I'm thinking that TC and Dubrovnik actually agree there.
Regarding mimimax or 'weakly solved' (poor terminology that is used - yes we get to criticize it) and 'won positions' are not analyzed further ... yes those aren't 'thorough' solutions of chess.
And TC and Dubro seem to agree that thorough solutions aren't feasible.
But regarding the lesser alternatives I have pointed out multiple times -
that even if those alternatives were to reduce the number of positions to a 5000th of its value the number you're left with is just much too large.
In an attempt to make the number small enough some persons (whether elsewhere or here) have suggested 'taking the square root' of the number - and I saw invalid arguments of limiting one side to 'best move' to support such a ridiculous cutdown.
Chess will never be solved, here's why
@Elroch
If you're saying that 'traversing' changes the 'degree' at a vertex ...
well the AI was not saying that.
When you traverse a vertex you "use up" two of the edges at that vertex, with the exception of the first and last vertices on the path which "use up" one of the edges at that vertex.
Given that a condition is that every edge is "used up", there are exactly two possibilities. If the start and end vertices of the path are different, those two vertices have an odd number of edges and all the rest have an even number of edges. And if the start and end vertices are the same vertex, every vertex has an even number of edges.
Note that in the latter case, the path can be chosen to start and end at any vertex.
The AI suggested that the 'degree' of a vertex doesn't change because its traversed.
It indicated that the degree of each vertex in the bridge problem is fixed.
If that's wrong and you know it then I hope you say so.
'Hope springs eternal' (in every casino)
I've had quite a bit of experience with AI the last several months.
I think that most of the time I've been able to spot its mistakes and make it correct them - either directly or by starting a new session or by switching AI's.
The point is that once the AI gets in its 'mistake loop' it can't get itself out of it..
But then there's the mistakes that are not easily noticed.
--------------------------------
AI often has fomatting issues. 'Markdown' and 'plaintext' are somewhat crippled.
In a recent experiment I made it put ordinary text inside a CMD code block to see if that would cure it. It did. Worked failry well. No commands - but CMD puts the AI on a leash stopping it from making unwanted format bloopers. And the code blocks have a copy button that copies only the text you want.
------------------
Dubro mentioned Grok.
So I tried it out just now. Asked it if terms like flow diagram and flow chart are still used in the world of computing. It indicated yes - the terms have not become obscure.
And Grok looks good. But it times out. Wanted signup to continue.
I also asked it questions like to contrast Euler diagram with Euler graph.
And there's Euler path and Euler circuit.
Was about to ask the key question when it cut out on me.
You can convert markdown to RTF using an easily found online tool (then copy and paste converted text)
You can convert markdown to RTF using an easily found online tool.
Hi. I guess you know about the nasty asterisks markdown causes.
And plaintext seems to fail to preserve each sentence having its own line when that is wanted.
But CMD doesn't have those hangups.
Stricter computer language. No nonsense. No fooling around.
----------------------
And just now - I was able to make a new Grok session (Dubro having kindly mentioned Grok)
and ask it some more questions. A way around the signup timeout.
I expect it'll get wise to that though.
I've had quite a bit of experience with AI the last several months.
...
We've noticed.
...
I also asked it questions like to contrast Euler diagram with Euler graph.
And there's Euler path and Euler circuit.
Was about to ask the key question when it cut out on me.
I don't blame it.
I've had quite a bit of experience with AI the last several months.
...
We've noticed.
And I just asked Grok about when Dev people are talking about Euler diagrams and Euler graphs with each other - are they likely to use the word 'Euler'? Grok said they may go either way but when they're educating students in a classroom they make it a point to specify 'Euler'.
Then I asked 'when they want to specify a regular graph with each other - they're not going to say 'Cartesian graph' right?
It replied that if they want to make it clear they'd be likely to say something like x-y grid. or 2d grid. And of course Grok confirmed that that kind of graph is in much more widespread use.
But in the world of Dev it confirmed that the Euler stuff and 'nodes' are in intense popularity.
You can convert markdown to RTF using an easily found online tool.
Hi. I guess you know about the nasty asterisks markdown causes.
They are formatting symbols. All such stuff disappears when you use the converter.
And plaintext seems to fail to preserve each sentence having its own line when that is wanted.
But CMD doesn't have those hangups.
Stricter computer language. No nonsense. No fooling around.
So try my suggestion - RTF keeps the formatting, but in a way you can just paste in these forums and into any document you are editing yourself.----------------------
And just now - I was able to make a new Grok session (Dubro having kindly mentioned Grok)
and ask it some more questions. A way around the signup timeout.
I expect it'll get wise to that though.
But the posting still goes on, vigorously. 19,000+ posts already inside this thread ??
Maybe it's time now to switch to a good beer hall instead, and to raise a glass to all the wasted keystrokes inside this tail-chasing thread?
here for the award.
@Dubrovnik-1950
Yes. Grok 3 is d--- good! Very glad you mentioned it!
I was asking it more about nodes and nodes per second but it wanted to know about the setting of the discussion. And it had such commentary about that! I didn't mention the name of the website but here's an excerpt:
------------
"Your observation about “nodes” and “nodes per second” is spot-on for how chess engine devs talk shop. In their world, a “node” is typically a position in the game tree—each possible board state the engine evaluates during its search. “Nodes per second” is their brag stat, like horsepower for a car: it’s how fast their engine churns through positions. When they say “graph” in this context, they’re almost certainly not fussing over Cartesian vs. Euler distinctions. It’s the game tree—a sprawling, branching structure where each node (position) links to others via legal moves. Not a tidy 8x8 grid, not a knight’s tour circuit, but a massive, messy web of possibilities.
