Can Humans Evolve to Beat the Engines

[Peter Berger]

These statements mostly feel kind of ridiculous.

When we talk humans it feels sensible to think of the very top human players. The good thing is: we can read and view what they think in these internet days. Classical chess had become kind of boring conceptwise, until along came Alpha Zero. Magnus evan began to think about getting to 2900 in classical as he realized others were a bit slow in updating their general concept of the game in response at first. Everyone who wanted to stay at the top had to update his game to adapt to the new concepts introduced by the computer.

[towforce]

It will take longer than 7 years for humans to start growing graphics cards in their brain.

After the first Kasparov v Deep Blue game in 1996, I saw several chess experts say, "Not in my lifetime" on RGCC. It seemed as though the challenge of beating the human world champion was beyond the reach of engines in the foreseeable future. In the rematch 12 months later, it happened.

We also know of cases where it has happened in reverse. It might not happen in chess, because chess has a much lower branching factor than Go, making it a more computer friendly game, but in Go, I can give you two examples of humans beating seemingly invincible engines because of weaknesses in the top Go engines:

1. Lee Sedol in 2016: While AlphaGo, developed by Google DeepMind, won the overall match 4-1, Lee Sedol secured a historic victory in the fourth game by strategically exploiting a weakness in the AI's understanding of certain Go patterns. This victory, though not changing the overall outcome, highlighted a potential vulnerability in even the most advanced AI systems and demonstrated the enduring power of human intuition and strategic thinking

2. Kellin Pelrine in 2023: this amateur player was able to beat his machine opponent, KataGo, a top open-source computer Go program which is stronger than AlphaGo (link), by using suggestions from another computer program that was designed by the research firm FAR AI to find a shortcoming in Go-playing AI systems. Pelrine tricked the AI by creating a large “loop” of stones to encircle one of the opponent’s groups while distracting it by also moving stones in other corners of the board. Pelrine told The Financial Times that the Go-playing bot did not notice its vulnerability even when the encirclement was nearly complete. This, to him, would’ve been something easily spotted by a human player (link).

[Uri]

I believe that chess engines still have some small weaknesses compared to the best and most talented human chess players.

Of course if it was a million years from now I'd say otherwise because I believe that in a 1,000,000 years from now engines will become truly unbeatable by any human player. But as of 2025 this is still not the case I think.

For example chess engines are not as good in chess strategy and forming a long-term chess plan as compared to human champions. Also engines don't have the same opening and endgame expertise and technique as the most talented and experienced human chess players.

[OliverBr]

The main question is: Strategy vs Tactics.

In "Tactics" engines are far ahead of humans since long long ago, but in Strategy they are not.

I remember very well my Doctor-Father who was an amazing chess player. He had "only" 2300 ELO, was German team master, he was strategically one of the best players.
I witnessed multiple time that he completely outsmarted any engine (ok, Stockfish didn't exist, it was the times of Glaurung and Crafty) into a wall. He could easily put them into big troubles, but in the end he lost. Why?

Because of tactics. At one time engine found a tactical manoever to take material and the win was lost. But he was the better player.

[jefk]

strategy ?
well -most- engines were weak in strategy but that has been fixed with
the neural nets. MTCS already existed, but nowadays Leelazero is
better than the humans in 'strategy'. As can be seen eg from the leela knight
odds games against humans. SF isn't that bad either.

P.Berger yep this discussion is a bit ridiculous which is why stopped engaging; but after
the -seemingly reasonable comment by mr TF i couldn't resist giving a quick reaction.

Mmr Towforce, chess is not Go. As i understand, not being an expert in (computer)Go,
first of all, there (rare) new improvements in 'strategy' compared to Alphago,
even Alphazero-Go, were invented by other programs/nets (katago), then
in the end one researcher still found a tiny glitch, which he could exploit
against this katago; no big deal; and this will quickly fixed simply by new
release of this katago; or similar.
Wanna find a similar glitch in chess ? well good luck(*)

(you talked about 'steering' the game into positions which the engine does'nt understand;
well apparently you have not played (much) correspondence chess; almost all
relevant opening strategies have been analyzed nowadays; want to explore other
routes, eg. after 1.h3 or so, well good luck again but look at the chinese database
also there such positions are there to quite some depth and they lead to nothing
(ofcourse one can say that this is only an opening base, blablabla, well the
game starts with the opening, some -known- opening have reached the endgame
and once your reached the endgame, the result is known. Can you 'steer' the
game via eg h3 into a winning position. No you can't.

