Latest Updates:
Page Index Toggle Pages: [1] 2 
Topic Tools
Very Hot Topic (More than 25 Replies) New book "Game Changer" on Deepmind's AlphaZero (Read 7454 times)
Jupp53
God Member
*****
Offline


be

Posts: 864
Location: Frankfurt/Main
Joined: 01/04/09
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #28 - 03/30/19 at 10:18:34
Post Tools
an ordinary chessplayer wrote on 03/28/19 at 21:45:53:
ErictheRed wrote on 02/14/19 at 16:29:34:
Humans will never learn to "think like computers;" it's simply impossible.  That doesn't mean that we can't learn something from a computer's games, however.


barnaby wrote on 03/28/19 at 13:58:02:
With much respect for you and your thoughts and ideas here both chess and otherwise, I feel I have to disagree here on this concept and I will try to explain why. (/big snip)

I think you just expressed the same idea as ErictheRed, only using a lot more words.


Which very good in some cases and for some purposes. Here it is the reflection of "thinking", "concept" and related terms.

If I see some discussions of the book, I feel that old concepts are seen as product of the new engine and here recommended by the New In Chess newsletter for marketing purposes. https://www.youtube.com/watch?v=u552M_7T8gA&mc_cid=235f4d3f85&mc_eid=b3953ab336. Open lines and diagonals are part of Lasker's list of Steinitz's elements of strategy. King's security belong to it too. So it is not necessary to see something new in every example, which is presented now.
  

Medical textbooks say I should be dead since April 2002.
Dum spiro spero. Smiley
Back to top
 
IP Logged
 
an ordinary chessplayer
God Member
*****
Offline


I used to be not bad.

Posts: 563
Location: Columbus, OH (USA)
Joined: 01/02/15
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #27 - 03/28/19 at 21:45:53
Post Tools
ErictheRed wrote on 02/14/19 at 16:29:34:
Humans will never learn to "think like computers;" it's simply impossible.  That doesn't mean that we can't learn something from a computer's games, however.


barnaby wrote on 03/28/19 at 13:58:02:
With much respect for you and your thoughts and ideas here both chess and otherwise, I feel I have to disagree here on this concept and I will try to explain why. (/big snip)

I think you just expressed the same idea as ErictheRed, only using a lot more words.
  
Back to top
 
IP Logged
 
barnaby
Senior Member
****
Offline


The night is dark and
full of terrors.

Posts: 345
Joined: 01/09/12
Gender: Female
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #26 - 03/28/19 at 13:58:02
Post Tools
ErictheRed wrote on 02/14/19 at 16:29:34:
Humans will never learn to "think like computers;" it's simply impossible.  That doesn't mean that we can't learn something from a computer's games, however.



With much respect for you and your thoughts and ideas here both chess and otherwise, I feel I have to disagree here on this concept and I will try to explain why.

1.  Computers do not 'think.'  They do not come up with ideas. 

https://www.psychologytoday.com/us/blog/the-voices-within/201008/what-do-we-mean...

2. Humans beings can process information the same way a computer does but cannot operate at the same speed or with the same consistency or for as long a period of time.

3. It is the computer which cannot think like a human.

What we can 'learn' from computers and the way in which they process information is for human beings to process the information in different ways and to create new ways of evaluating and analysing which leads to wider taxonomic applications.

Computers can now create art, write books, music, etc.  But they cannot critique those creations and decide which art is culturally relevant nor can they offer value judgments regarding those creations. 

They can create things but not concepts.  They cannot think.

  
Back to top
 
IP Logged
 
CanadianClub
Senior Member
****
Offline


Greetings from Catalonia!

Posts: 352
Joined: 11/11/12
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #25 - 03/22/19 at 15:07:04
Post Tools
Leon_Trotsky wrote on 02/14/19 at 05:09:01:
How do you feel that reading the book changed your opening repertoire ¿


I haven't read the book yet. But I usually feel myself "more bad" playing as Black when I play against Larsen (1.b3) or against simple non-theory-Bc4-Sicilian, or against shy English tries with b3 or e3... because I am out of my comfort zone.

This is the idea that (if I am not mistaken) was said here in the post. Not focusing on knowing theory (where Black can -and probably will- answer with 15+ theory move to equalize completely) but on position where previous knowledge and (even equal) better piece placement, and better understanding where we can dig in....

I have changed as White my mind in this way and I play Nf3 to get this.
  