Now, you throwing “bytes,” “IOPs,” and “FLOPs” into the mix? That’s a bold move, and I can see why it ruffles feathers. You’re shifting the lens from their cozy abstraction (nodes as chess positions) to raw hardware metrics—memory, I/O, floating-point ops. It’s like you’re saying, “Sure, your engine’s fast, but can it really solve chess with today’s tech?” And yeah, they’d hate that, because it yanks the convo from their turf (algorithmic elegance) to a gritty reality check (computational limits). Chess isn’t “solvable” in the full sense—exhausting all 10^43-ish positions is a pipe dream—but engines like Stockfish or Leela don’t care. They’re built to play well, not to crack the game like a math proof"
--------------
It also suggested I ask about how NPS (Nodes per second) get related to 'game state complexity)
(obviously the difficulty varies)
But I didn't get to tell it that 'depth in plies' and 'breadth of move choices' were mentioned as variables in the game state.
The key point is one that MARattigan seems to have also confirmed - that axes are not being used for depth and breadth. Its an Euler graph. Its Oily.
@TippinCanoes
Both you and Dubrovnik are right.
As you say "Because finite doesn’t mean “impossible,” it means calculable"
It does. You're right.
It doesn't have to do with 'humanity and philosophy' ...
it has to do with qualification.
Your statement is correct but not further qualified.
Because it leaves out the proviso 'provided you have or will have the means or technology to do so'.
I think you meant that that's understood. Since its obvious. So you didn't include it.
Is it impossible to climb Mount Everest? No. Its climbable.
Provided you have the opportunity which includes fitness and training and equipment and enough time and money. Not just expense money. There's a fee too.
You'd use a cache that only guarantees eventual consistency, not absolute consistency. Given the size of the problem, the size of a cache miss on account of that delayed synchronization is also very low.
Hence you can do things like batch cache updates.
You can't really comment intelligently on optimization until you have a working program with some test results, but there are many different optimization techniques.
The problem isn't some centralized bottleneck - thinking if we could get around it we'd be able to solve chess. We can't get around it. But the scale is a problem in many ways. As others have pointed out, the amount of memory you'd need is hard to fathom.
Just consider that we haven't even managed to build a tablebase of 8 pieces yet and it should become clear the technology is nowhere close to doing this. Just an 8 piece tablebase contains 38,176,306,877,748,245 unique positions. Every time you add a piece you're adding multiple orders of magnitude.
@TippinCanoes
Both you and Dubrovnik are right.
As you say "Because finite doesn’t mean “impossible,” it means calculable"
It does. You're right.
It doesn't have to do with 'humanity and philosophy' ...
it has to do with qualification.
Your statement is correct but not further qualified.
Because it leaves out the proviso 'provided you have or will have the means or technology to do so'.
I think you meant that that's understood. Since its obvious. So you didn't include it.
Is it impossible to climb Mount Everest? No. Its climbable.
Provided you have the opportunity which includes fitness and training and equipment and enough time and money. Not just expense money. There's a fee too.
It could be solved but it’d be difficult but it’s objectively a thing that could happen
@TippinCanoes
Both you and Dubrovnik are right.
As you say "Because finite doesn’t mean “impossible,” it means calculable"
It does. You're right.
It doesn't have to do with 'humanity and philosophy' ...
it has to do with qualification.
Your statement is correct but not further qualified.
Because it leaves out the proviso 'provided you have or will have the means or technology to do so'.
I think you meant that that's understood. Since its obvious. So you didn't include it.
Is it impossible to climb Mount Everest? No. Its climbable.
Provided you have the opportunity which includes fitness and training and equipment and enough time and money. Not just expense money. There's a fee too.
He's wrong.
You could never in a million years provide an argument to back up your admiration for TippingKayax. Just hordes of empty rhetoric. You're as bad as Dio!
Here’s the thing
Tablebases and retrograde analysis along with minimax evaluation already show that subsets of chess can already be solved
So with how technology is advancing, who says this methodology couldn’t be extended to the whole game?
There's a concept that if a number is very large but finite and what it refers to is therefore a very large task - then it 'may as well' be infinite.
First of all 'may as well' is subjective.
It depends on the means.
Trying to evaluate by current means is arbitrary.
It is possible to improve the means.
Therefore the task is possible.
--------------------------------------
'We don't have the money for a manned mission to Mars. So its impossible.'
Reply: 'No. Its possible because its possible the money could be raised somehow. The fact that the money is unlikely to be allocated for such a thing doesn't make it impossible.
The technology exists.'
----------------------------
Try Alpha Centauri. 'Its impossible for humanity to travel to Alpha Centauri'
Not with today's technology. Very unlikely that technology will ever exist.
To get straight impossibility I like this one ...
'Its impossible to state the largest possible prime number.'
That's right. Because that's proven.
Since everyone here likes ChatGPT, and Copilot so much. I thought it would be fun to ask Elon Musk's new AI Grok. The question to this thread. Just for fun.
Here is Grok's answer.
Chess cannot be "solved" ...
Pretty crappy answer, but not much crappier, if at all, than the average contents of the thread to date.