Page 2 of 7

Re: Alpha subZero - NN for CPU

Posted: Wed Jan 15, 2020 12:18 am
by carldaman
A few folks over at chess2u even suspect some possible malware. Ugh! :x

Re: Alpha subZero - NN for CPU

Posted: Thu Jan 16, 2020 4:26 pm
by dkappe
I’ve requested the source code over at chess2u from the “developer” per the terms of the GPL.

With all the antivirus warning signs and client.exe (which should be a golang program used only for self-play training but apparently is required), I wouldn’t touch this with a ten foot pole.

Re: Alpha subZero - NN for CPU

Posted: Fri Jan 17, 2020 9:40 am
by marsell
What will you do if the source code is not available? Are you going to knock on his door with a gun loaded? 8-)

Re: Alpha subZero - NN for CPU

Posted: Fri Jan 17, 2020 9:56 am
by Damir
I would not worry about what dkappe want… after all when he put source of a certain engine on GitHub it is only there to collect dust…. same way he did with ThothFish... :roll: :roll:

Re: Alpha subZero - NN for CPU

Posted: Fri Jan 17, 2020 9:57 am
by Damir
Giving him the source would be beyond foolish….

Re: Alpha subZero - NN for CPU

Posted: Fri Jan 17, 2020 3:12 pm
by dkappe
marsell wrote:
Fri Jan 17, 2020 9:40 am
What will you do if the source code is not available? Are you going to knock on his door with a gun loaded? 8-)
Yikes! How do you go from a nice, peaceful request to such a bloodthirsty concept?

Re: Alpha subZero - NN for CPU

Posted: Sun Jan 19, 2020 11:51 pm
by cucumber
supersharp77 wrote:
Mon Jan 13, 2020 9:16 pm
cucumber wrote:
Sun Jan 12, 2020 10:44 pm
I can confirm that it's a modified Stockfish (or Brainfish...) that reports an artificially lowered depth.
You sure its really SF? The size seems small 468k for a Stockfish compile........asmFish Perhaps?...Thx AR :) :wink:
Yes, it's packed with something like UPX to bring the file size down. It just divides nodesCounted by a large number (1000 if I remember correctly?) in the first binary he posted.

Re: Alpha subZero - NN for CPU

Posted: Tue Jan 21, 2020 9:57 pm
by Scacchista1977
there are many doubts about this engine, is possible that it is just a simple clone ?!
In addition to the countless evidences published at chess2U.com forum, there is this test which shows an amazing result:

Image

games:
http://www.mediafire.com/file/kel3doh6s ... es.7z/file

is it reliable in your opinion?

meanwhile here there is identical performance, for now ...
http://infinitybook.pl/stream.html

Re: Alpha subZero - NN for CPU

Posted: Wed Jan 22, 2020 9:58 am
by Guenther
Scacchista1977 wrote:
Tue Jan 21, 2020 9:57 pm
there are many doubts about this engine, is possible that it is just a simple clone ?!
But this is long proven... the first version obviously was a plain copy of SF dev from December,
with a manipulated node count (divided by x).

Now it morphes into something else, probably cloning LC0 with other obfuscations.
Scacchista1977 wrote:
Tue Jan 21, 2020 9:57 pm
In addition to the countless evidences published at chess2U.com forum, there is this test which shows an amazing result:

Image

games:
http://www.mediafire.com/file/kel3doh6s ... es.7z/file

is it reliable in your opinion?

meanwhile here there is identical performance, for now ...
http://infinitybook.pl/stream.html
Tests w/o hardware etc. specifications are meaningless, if it is meanwhile a cloned LC0 playing on highend GPUs
it is obvious it will win against any SF on 10 cores by some margin.
(also one of those tests is from the alleged 'author' and the other from one of his 'buddies')

Some people have too much time to waste and some believe anything they read in the net.

Re: Alpha subZero - NN for CPU

Posted: Wed Jan 22, 2020 3:52 pm
by Damir