Back to top
 
IP Logged
 
an ordinary chessplayer
God Member
*****
Offline


I used to be not bad.

Posts: 563
Location: Columbus, OH (USA)
Joined: 01/02/15
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #24 - 03/20/19 at 17:46:40
Post Tools
Unannotated games from the book are available:
https://www.newinchess.com/nic-books-games-downloads
A quick count of the contents ...
  • 69 games AlphaZero vs Stockfish 8
  • 20 games grandmaster vs grandmaster
  • 4 games Sadler vs Smallfish
  • 1 game Houdini vs Stockfish
  • 10 positions without moves
  • 2 AlphaZero/Sadler analyses
  
Back to top
 
IP Logged
 
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2449
Location: USA
Joined: 10/02/05
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #23 - 02/14/19 at 16:29:34
Post Tools
Humans will never learn to "think like computers;" it's simply impossible.  That doesn't mean that we can't learn something from a computer's games, however.
  
Back to top
 
IP Logged
 
Leon_Trotsky
Senior Member
****
Offline


Кто был никем — тот станет
всем!

Posts: 434
Location: Barcelona, CAT
Joined: 08/11/17
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #22 - 02/14/19 at 05:09:01
Post Tools
How do you feel that reading the book changed your opening repertoire ¿
  
Back to top
 
IP Logged
 
trw
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1260
Joined: 05/06/08
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #21 - 02/14/19 at 05:03:24
Post Tools
Looking forward to cracking my copy in two days.  Cheesy
  
Back to top
IP Logged
 
phonological_loop
YaBB Newbies
*
Offline


I Love ChessPublishing!

Posts: 37
Joined: 09/10/17
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #20 - 02/07/19 at 20:16:34
Post Tools
katar wrote on 02/07/19 at 18:38:50:
I got the impression that AZ does not play 1.e4 because of 1...e5 - specifically the Berlin, which AZ favors as Black.  I have wondered whether AZ would adopt 1.e4 if the opponent had to play anything other than 1...e5.


The current best nets for Lc0 (T30 and T40) both favor 1. e4, and both are stronger than A0. Refer to the second graphs on these pages:
https://script.google.com/macros/s/AKfycbz3cw6fXOQ2T2QIpMzSEdUNwVaqfmYpR0LWvP3zw...

https://script.google.com/macros/s/AKfycbwAVsblBYs17sxoao9wKiz88m2FfBG-154ncQ1lP...
  
Back to top
 
IP Logged
 
ReneDescartes
God Member
*****
Offline


Qu'est-ce donc que je
suis? Une chose qui pense.

Posts: 1035
Joined: 05/17/10
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #19 - 02/07/19 at 19:25:53
Post Tools
phonological_loop wrote on 02/07/19 at 18:18:44:
barnaby wrote on 02/07/19 at 17:19:52:
Sacrificing pawns more freely now, playing more on the wings by using a and h pawns more, changed opening repertoire from 1.d4 to 1. c4 and 1. Nf3 and stopped playing for advantages as White through known theory and now focusing on creating better outpost squares for pieces.  More actively thinking now: "what would AZ do here" and ," how can I give up material to gain positional advantage."


Thank you!

I am a (bad) amateur who plays exclusively 1. e4, so the following question comes from a place of total ignorance. What is so special about 1. c4 and Nf3 as compared to d4? My impression is that many of the lines of the former two options transpose to d4 lines anyway, so that they are tightly related.

Do you think the c4/Nf3 are somehow objectively better (and if so, for what reason? lc0, which a stronger neural network-based engine than A0, does not seem to favor c4/Nf3). Or are you making a practical decision, trying to enter positions you might be more familiar with than your opponent, and which may give them more opportunities to go wrong (because typically people are less booked up and experienced against these less common openings, for instance)? Or some other reason?


People play 1.Nf3 and 1.c4 for different reasons. Many want to play main 1.d4 lines against most defenses, but, say, avoid certain sharp defenses, playing (after 1.Nf3) an Anti-Benoni rather than a Benoni, or (after 1.c4) an Anti-Gruenfeld rather than a Gruenfeld. Of course, they have to be ready for 1..c5, in which case they cannot reach a 1.d4 mainline if their opponents do not want them to--and, in the case of 1.c4, also 1...e5 with some of the English main lines.

Other players  have no intention of playing any main 1.d4 lines, and want to go into a Reti, KIA, English, etc.

