AdminX wrote:
Thank you for the clarification Guenther.
Interesting how it attracted stupid dumbhead cheaters ;-)
The first game I watched already I was very soon sure that it was a cheater
and his rapid games revealed even more cheating. (later he was marked)
Now I see that already three cheaters were caught alone in playing this
(weaker version) of LeelaChess. How dumb can you be to cheat in front
of dozens and up to hundred observers against a computer program??
Also experienced lately that around 20% of my opponents were cheaters.
(playing in the range of 2050-2300 at lichess)
It is stalled. Self-playing Elo results mean nothing in real life. For actual performance check Kai's results.
It is not stalled. The CCLS gauntlet clearly still shows progress, it was actually just upgraded to a stronger opponent selection as a result. And by the way, once there really IS a stall, we'll simply upgrade to the next larger network architecture (most likely 192x15 at this point).
GMs play her because she is an AI and different from standard chess engines it is good to see her progress versus various players. In due time she will beat Stockfish. I would like to see them put a HUMAN ELO on the graph and it does seem her progress is slow.GM Tang is a good sport!
FWCC wrote:GMs play her because she is an AI and different from standard chess engines it is good to see her progress versus various players. In due time she will beat Stockfish. I would like to see them put a HUMAN ELO on the graph and it does seem her progress is slow.GM Tang is a good sport!
Her progress is slow? Compared to what exactly? Sure it's slow compared to when she was a very weak player, but compared to basically any other chess engine that's improved by code tweaking? Really?
LCzero and Alphazero have a lot to lose by playing chess against anti-machine specialists without titles and elo Fide. Alphazero and LCzero are very busy taking care of and making up their image, they do exactly what DeepBluee did, they avoid at all costs that a heretic and renegade player ridicules their computers. The explanation is simple: The myth would come to an end. In the modern world, it is not necessary to have the titles or titles to face the best computers in the world.
I am thinking chess is in a coin.Human beings for ever playing in one face.Now I am playing in the other face:"Antichess". Computers are as a fortres where owner forgot to close a little door behind. You must enter across this door.Forget the front.
Father wrote:LCzero and Alphazero have a lot to lose by playing chess against anti-machine specialists without titles and elo Fide. Alphazero and LCzero are very busy taking care of and making up their image, they do exactly what DeepBluee did, they avoid at all costs that a heretic and renegade player ridicules their computers. The explanation is simple: The myth would come to an end. In the modern world, it is not necessary to have the titles or titles to face the best computers in the world.
Do you understand your own rants?
Actually it seems you fear to play not vice versa, we told you how to play vs. LCZero at lichess, but I guess your excuse is now you are not able to create an account there?
The surprise for me is that a NN that learned through self-play could after 4M self play games dominate an IM (lovik/lovlas) and a GM (Tang/penguin).
It also plays attractive and interesting, if flawed by SF standards, chess.
BTW the version was ID125, which IIRC had the promotion bug that was introduced in the code after ID123 but not yet heavily affecting the trained net as far as I understand.
The current net is much stronger, but still relatively weak tactically and at endgames, where it cannot compete with egtbs of course.
Nevertheless, think about what has been done in such a short space of time. No one told this engine how to play.
The progress is exponential.
LCZero is going to be really, really good in a few months.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
I agree about GM Tang being a good sport. Indeed, it was nice to watch him obviously having fun while trying to get a win off Leela.
While watching Tang play without even coming close to losing his cool and whose ego appeared to remain intact while losing multiple games, I couldn't help but think about the Go masters whose disappointment in their own losses against AlphaGo/AGZero was clearly crushing their egos and affecting them emotionally.
I enjoy watching the documentaries and live action as top players test themselves against the new NN-inspired software. Me, I'm content to test myself against my capable eleven-year old grandson, who checkmated me in 31 moves the last time we played.