as engaging as possible so that the puzzle solver will not notice the robot insurrection
I can’t understand why one is so much more difficult than the other when they appear to be so similar. That’s the real question.
Pretty amazing!!
Nice to see you. Hope all is good Mr.Puzzle 😀☕️
because you're wrong and they're not similar
I think the difference is the no. of pieces that only fit the center
Easy probably has 4 of them.
@@Mr.Puzzle that makes sense, I think you are right 😀
One thing to make note is that flat sides can connect with flat sides, I make note of this because nobody considers the flat side
I'm pretty sure what makes the Hard puzzle more difficult is the fact that most of the pieces have flats sides on them. This makes them harder to places on the inside of the puzzle.
having a flat-side puzzle in the middle will add more probability of combination because you're not limiting them on the sides.
Alleluia..... 😇
Very cool puzzles. I found a different solution for the blue one, so there are at least 2 possible answers.
If you have one of these puzzles keep in mind that NO TWO PUZZLES are the same.
I think the dificulty is based on the posibel solutions to the puzzle the more solutions the easier it is
The hard puzzle is really difficult to solve.
Absolutely love the concept of the puzzle. The uniqueness makes every puzzle very personal.
I was wondering while watching this video... will Chris Ramsey solve the hard puzzle faster than you?
Mr Puzzle defeated the Puzzle Design AI !
the reflection on that puzzle looks like a space station.
It looks like the medium puzzle has some 4 sided pieces unlike the hard puzzle which only has 3 sided and corner so the 4 sided are easier to identify as middle pieces
I think your first problem is that it took you way too long to vary your corner locations.
Second, logically the pieces that have the double female ends on opposite sides should be somewhere in the middle as that is where it offers the most possible combinations of connections. And that is only true because you have more flat sides than there are border pieces. If you had counted the flat pieces instead of the connectors you would have seen there was 4 more which would dictate that they need to be on the inside. That combined with the uniqueness of the two pieces should suggest that at the minimum they should be connected to each other on the flat sides since it removes 2 possibilities from the total combinations because they are symmetric.
It's the same on the medium puzzle. The unique pieces are on tbe inside
it looks liek it's more diffecult because every piece has a flat side, unliek the one before that
Make sure to put the corners in last.
Maybe the difference in the difficulty comes from the corner pieces.
Perhaps in the easy one there are solutions possible for more than one corner configuration, but in the hard one maybe there is only one solution.
But that's just speculations.
It says one of a kind, but there is obviously a finite number of puzzles to be made using 16 pieces and five possible edge types, especially considering it seems that a maximum of 2 identical edge types can be used on any single piece. Any math wizards able to figure out the number of possible puzzles one could create with those constraints? Also, if they chose to make one of the pieces with four flat edges, would that make the puzzle easier or harder?
The differences between medium and hard are all about the line of logic. The so called medium puzzle has all its clues presented to you on the outset, meaning if you start at any logical starting point (i.e. corners or particularly unique pieces) you’ll find the solution by luck or by following your own thinking.
The hard one is actually very smart. It breaks its own loop of logic into more of a spiral, starting somewhere in the middle and looping around into something akin to an infinity symbol, tucking under its self in a manner of speaking. It’s easiest to see with the interior straight lines, using them as a guideline where the lines of logic “overlap”
Essentially, the medium one is a closed loop of thinking, guiding you around in a more traditional puzzle way. The hard one requires more meditating on the possible ways to make a puzzle, more than solve it
Honestly these seem quite simply made by bruteforcing
Not a Neural network but just taking a 4×4 grid and then carving out a random arrangement of pieces.
This would still create a feasible amount of unique puzzles due to the 3^24(?) possible connectors (4^24 on hard).
The reason I think this approach is very likely is due to the way the hard vs medium difficulty is made.
On hard it just adds the flat corner as a requirement for every piece. That is *not* something the algorithm would do on its own, but something the algorithm would be written to do.
On a larger setboard you might even find artifacts of this approach if done inperfectly (like all the flat sides on the hard puzzle being in a line), but on this setboard I couldn't identify such a problem.
But overall a really neat puzzle. Just having a bit of an annoyance with the marketing saying it was made by a neural network instead of just randomized (in general there's just a shitton of misinformation about AI).
Also of course great they put a good bit of the profit to good causes.
Screen shot........ Tinkercad.......3d print.....easy .......😂
In this case it would seem very logical to me if there's more than 1 way to connect the 4 center pieces, and then move to the sides.
"It was really easy after I spent an hour doing it wrong" doesn't really seem like a good metric for giving something a four.
Why? One hours is perfectly fine for a difficult puzzle. For a 5 it must leave me clueless or drive me crazy. 😁
@@Mr.Puzzle I don't know, just the way you said it in the video seemed to imply that because the hour long session was on the previous day that it didn't count the same.
This video made me really angry haha. I wish I could have pushed him out of the way on this one. He was over thinking
Mr.Puzzle's puzzles.
Maybe its easier to start from a corner and make a path until something doesnt match, and then try something else.
Good man✌🏻
I sincerely doubt this has anything to do with AI. It looks algorithmically generated, that doesn't imply "AI" of any kind.
A cursory glance, difficulty is probably related to the number of similar pieces (the "hard" puzzle has quite a few nearly identical pieces).
Wait what happened to the sound? :o
I didn't notice any problem with the sound, other than that I found the background music (with constant clicking percussion) to be rather annoying.
Forgot to hit record during the intro. After the intro the sound is normal.
@@Mr.Puzzle Ah, was the intro just on your camera mic then? I was wondering why it was terrible compared to the audio for the rest of the video.
Just at two and a half minutes, my guess is the 'harder' turns out to be harder? lol
You appeared to have trouble dealing with the concept that each piece is 'double-sided' rather than just having a front-and-back . . .
anytime you see me stream ..come into the live chat,say hi and leave ur link to your page or live stream 💪💪💪💪
You should give the Diamond puzzle a try ;)
Solid background beats