recent article on alphazero ... 12/11/2017 ...

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

Moderators: hgm, Rebel, chrisw

duncan
Posts: 12038
Joined: Mon Jul 07, 2008 10:50 pm

Re: recent article on alphazero ... 12/11/2017 ...

Post by duncan »

hgm wrote:It seems a complete bullshit paper...
perhaps you could go into a bit more detail
User avatar
hgm
Posts: 28134
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: recent article on alphazero ... 12/11/2017 ...

Post by hgm »

Well, for instance the idea that it would make the result less scientific because others cannot reproduce it, because they lack the equipment to do so, is utterly ridiculous, and shows that the guy has no clue as to how science works. It is well known how to neural nets work, and others can run it on their own hardware, and build the necessary hardware if their's is not powerful enough. That they likely don't want to spend the money is never an issue. Cutting-edge scientific discoveries are always done by the people with the best equipment. There also is no other accelerator like CERN's Large Hadron Collider (yet).
Dariusz Orzechowski
Posts: 44
Joined: Thu May 02, 2013 5:23 pm

Re: recent article on alphazero ... 12/11/2017 ...

Post by Dariusz Orzechowski »

hgm wrote:Well, for instance the idea that it would make the result less scientific because others cannot reproduce it, because they lack the equipment to do so, is utterly ridiculous, and shows that the guy has no clue as to how science works. It is well known how to neural nets work, and others can run it on their own hardware, and build the necessary hardware if their's is not powerful enough. That they likely don't want to spend the money is never an issue. Cutting-edge scientific discoveries are always done by the people with the best equipment. There also is no other accelerator like CERN's Large Hadron Collider (yet).
You will be right as soon as Google publishes their raw data (which is not going to happen). CERN even published few hundred terabytes of their raw data on the internet last year.
Leo
Posts: 1092
Joined: Fri Sep 16, 2016 6:55 pm
Location: USA/Minnesota
Full name: Leo Anger

Re: recent article on alphazero ... 12/11/2017 ...

Post by Leo »

I agree with this paper. Way to many flaws in the way the contest was run. I agree with HGs comment about the lack of equipment.
Last edited by Leo on Thu Dec 14, 2017 6:12 pm, edited 2 times in total.
Advanced Micro Devices fan.
Leo
Posts: 1092
Joined: Fri Sep 16, 2016 6:55 pm
Location: USA/Minnesota
Full name: Leo Anger

Re: recent article on alphazero ... 12/11/2017 ...

Post by Leo »

Rodolfo Leoni wrote:
shrapnel wrote:Unfortunately, since Alphazero has not been commercialized nor has more information been released, the voices of its detractors will grow only stronger.
Sad, because computer chess players were given a glimpse of the Future ( A Ferrari ! ) and now we are forced to play with obsolete chess engines ( Wooden Bullock Carts !).
Being not a detractor, I really think hardware is a breakthrough. I'm not sure about AlphaZero, tough. It could become like one of the old videogames in our C64 home computers, if compared to an engine by top programmers (when/if hardware will be available).

If we think at self-learning procedure and philosophy, Romichess already performed such a task more than 10 years ago. Give Romi features to a top engine, let it train both vs himself and a group of top opponents for enough games, then you'll get the same effect. With all the limits of conventional technology, of course.

The strenght difference of AlphaZero could depend only because of hardware, IMO. But I'll be very curious to see if Deep Mind decides to let it participate to WCCC, to legitimate its superiority. :wink:
+1 I would be really surprised if they compete in any sanctioned tournament.
Advanced Micro Devices fan.
Damir
Posts: 2823
Joined: Mon Feb 11, 2008 3:53 pm
Location: Denmark
Full name: Damir Desevac

Re: recent article on alphazero ... 12/11/2017 ...

Post by Damir »

Moderators please delete multiple AlphaZero threads. It makes no sense to talk about the engine that runs on a hardware nobody is ever gonna have. :) :)
FWCC
Posts: 117
Joined: Wed Aug 22, 2007 4:39 pm

Re: recent article on alphazero ... 12/11/2017 ...

Post by FWCC »

I was wondering if Alpha Zero could give a GM Queen odds,but in all likelihood that would be too much.Perhaps a two minor piece handicap.
Uri Blass
Posts: 10469
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: recent article on alphazero ... 12/11/2017 ...

Post by Uri Blass »

FWCC wrote:I was wondering if Alpha Zero could give a GM Queen odds,but in all likelihood that would be too much.Perhaps a two minor piece handicap.
I am sure that it cannot give a GM queen odds or two minors pieces.
I feel cerain that even I can beat every player in these conditions.

People who think that it is possible to beat GM's with 2 minors pieces handicap probably are not strong chess players.

Even one minor piece handicap is probably too much against GM's.

Uri
User avatar
hgm
Posts: 28134
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: recent article on alphazero ... 12/11/2017 ...

Post by hgm »

Dariusz Orzechowski wrote:You will be right as soon as Google publishes their raw data (which is not going to happen). CERN even published few hundred terabytes of their raw data on the internet last year.
It is not customary in sciece to publish raw data. I never did it, and no one ever asked for it. People that want to reproduce the experiment should collect their own raw data.

The only reason I can see that would make it necessary to publish raw data is when the data analysis could be controversial or non-standard. I can very well imagine that this is the case in some of the CERN experiments, where discovery of a new particle often is claimed based on extremely scanty evidence of just a hand-ful of 'events'. It does't seem the case for the AlphaZero experiments.
Rein Halbersma
Posts: 749
Joined: Tue May 22, 2007 11:13 am

Re: recent article on alphazero ... 12/11/2017 ...

Post by Rein Halbersma »

hgm wrote:
Dariusz Orzechowski wrote:You will be right as soon as Google publishes their raw data (which is not going to happen). CERN even published few hundred terabytes of their raw data on the internet last year.
It is not customary in sciece to publish raw data. I never did it, and no one ever asked for it. People that want to reproduce the experiment should collect their own raw data.
Depends on the field. In economics (yeah, it's not physics, so don't go all Rutherford on me, I'm a physicists by training, economist by profession :) ), the top journals now have strict replication policies, requiring authors to submit all data, programs and documentation sufficient to replicate the findings in a paper.

https://www.aeaweb.org/journals/policie ... ity-policy

In any case, the AlphaGo-Zero paper in Nature (also their earlier Nature paper) were proper papers with enough details to be able to replicate the experiments. Hence, open source projects such as Leela-Zero.

The last paper on AlphaZero, however, is very much an early draft, with just enough results to generate a lot of publicity, but not nearly enough details to be able to replicate. IIRC, Deepmind announced the fully peer-reviewed paper to be available soon (no date given). Maybe the Arxiv-version of the Alpha-Zero paper was rushed to generate the publicity during the London Chess Classic last week.