Too much LCZero

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
hgm
Posts: 24736
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Too much LCZero

Post by hgm » Wed Apr 25, 2018 9:32 am

In flat view more than half the threads on the first page are about other topics than LCZero. Is it that you are usig threadeed view? One of the LCZero threads is getting quite long now; perhaps it should be locked an a new one started.

I guess the main problem is that there isn't much to report beyodn LCZero, an the posting intensity really reflects what the TalChess community is doing at the moment..

Dann Corbit
Posts: 11510
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Too much LCZero

Post by Dann Corbit » Wed Apr 25, 2018 10:07 am

LCZero is the most interesting thing that happened in computer chess in the last 30 years.

Maybe everything but LCZero posts could be moved to some other forum.
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.

jkiliani
Posts: 143
Joined: Wed Jan 17, 2018 12:26 pm

Re: Too much LCZero

Post by jkiliani » Wed Apr 25, 2018 10:50 am

While I can understand developers of other engines getting frustrated by the attention given to LCZero, it's as Dann stated: A paradigm shift is currently taking place in computer chess, triggered by the AlphaZero paper and now proven by Leela. While I wouldn't be surprised at all if developers came up with good ways to combine neural net based engines with Alpha-Beta, I think it's safe to say that in 5-10 years, no serious chess engine will be able to avoid using a NN in some way.

User avatar
hgm
Posts: 24736
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Too much LCZero

Post by hgm » Wed Apr 25, 2018 11:55 am

LCZero is using my cores, not my GPU.
LCZero: Progress and Scaling. Relation to CCRL Elo

Need assistance installing Der Bringer 1.9 in Shredder GUI
Too much LCZero
Yet another borderline position
Why GPUs are well-suited to deep learning
Can't publish in visual studio community 2017 for no .exe
LCZero is annihilating Scorpio MCTS on TCEC right now
Deep Blue takes the back seat
Using GPU for chess calculation
GM Andrew Tang vs Leela Chess Zero

SF spit out a few teeth, but eventually got it right.
Stockfish 9 vs Stockfish 9 - Game with fixed depth = 50
You can run your engine on lichess.org now
Upcoming feature: move animation
Lczero setup help
CDrill 1800
No more ultimaiq SF compiles?
Titan V and LCZero
STS Test Results
Above is the list of subject threads on page one of the general section; I don't consider the situation so imbalanced that it warrants moerator interference.

People are encouraged, however, to share their LCZero stuff in existing threads, an not start new ones for things that are alreay being discussed.

Milos
Posts: 3832
Joined: Wed Nov 25, 2009 12:47 am

Re: Too much LCZero

Post by Milos » Wed Apr 25, 2018 11:55 am

jkiliani wrote:triggered by the AlphaZero paper and now proven by Leela.
I guess you are seeking to steer the pot with statements like that, but that is on par with statements that it is proven that Earth is flat.

Robert Flesher
Posts: 1239
Joined: Tue Aug 18, 2009 1:06 am

Re: Too much LCZero

Post by Robert Flesher » Wed Apr 25, 2018 12:32 pm

Dann Corbit wrote:LCZero is the most interesting thing that happened in computer chess in the last 30 years.

Maybe everything but LCZero posts could be moved to some other forum.


+1

James Constance
Posts: 356
Joined: Wed Mar 08, 2006 7:36 pm
Location: UK

Re: Too much LCZero

Post by James Constance » Wed Apr 25, 2018 12:51 pm

This is an LCZero post :D

peter
Posts: 2132
Joined: Sat Feb 16, 2008 6:38 am
Full name: Peter Martan

Re: Too much LCZero

Post by peter » Wed Apr 25, 2018 1:23 pm

James Constance wrote:This is an LCZero post :D
:)
I don't have any problems with too much postings about LC0, I neither had about A0, about which there was even much less real info and even more hype then.

Yet I think the hopes and dreams and elosions (elo- illusions :)) come from one and the same big old computerchess-self-deception.

It's simply the wish to delegate chess to any kind of engine, "AI" or other automat, so you wouldn't have to bother with the game itself any more as a human being.
:)
Peter.

User avatar
Laskos
Posts: 10620
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: Too much LCZero

Post by Laskos » Wed Apr 25, 2018 1:52 pm

peter wrote:
James Constance wrote:This is an LCZero post :D
:)
I don't have any problems with too much postings about LC0, I neither had about A0, about which there was even much less real info and even more hype then.

Yet I think the hopes and dreams and elosions (elo- illusions :)) come from one and the same big old computerchess-self-deception.

It's simply the wish to delegate chess to any kind of engine, "AI" or other automat, so you wouldn't have to bother with the game itself any more as a human being.
:)
Why you are saying that? In 2-3 months LC0 could rewrite the opening theory, a thing mostly humans developed in 100+ years. Engines up to now were pretty hapless without human input in the evaluation for that. It could still be tactically weak even in 2 months, but one takes Stockfish for checking hidden tactics and LC0 for purely positional play. Have you seen the games played by LC0? It's some sort of strong and romantic, careless human, more than a regular engine.

Albert Silver
Posts: 2917
Joined: Wed Mar 08, 2006 8:57 pm
Location: Rio de Janeiro, Brazil

Re: Too much LCZero

Post by Albert Silver » Wed Apr 25, 2018 2:09 pm

Laskos wrote:It could still be tactically weak even in 2 months, but one takes Stockfish for checking hidden tactics and LC0 for purely positional play. Have you seen the games played by LC0? It's some sort of strong and romantic, careless human, more than a regular engine.
This is truer than you know, and LC0 has completely shattered the paradigm. I am running some tests on my laptop, which has a pretty good CPU and GPU (i7-6600QM with 16GB RAM and GTX980M with 8GB VRAM), and using a very basic tactical suite, compared it to FIreFly on 100 positions. Firefly scored 93/100 and LC0 scored 43/100.

I then started a match using my openings suite (50 positions, 100 games) but stopped it short with Firefly losing 0-10 and no relief in sight.

I have seen engines in the past compensate for some slight tactical inferiority, but the keyword is slight. I have never seen anything like this though. Never.

Firefly is rated 2250 on CCRL, so I am now testing it against Delphil 3.2, which is 2494, and the score is more balanced.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."

Locked