Chess will never be solved, here's why

I stand corrected. And you're making great posts.
See my previous post with question about weight ratio to Grok.
I asked what a Threadripper is and got this:

"A Threadripper isn’t a desktop computer itself—it’s a line of high-performance CPUs (central processing units) made by AMD, specifically designed for desktops and workstations.
The Ryzen Threadripper series is known for its insane core counts, multithreading capabilities, and raw power, aimed at enthusiasts, gamers, content creators, and professionals who need serious computing muscle.
So, you’d find a Threadripper inside a desktop computer, typically a custom-built one, not as a standalone device. Think of it as the brain that could power a beastly desktop rig, not the whole machine"
------------------------------------
@MARattigan I saw this too:
"Additionally, there's a JavaScript task runner called Grunt, which is not an AI but a tool used for automating tasks in web development projects."

It definitely seems to have an edge over chatgpt which has a slight edge over copilot.

@ Dubrovnik
I got this:

Ratio of FLOPS per Pound (Sycamore to El Capitan)
Calculation:
(2.17 × 10¹⁴ FLOPS/lb) ÷ (2.79 × 10¹² FLOPS/lb) ≈ 77.78.
Ratio: Sycamore’s FLOPS per pound is approximately 77.8 times higher than El Capitan’s.
(per pound of hardware)
--------------------
Sycamore is one of the top Qubit computers - but its range of application is very narrow compared to the supercomputers like Capitan and Frontier.
Perhaps that will change and Quantum computers will ratchet up the speed by a factor of 1000.
But that isn't a big enough 'ratchet'.
That rachet needs a hatchet. (can't resist pun connection to Nurse Ratched and Ratchett in the Orient Express movie)

Looks like the AI takeover is progressing faster than I thought, it has already taken over this thread

AI is relevant to computers and computers are relevant to the forum subject.
AI is also a source of relevant information.
Someone mentions AI and the moon is going to explode?
Should we put a ban on the words 'nodes' and 'clock cycles per second' and 'AI'?
How about a ban on the phrase 'current software' or on 'Stockfish' ...
What about a ban on the word 'that'? Egads !

White wins in 64 moves.

Dubrovnik, that is rich coming from you... calling other members names, having yourself posted unannounced AI essays and wikipedia articles with no elaborated opinion, cheating accusations based on wrongly interpreted data, and various other dubious things. You have made interesting posts, but clearly the wide range of your standards includes some total bs as well.

Grok 3 is up there with the best, but not clearly better than them. It also has a weakness - Elon Musk.

The model has already been hastily crippled to modify its responses to questions relating to Musk, much like the CCP enforced Deepseek to not give answers it did not like (instead of honest answers to questions about "sensitive" topics, it slurges a stock CCP statement. I can understand that when Musk discovered that if you asked Grok 3 "who are the biggest spreaders of lies and misinformation in the modern world" his name was on a shortlist of two, but "fixing" this unwanted honesty is not something desired by users of an AI for dealing with real world facts.

ChatGPT-4o will give an answer that is approximately as good. It's not a particularly good test, because there isn't much substance to the answer.

I hear that where ChatGPT beats Grok 3 hands down is with the Deep Research extension. If one thing is worth subscribing for, it is probably this.

Personally I like the fact that when I run into my limit of usage for ChatGPT-4o, there are always alternatives like Grok 3!

You say it is a great test, then you agree with me - it's not very difficult (it's also not very precise).

I asked Grok 3:

"Is it realistic that computing would advance enough to fullfil the requirements for a weak solution of chess?"

Grok 3 responds:

"Realistic Timeline - If computing power doubles every two years (a stretch, given current trends), reaching 10^43 flops would take about 140 years from 10^18 flops (log2(10^25) × 2 years). That’s mid-22nd century, assuming no plateau. But with algorithmic leaps, it could happen sooner. Solving 7-piece endgames took decades of incremental progress; scaling to 32 pieces might need a paradigm shift—like a theory of chess that simplifies the game into solvable chunks. Some estimate a weak solution could be feasible by 2100 if both hardware and software align, but others argue it’s centuries away—or even intractable without a revolutionary approach.

My Take - It’s realistic that computing could advance enough, but not soon—think late 21st or 22nd century at best, barring a genius breakthrough. The bigger limiter isn’t hardware but our ability to outsmart the problem. Chess’s beauty lies in its depth; solving it might require not just faster machines but a deeper understanding of its structure. Do you think we’ll get there by cracking the math first, or just throwing more silicon at it?"

Interestingly its answer does consider the possibilities of conceptualizing and modeling chess positions mathematically ie. "algorithmic leaps". Grok 3 considers a weak solution a realistic though unlikely possibility if the timeline is a 100+ years, maybe sooner depending on how AIs mathematical/conceptual ability advances.

I absolutely hate to post AI stuff like this but thats the theme atm.

It was precise, and the facts and logic were correct.

I have noticed Grok's reasons and logic, and facts. Have still yet to be refuted....we are waiting.

Patience.

I'm working on it. I don't want to post a response that's just as superficial and inaccurate as the post I'm criticising.

You can, and appear to be intent on, posting many long screeds of such stuff, because you can churn it out of your threadripper at 1000 posts/sec.

I work at a slower speed and aim for quality over quantity.

I dont know what you mean but Grok 3 understands what a weak solution is:

A game is considered "weakly solved" when, for any given position, an algorithm can determine a strategy that guarantees either a win or a draw for one of the players, assuming perfect play from that point onward. For chess, this would mean finding a strategy that, from the starting position, ensures at least a draw (or possibly a win) for White, and similarly for Black in response.

Which means Grok 3 doesn't understand what a weak solution is. (Do you?)

No such thing as the 7 man tablebase. There are 7 man DTC tablebases or 7 man Syzygy tablebases and there used to be 7 man Lomonosov DTM tablebases. Whether they are weak solutions depends what you mean by "chess". Under what I generally refer to as "competition rules" chess, the DTC and Lomonosov tablebases are not any kind of solution to some of the positions. (Ask Grunt.)

Anyhow, clearly Grok 3 doesnt agree with your view that weakly solving chess from the starting position is physically impossible in some humanly realistic timeline.

Just out of interest, was the other candidate Gronk 3 or the man with the funny haircut?

To be less biased it does here and it doesn't here.

That's why you need to stop posting the stuff and address the question with real, rather than superficial, intelligence.

Chess cannot be solved by classical methods, but maybe future A.I. in conjunction with brute force. But the issue is that humans will always play mind games, and play sub-par moves, and make mistakes. So it cannot truly be solved for human play. Further, A.I. itself makes weird moves to try and 'trick' the opponent engine. That is not strictly part of Chess being 'solved' in classical terms. That is something more human, low-level human, in fact. No Grandmaster would do that. For example, Carlsen talks on Rogan's podcast that they might promote to a Knight, not a Queen. This way, the opponent engine might be confused and not take the Knight, which would be a blunder. If it was a Queen, it would automatically take it due to the value difference. That is very sneaky and interesting, and the sort of things A.I. systems can do now.

We should define solved as, 'every or draw every game assuming no coding error'. In that case, Chess will be solved very soon by Stockfish, by the seems of it. Likewise, humans are not far behind, but are held back by bodily issues and general mistakes. But, Carlsen already cannot lose a game when he is 100% in body and mind. I think he has the greatest streak in history, 150 top games without a loss? Before that, it was about 100, and before that 95 (Tal), and before that Fischer (21, I think). In the old days, also, Lasker and Morphy went about 100 games without a loss. I think Alekhine and Capa might have, too; cannot confirm.