PS thousand centuries before an engine won't lose against a human ? Lol, at current
strength levels with sufficient little time/move the top engines are already invincible,
whether it's against a human, another engine, or G itself simply because chess is a draw
(with 'perfect play' ie not making mistakes thus staying in the drawing margin);
apparently i'm ahead of my time, but 100000 years ? Whew, that's a surprise

PS although i'm not always in line with mathman syzygy, what he meant with a
graphics (GPU or more modern) card is -i suppose- the processing (in fact calculations) it
can do with the neural nets; which are much more exact/accurate than humans what can;
it's not about vision (accuracy) whereby human biology indeed has amazing capacities,
regarding pixels, colors, stereo (depth), etc. But you can make a picture of a chess
board in zillions of megapixels, if you don't know how to interpret that (search)
this means nothing. In other words, it's about interpreting the (vision) results;
we have in our brain certain capacities, eg. recognizing faces, and (if your not autistic)
emotions, but doing this with chess positions is rare unless you're someone as Magnus;
and he 'trained' a lot you know, observing many positions, in his youth and thereafter.
But not at the level of the modern nnue machines, like it or not. Ordinary humans
will anyway -probably- not 'evolve' to such capacities unless you're going the
cyborg route (not going to dwell in such topics about transhumanism).

(*) as you may know i offered here ten grand to the person who can come
up with a winning line for White in chess which cannot be refuted.

[syzygy]

I believe that chess engines still have some small weaknesses compared to the best and most talented human chess players.

Of course if it was a million years from now I'd say otherwise because I believe that in a 1,000,000 years from now engines will become truly unbeatable by any human player. But as of 2025 this is still not the case I think.

Are you sure you are living in 2025?
Your posts here, including your last one, suggest you are somehow stuck in 2002.
I am not going to speculate, but let's say you are not up to date.

[jefk]

syzygy, mr RdM,\
this guy is not Uri B, this 'uri' (don't know him) once talked about dementia in
the kindergarten, so there are various sorts of persons; be polite (basic respect);
as for towforce, i can -possibly-0 discuss with him (if he would listen).
jef
PS i'm still working on my 'proof' (yeah the ultraweak solution, now i cannot
post in the KG anymore on Ras' advice, here's the latest ; 1) i have a
draft article either for research gate or authorea, but not here), maybe i should
use a system as Lean which you probably know (lol); i do have respect for math(*) ,
2) i made conceptual progress again regarding4-in-a-row (remember my
expanding trees if you noticed) for other configurations eg when the first mover
is losing by force, and then again, you can see this quite soon in the beginning
when you make the trees (losing/winning/draw) thus zugzwang (remember
my statements about Zermelo; if there's not winning strategy it is a draw)
(*) actually i was asked in secondary school if i wanted to to math olympics
but i politely declined (not having the faintest idea what this was and i was lazy
then except for some table tennis instead of homework believe it or not); they also
didn't make if very attractive, in Russian school i would probably have been pulled out
and forced to join some training program (before such olympics); my teacher also was
not so helpful either, typical Dutch school system, they help the average, and the rest can
rot (or can go play chess (or do computer chess?) which i only started later in my life,
at least at a more serious level;
PS2 new update of Bookbuilder coming soon, now also with SF engine analysis (!) at sourceforge.net
(i will announce that later more properly some last cosmetics to fix and then we have something
PS3 math is not politics, i know that, so why is the (academic) math world
(apparently) tough ? logical restrictions, truth, etc, yeah i understand, but
respect please (look at eg a guy as T.Tao); in physics there also have been intellectual
disputes (eg about string theory black holes) yet when i met some (top)
theoretical physicist in real life the actually seemed (in my perspective were)
nice people; even in chess at higher levels the GM's usually are more civilized
(although there are some arrogant blokes as well, and the hierarchy in some
old fashioned chess clubs continue to amaze me until today (lol))
PS4 want to see my draft paper (about the ultraweak solution for chess)
no big deal; we can figure this out with the polder model. if you agree i'll
probably find one of your mail adresses somewhere and might mail it to you
Peace
(difference between eg a guy as Trump and me is that i post sometimes
when i had a beer (they should allow us on this forum to edit our posts
lateron as well) whereas mr Donnie and the great us leader never drinks
https://trumpstruth.org/
yep this is off topic i know but then it imo are interesting times ;
chess is a draw; would this also be between good and 'evil'?
No further comment(s)

[Werewolf]

The good news: uncovering new chess knowledge (deeper patterns in chess) would enable humans to easily beat today's chess computers.

Obviously this is false.
Chess is a concrete game with a search tree. Of course a human cannot match a machine which looks at millions of times more nodes, anymore than a runner can say "I'm sure in the future humans will run as fast as Ferraris".

[OliverBr]

be polite (basic respect);
as for towforce

towforce, for aka Graham Bank, is one of the most politest and respectful persons I ever had the joy to make contact.

[OliverBr]

[Of course a human cannot match a machine which looks at millions of times more nodes

I can not complete concur. I know a very good chess player (German Team master) who regularly played big engines "on the wall". because he was strategically far superior.
I saw it live (he was my doctor father) and I was so impressed each time. Engines in that time (1997-2002) were far inforior to him, even they could look at millions of time more nodes. Their pruning was inferior in every aspect. They strategical knowledge also.

Only thing that he "oversaw" tactical things, which robbed him most time of the deserved win.
So, no, I would not say you can easily say "Of course a human cannot match". He could.

Maybe to add: He did not use anti-engine tactics, the didn't even know such. He didn't seem to think about any move, he "felt" them. Once we tried together against a machine and he said to me "push that pawn". Now there were two pawns who could have been pushed, so I pushed one. And then he "Oh, no. Not this one. You spoiled it."