# Can an AI be taught to solve puzzles?

Only 60 people in the world have solved this puzzle. But they managed to do it because they were intelligent. If AI is an artificial intelligence, could the AI also solve this or any other puzzle without using brute force methods like computers normally do?

1. 5 months ago
Anonymous

Yeah but this is something that you can just solve with a constraint solver once the rules are properly encoded. Like just put this shit in prolog and let it run. That's what the field of AI was like some decades ago.

• 5 months ago
Anonymous

But if you're just going to run a brute force search, that doesn't require any intelligence. And the AI is called an intelligence.

• 5 months ago
Anonymous

It is not a brute force search if it takes into account all the given constraints during the search. And by the way how would you even define an "intelligent" solution to such a puzzle? There is no magic element to intelligence that makes you just know the solutions of puzzles without any reasoning behind it.

• 5 months ago
Anonymous

Regardless of whether you use a brute force search or not, you're still taking into account all the constraints of the puzzle am I right? Because otherwise you would not have a unique solution and you would possibly end up with a wrong solution.

"Intelligent solution" would be a solution which you were able to find as easily as possible. That's how I would define it. Something like picrelated as an illustration of my point. A dumb person would use the quartic formula or something to solve the equation in a complicated and robotic way. But a more intelligent person would realize that the problem can be reduced to something simpler and solved in a much easier way. How would you program a computer to do something like that?

• 5 months ago
Anonymous

is "there is a unique solution" part of the clues/puzzle?
>How would you program a computer to do something like that?
It depends. A constraint solver could start by listing the amount of possible numbers according to the current state of the puzzle, then start with the one that has the least number of possibilities (i.e. I see at least 1 square that you can fill in right at the start and know for sure that it's correct). But for some puzzles you can't know what's the easiest path to the solution without actually trying many solutions.

2. 5 months ago
Anonymous

> Only 60 people in the world have solved this puzzle.
That’s because only 60 people have tried. It’s an easy puzzle with no intellectual merit. If you want a real puzzle then do all the Advent of Code problems!

3. 5 months ago
Anonymous

>Only 60 people
*61

That was trivial.

• 5 months ago
Anonymous

Are you in picrel?

4. 5 months ago
Anon0

Are the people who claimed they solved this puzzle actually only using their brains, or just using their brains to write a computer algorithm to do it for them?

5. 5 months ago
Anonymous

fun

• 5 months ago
Anonymous

• 5 months ago
Anonymous

https://app.crackingthecryptic.com/sudoku/Hf9QqQMD7J

6. 5 months ago
Anonymous

mmm, challenge accepted. will post solution if/when I get it

7. 5 months ago
Anonymous

I don't get it. It seems like A should be 1 or 2, and B should be 1,2, or 3.

If B is 3 then that forces a "region" of size 9 (WLOG the blue region in pic). But now A has to be 1, and its own region is stuck between the purple region of size 9, so A's own region must have size <9, contradiction.

If B is 2 or 1, then A must be the opposite, and we quickly get a contradiction based on regions as well. This exhausts all cases, so there is no solution

• 5 months ago
Anonymous

>If B is 3 then that forces a "region" of size 9 (WLOG the blue region in pic). But now A has to be 1, and its own region is stuck between the purple region of size 9
Why couldn't the cell a itself be part of the region that a points to (i.e. the purple region in your diag)?

• 5 months ago
Anonymous

>Why couldn't the cell a itself be part of the region that a points to (i.e. the purple region in your diag)?
Because if A is in region N then those are in region N+1.
>not true
Yes true, because the same number cannot appear twice in that row.

• 5 months ago
Anonymous

d'oh, forgot the basic rules of sudoku while focusing on the weird shit

• 5 months ago
Anonymous

>If B is 2 or 1, then A must be the opposite,
not true

• 5 months ago
Anonymous

>Why couldn't the cell a itself be part of the region that a points to (i.e. the purple region in your diag)?
Because if A is in region N then those are in region N+1.
>not true
Yes true, because the same number cannot appear twice in that row.

• 5 months ago
Anonymous

>If B is 3 then that forces a "region" of size 9 (WLOG the blue region in pic).
>But now A has to be 1
says who

• 5 months ago
Anonymous

>WLOG
blue region could contain all 3 cells right of “B”. this changes everything. retard

8. 5 months ago
Anonymous

"AI" doesn't exist. You can count on it that some corporate chimp is going to train a bot to solve this kind of puzzle at some point and declare some kind of imaginary victory, but no existing or even realistically conceivable iteration of current technologies can deal with this without being specifically trained to.

9. 5 months ago
Anonymous

I'm confident I can solve that, but if a solution is already public, I'm not interested.

10. 5 months ago
Anonymous

1. How do you know that only 60 people have solved it?
2. Is a solution already available? If so, this puzzle offers nothing to the intellectual flexer.

• 5 months ago
Anonymous

I think it's from here
https://logic-masters.de/Raetselportal/Raetsel/zeigen.php?id=000BD7
but when I follow the penpa+ link the UI looks different from what's in the OP