How far away are we from deep learning Stockfish, Komodo,...

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Hai
Posts: 455
Joined: Sun Aug 04, 2013 11:19 am

How far away are we from deep learning Stockfish, Komodo,...

Post by Hai » Fri May 19, 2017 10:55 am

3, 6, 9, 12 months?

Does it make sense to have a
Stockfish
Asmfish
Pedantfish
...
Deep learning Stockfish
Deep learning Komodo
Deep learning Houdini
...

AlvaroBegue
Posts: 920
Joined: Tue Mar 09, 2010 2:46 pm
Location: New York
Full name: Álvaro Begué (RuyDos)

Re: How far away are we from deep learning Stockfish, Komodo

Post by AlvaroBegue » Fri May 19, 2017 11:13 am

I am confused. Is it inevitable that deep learning will be useful for top engines? What's the evidence?

CheckersGuy
Posts: 273
Joined: Wed Aug 24, 2016 7:49 pm

Re: How far away are we from deep learning Stockfish, Komodo

Post by CheckersGuy » Fri May 19, 2017 11:29 am

In my opnion we might not see deep learning at all if we still have very consistent elo improvements year after year. Using neural networks is still pretty slow and probably not quite likely to be seen in chess engines using the current hardware generation.

Hai
Posts: 455
Joined: Sun Aug 04, 2013 11:19 am

Re: How far away are we from deep learning Stockfish, Komodo

Post by Hai » Fri May 19, 2017 2:59 pm

AlvaroBegue wrote:I am confused. Is it inevitable that deep learning will be useful for top engines? What's the evidence?
1.Who is stronger: Stockfish 8 vs Stockfish 8 + deep learning.
What is your opinion?

2.Take for example infinitychess tournaments:
To get the best results, Stockfish + human(=deep learning) was always better than only Stockfish.

AlvaroBegue
Posts: 920
Joined: Tue Mar 09, 2010 2:46 pm
Location: New York
Full name: Álvaro Begué (RuyDos)

Re: How far away are we from deep learning Stockfish, Komodo

Post by AlvaroBegue » Fri May 19, 2017 8:06 pm

Hai wrote:
AlvaroBegue wrote:I am confused. Is it inevitable that deep learning will be useful for top engines? What's the evidence?
1.Who is stronger: Stockfish 8 vs Stockfish 8 + deep learning.
What is your opinion?
I can't form an opinion because I don't know what "Stockfish 8 + deep learning" even means. You could use a CNN as part of the evaluation function, and you might make it better, but you would also slow it down by an order of magnitude, so the resulting player would be worse. If you can think of some other way of using deep learning here, please let us know so we all know what we are talking about.
2.Take for example infinitychess tournaments:
To get the best results, Stockfish + human(=deep learning) was always better than only Stockfish.
Two objections: 1) The `human=deep learning' part of that equation is just absurd. 2) I don't think we have enough statistical evidence that the human adds anything of value.

noobpwnftw
Posts: 360
Joined: Sun Nov 08, 2015 10:10 pm

Re: How far away are we from deep learning Stockfish, Komodo

Post by noobpwnftw » Fri May 19, 2017 8:36 pm

After some garage experiments, I think NN-based eval for chess is pretty much a joke even compared to current PST+MAT eval alone.

At most, it might be good for move pruning and sorting, prove me wrong.

zenpawn
Posts: 296
Joined: Sat Aug 06, 2016 6:31 pm
Location: United States

Re: How far away are we from deep learning Stockfish, Komodo

Post by zenpawn » Fri May 19, 2017 8:59 pm

AlvaroBegue wrote:2) I don't think we have enough statistical evidence that the human adds anything of value.
That said, I remember fondly this result as I knew these guys from local tournaments. :)

http://en.chessbase.com/post/dark-horse ... tournament

ZirconiumX
Posts: 1327
Joined: Sun Jul 17, 2011 9:14 am

Re: How far away are we from deep learning Stockfish, Komodo

Post by ZirconiumX » Fri May 19, 2017 9:41 pm

noobpwnftw wrote:After some garage experiments, I think NN-based eval for chess is pretty much a joke even compared to current PST+MAT eval alone.

At most, it might be good for move pruning and sorting, prove me wrong.
Gladly.

I don't think the answer to "Are neural networks feasible for computer chess?" is "No", I think it's "Not yet".

AlphaGo required specialised hardware to win at Go, remember.
Some believe in the almighty dollar.

I believe in the almighty printf statement.

Milos
Posts: 3387
Joined: Wed Nov 25, 2009 12:47 am

Re: How far away are we from deep learning Stockfish, Komodo

Post by Milos » Fri May 19, 2017 11:38 pm

ZirconiumX wrote:
noobpwnftw wrote:After some garage experiments, I think NN-based eval for chess is pretty much a joke even compared to current PST+MAT eval alone.

At most, it might be good for move pruning and sorting, prove me wrong.
Gladly.

I don't think the answer to "Are neural networks feasible for computer chess?" is "No", I think it's "Not yet".

AlphaGo required specialised hardware to win at Go, remember.
Answer is No and Never. Bojun is 100% correct.
If you took Giraffe and replaced its eval with SFs eval it would gain few hundred Elo easily.
Similarity between Go and chess is that they are both played at the board. That's exactly where any similarity ends.
So mentioning Go in context of chess is absolutely pointless.

Milos
Posts: 3387
Joined: Wed Nov 25, 2009 12:47 am

Re: How far away are we from deep learning Stockfish, Komodo

Post by Milos » Fri May 19, 2017 11:43 pm

Hai wrote:
AlvaroBegue wrote:I am confused. Is it inevitable that deep learning will be useful for top engines? What's the evidence?
1.Who is stronger: Stockfish 8 vs Stockfish 8 + deep learning.
What is your opinion?

2.Take for example infinitychess tournaments:
To get the best results, Stockfish + human(=deep learning) was always better than only Stockfish.
Are you for real?
This is programming forum for god sake.
Human=deep learning????
Humans have much less similarity with deep learning than spiders or ants or bees.
Unfortunately bee+SF is still stronger than deep learning+SF because bee would not make Stockfish weaker while deep learning certainly would.

Post Reply