Fruits? What are you talking about. Maia isn't about averaging all players. Read the paper
It seems like I'm not cut out for chess.
Sort:
The paper is one paragraph, and it says that Maia was trained on human games so that it plays more like a human. This means that Maia would learn the average human's strategy, and the average human doesn't play like an individual human.
No, it's not one paragraph.
you may say it's much longer than a paragraph, I say it's just a really long paragraph
If you mean this stuff included 
look at that scale, it barely goes above 50%
A human doesn't make the human move 50% of the time, it makes the human move either a lot more or a lot less.
RandomChessPlayer62 wrote:
you may say it's much longer than a paragraph, I say it's just a really long paragraph
Lol. Read the paper, bro. Look up "Access paper - View PDF"
This 11 page paper contains about 2 paragraphs worth of relevant information, and nothing in the paper magically shows that Maia isn't just an average amalgamation of human strategy that ends up not really playing like a human.
So you've spent 18 minutes reading your first scientific paper ever and already making some conclusions? It's not that simple. Just because their training evaluation system can see 50% match, doesn't mean you will distinguish Maia from human in 50% of the cases. I doubt you will do even 5% detection. How many chess engines did you write? I wrote some. One of my recent engines was written very quickly with lot of bugs but still convinced a local chess engine expert, he was absolutely sure that's a human play. So if your point is humans are not replaceable, maybe they aren't but can you notice it and does it matter in chess development? You actually don't even need an engine to develop your chess skill. Just self-play is enough.
Give me 100 games, half between Maia and a human, half between a human and a human.
Oct 19, 2024
0
#72
Ok guys lets not waste our time arguing against a randumb 360.
I'm gonna make this quick.
Chess.com has very reliable anti cheat mechanisms.
Bots play different than humans. For example higher rated bots alternate between long streaks of stockfish moves and a few really bad blunders that no one at their level would make. They alternate being able to find genius moves, and being unable to find easier moves.
Now self playing is one thing, but you have the disadvantage of knowing the other sides plans, then there's the bias of wanting one side to win.
All in all the best ways of improving is to alternate between studying, playing, and reviewing your games.
a smart player can cheat on this site very easily
lol his infinite supply of random surveys and papers didn't include a human-bot distinguishing test
RandomChessPlayer62 wrote:
lol his infinite supply of random surveys and papers didn't include a human-bot distinguishing test
Because humans can't do that lol. Otherwise anticheating would be SOOO easy.
falcon39 wrote:
Ok guys lets not waste our time arguing against a randumb 360.
I'm gonna make this quick.
Chess.com has very reliable anti cheat mechanisms.
Bots play different than humans. For example higher rated bots alternate between long streaks of stockfish moves and a few really bad blunders that no one at their level would make. They alternate being able to find genius moves, and being unable to find easier moves.
Now self playing is one thing, but you have the disadvantage of knowing the other sides plans, then there's the bias of wanting one side to win.
All in all the best ways of improving is to alternate between studying, playing, and reviewing your games.
Randumb 360? Reported. You shouldn't use insults. Your view on bots is just a repetition of common myths, that's not how bots work. It's actually the opposite, human play often looks like top stockfish line with occasional mistakes and severe blunders. While downskilled Stockfish just picks moves that are slightly worse, avoiding blunders and mistakes.
In my recent game against Stockfish that was limited to 1600 Elo, in moves 10..20 (after the opening)
I made
three top-1 stockfish moves,
three top-2 stockfish moves,
only one inaccuracy,
three mistakes and
one blunder,
other moves were of mixed accuracy.
Stockfish-1600 in same 10..20 moves made
only two top-1 stockfish moves,
only one top-2 stockfish move,
six inaccuracies,
only one mistake,
no blunders
My play looks like a bot play if your description of bot play was correct. Limited Elo Stockfish just plays some not very accurate moves but without too much mistakes and no blunders.
Now, stockfish is a bad example. There are no humanization features in Stockfish. Komodo*, Maia, The King known for humanization. Komodo and The King offer personalities, something Maia doesn't have (random chess player's concern about averageness).
Here on chess.com Komodo is used, not Stockfish and you can enjoy personalities although you don't have access to all settings (you can download Komodo to your PC and set it up as you wish, try different configs). But chess.com bots run using your PC/smartphone resources, not chess.com servers (chess.com is saving a lot in such way) and their behavior is not reliable because it depends on available resources of your PC/smartphone.
* by Komodo I mean Komodo Dragon engine, sometimes named just Dragon
Oct 19, 2024
0
#79
Pretty average 1700 game game, you are underated, I will give you that.
RandomChessPlayer62, Maia is an AI bot trained on lichess games. Your assessment of it as human like or not human like is subjective. While their (Maia authors) research paper has some actual numbers on how it matches human play.
Anyways, that's a weak bot and you will not learn too much by playing against it.
If I made a food that's the average fruit, would it taste like a fruit? Apples, grapes, bananas, cucumbers, mangos, oranges, etc. All combined. It probably wouldn't taste like a fruit, I could have data that matches every bit of flavour with every fruit, but it wouldn't taste like a fruit.