Chess will never be solved, here's why

The question is though, at those levels, is it possible for a chess program to actually be 1,000x better at computing, but still not be able to beat the lower rated engine 1000x more frequently. Just because the lower engine still isn't just going to lose easily. If it manages to find a defensive fortress for example, no matter how much further the higher level program can calculate, doesn't mean it it will be able to breakthrough to win. It may not require as high a level to play perfectly defensively, whereas the level of the higher one won't matter if there's no way to breakthrough. Ratings are affected by game results, so they may end up evening out eventually in rating even if one is 1000x the ability of the other, due to the fact that it can't win enough. To put it another way, the 4000 rated engine might be beating the 3000 rated engine 10x more frequently than it beats the 3400 rated engine, which is expected, but yet struggles to win 90%+ of games against the 3400 rated one itself.

I know the various directions AI is currently heading, and it's not looking good for solving chess anytime before our sun becomes a red giant. What is your belief in AI's quantum leap based on? Even storing the calculations along the way would take more matter than exists in our solar system, so that's maybe the first hurdle to jump over for you.

You're positing a forced draw. Seems kind of redundant on this thread . See the post above yours for a clear indicator that we are not demonstrably approaching some threshold where engine play will fall prey to diminishing returns.

In fact I don't even think any rating system can be the same for computers as it is for humans. Computer ability is more speed, nunbercrunching, and brute force calculating, whereas humans can use more logic, strategic reasoning, and pattern recognition. This is why many studies, puzzles, and compositions exist that humans can solve and even understand, but computers can't. The two types of players work too differently for the rating system to be consistent between them. Using a human rating system and assigning it to computers may not work. It was meant for human players. And given that the rating system can lose meaning at the extreme ends of the scale, that only amplifies this problem. We aren't as surprised when a 300 beats a 700, or when a 2500 beats a 2850, as when a 1200 beats a 1600 for example. The lower end of the scale can make so many blunders and random moves that it's too unpredictable, and a 2500 is still "really good", so it is conceivable that they can find some deep line that beats a 2800 in a very close game. But a 1200 beating a 1600-1700, we'll think it was some random big mistake at 1 move or a fluke. With computers over 3500, the numbers start to lose meaning as calculating a trillion positions per second isn't going to necessarily overpower another that's calculating "only" 300 billion per second. Unless there was some crazy line that the 1T/s would see that the 300B/s one wouldn't for some reason.

guys calm down

what r all these long explanations

This premise doesn't work. The player themselves doesn't matter at all, only their win rate, be they engine or human. Have you ever watched Moneyball? Doesn't matter how the player gets on base, only that they do get on base at a consistent and predictable rate.

Read them and find out.

CALM DOWN PEOPLES

dont think i bother reading all of that

Or, just stop reading and find something else to do that you can emotionally handle.

Wise decision, for you.

You're now using the Tygxc method of trying to build a bridge from both ends, mud on one side and steel girders on the other...it's not going hold any weight even if you somehow manage to connect them.

Ah but the win-rate is what ultimately determines people's rating. They aren't arbitrarily assigned. A new tournament player plays 20 or 30 games and they estimate it based on the number of 1000s, 1200, and 1500s..etc, that they won, drew, and lost against. By definition it's comparative. But with computers, how do we know stockfish is 3700? If it beat Magnus 99% of the time, how do we know whether it is 3700 or 37,000? Because no rating difference guarantees a 100% win-rate. How do we know that it's not really only 3200 but the difference is humans can make mistakes, computers can't. And those human mistakes are reflected in the rating. Since a computer can't blunder or forget about a threat on the board or "have a bad day". The fact that 2700s still make mistakes against 2500s in top tournaments might be what keeps them from not soaring into the 3000s (even if their real average ability is over 3000) But a computer will never make those mistakes, so it will never drop from that. So if we don't even know whether stockfish is 3200, 3,700, 4,700, or 470,000, how can we gauge the future supercomputers that beat it?

We gauge them by the ratings pools they play in, like any human player.

I have made the same point, that engines are largely in an insular ratings pool at this point, but it doesn't really matter ultimately, the engine pool was created by matching against humans, and engines continued to play from there and the pool relative to the engines playing in it is valid. The pools will diverge only so long as human GMs are too cowardly to play engines without getting odds . If you were to force a re-merge of the pools (say by allowing engines to compete in FIDE events and just not allowing them to win the title or prizes) the rating pools would stabilize fairly quickly.

EE you made another mistake again.
You made a strawman against Shannon.
Shannon never claimed that games couldn't go on past 40 moves.
Why did you try to dream up that he had?
You started your post with that false strawman/false premise - 
so I again didn't read the rest.
When are you going to start with 'valid'?
Six years from now?
----------------------
EE the translation of your bad premise is that you still don't get it about 'lower bound'.
Look up lower bounds and upper bounds in math so you have some kind of notion what they are.
Don't be afraid to use AI to do so.

Yes, but there's a difference that still applies. Once over "3500" in the computer realm, the ability difference may manifest in terms of not losing rather than winning. It's hard to explain without sounding like I'm just babbling/repeating myself, but a 3000 rated engine losing constantly against a 4000-5000 one, vs a 3500 engine not losing against a 4000-5000, possibly drawing it 80% of the time, because it has reached a level where it won't make any positional mistakes or miss even the deepest lines that the other computer can find. The difference then becomes crossing a threshold where the leap from one rating to another, makes it nearly impossible for the weaker computer to lose, therefore making it nearly impossible for the stronger one to win. And not losing = draws. But if as a result the 3500 gets 100% draws against the 4000-5000, their ratings should be the same. But they also shouldn't be because the 4000 one still beats the 3000 one much more frequently than the 3500 one beats that same 3000 one. It becomes a paradox. The 4000 one will constantly get points taken away from it by the 3500 one just by drawing, even if there are no winning lines to search for that the 3500 wouldn't still see. Their ratings would even out despite the ability difference just because of the fact that draws suck rating points out of the better player. Maybe draws shouldn't be treated the same in computer matches, I don't know.

The stronger computer can search trillions of more lines even if those extra lines can't accomplish anything better than that same draw. That's the best way to explain what I mean.

then idea: don't ask 'what is this?'
But you can. Now please post under this so I can post something else under you without having two consecutive posts.

I know the 40 is supposed to represent an average, but quite frankly most even lower level tournament games regularly go beyond that. I could see 50 or 60, but 40 feels short on all levels. When I used to play in the sub-1500 and sub-1800 sections of tournaments, my games were into 60 moves frequently. The 50 move rule is also similarly short-sighted compared to modern endgame research. This is why I don't even like how they setup the time controls in the WC match, they tag on 1 hour or 15 mins extra to random exact moves like #40 and #60 as if they actually mean something in the game. Like something special always happens at these moves that's consistent between games, which justifies the time boost, which couldn't be more false.

EE I didn't say 'average'. 
Dio probably knows for sure about the 40 moves.
Exactly what Shannon was doing with the 40 moves.
He wasn't 'claiming an average'. Maybe he assigned an average.
Or maybe he 'arbitrated a maximum' for the purpose of establishing a lower bound.
Don't say you know.
You couldn't because you don't get it about upper and lower bounds.
EE now trying to discredit Shannon in order to defend his false premise of 'you can't limit .... "I've seen EE attempt that many times.
Its called 'doubling down' on mistakes already made.
EE do you get it yet that professional mathematicians know 1000 times as much about math as you'll ever know?