Still others are winging it and just want to create fluid, unfamiliar (mostly hypermodern) situations for their opponents, as you suggested.

1.d4 2.Nf3 is a similar beast--some play it for d-pawn specials, others to play 3.c4 while avoiding certain sharp lines.

--All these lines are definitely no stronger than 1.d4 2.c4 in any objective sense--on the contrary. While that classic opening permits Black some sharp defenses (QGA, Benko, Albin, Schara), it also gives White some of his most theoretically powerful lines (the flick-knife Benoni, Saemisch KID, QGD Exchange with Ne2 and f3, etc), where pressure on Black is maintained for a long time even if eventually Black can thread through to equality.
  
Back to top
 
IP Logged
 
katar
Senior Member
****
Offline


look another year went
by

Posts: 424
Location: LA
Joined: 09/21/05
Gender: Male
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #18 - 02/07/19 at 18:38:50
Post Tools
I got the book on the FwdChess app.  It looks interesting.  The authors try to compare AZ's strategies with maneuvers by great human players in classic/historical games.  This adds to the pedagogic value and softens the criticism that readers should not try to play like a computer.

I got the impression that AZ does not play 1.e4 because of 1...e5 - specifically the Berlin, which AZ favors as Black.  I have wondered whether AZ would adopt 1.e4 if the opponent had to play anything other than 1...e5.

Sadler & Regan also recommended this flexible 1.Nf3 or 1.c4 or 1.d4 approach in "Chess for Life", citing Pia Cramling as a model player.  The general approach is to be shifty/crafty/flexible as White by employing move orders and transpositions.  And as Black, to study a mainline in extreme depth.  Sadler & Regan write: "It is easier to destroy [as Black] than create [as White]."

In a Youtube video, Natasha Regan stated that after working on the book she tried playing "like AZ" with poor results.  She commented that her victorious opponents, in the post-mortems, patronizingly tried to teach her that it is better to play in the center than to throw the rook-pawns. 
  

2078 uscf
Back to top
WWW  
IP Logged
 
phonological_loop
YaBB Newbies
*
Offline


I Love ChessPublishing!

Posts: 37
Joined: 09/10/17
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #17 - 02/07/19 at 18:18:44
Post Tools
barnaby wrote on 02/07/19 at 17:19:52:
Sacrificing pawns more freely now, playing more on the wings by using a and h pawns more, changed opening repertoire from 1.d4 to 1. c4 and 1. Nf3 and stopped playing for advantages as White through known theory and now focusing on creating better outpost squares for pieces.  More actively thinking now: "what would AZ do here" and ," how can I give up material to gain positional advantage."


Thank you!

I am a (bad) amateur who plays exclusively 1. e4, so the following question comes from a place of total ignorance. What is so special about 1. c4 and Nf3 as compared to d4? My impression is that many of the lines of the former two options transpose to d4 lines anyway, so that they are tightly related.

Do you think the c4/Nf3 are somehow objectively better (and if so, for what reason? lc0, which a stronger neural network-based engine than A0, does not seem to favor c4/Nf3). Or are you making a practical decision, trying to enter positions you might be more familiar with than your opponent, and which may give them more opportunities to go wrong (because typically people are less booked up and experienced against these less common openings, for instance)? Or some other reason?
  
Back to top
 
IP Logged
 
barnaby
Senior Member
****
Offline


The night is dark and
full of terrors.

Posts: 345
Joined: 01/09/12
Gender: Female
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #16 - 02/07/19 at 17:19:52
Post Tools
phonological_loop wrote on 02/05/19 at 17:15:20:
barnaby wrote on 02/05/19 at 15:54:04:
I have already significantly changed my own playing style including opening repertoire and middle game planning with highly successful results, all based on studying AZ games.  I just put my rating at its all time peak, which is something most people close to 60 years old do not find so easy to do, and I attribute much of it to the change in my play that I decided to make after spending months studying these games.

AZ is an absolute game changer (for those of us willing to change our game).


Could you comment on what you've learned and changed, if you have the time? I'd be extremely interested in what you have to say, and I'm sure others would too.



A over simplified take:


Sacrificing pawns more freely now, playing more on the wings by using a and h pawns more, changed opening repertoire from 1.d4 to 1. c4 and 1. Nf3 and stopped playing for advantages as White through known theory and now focusing on creating better outpost squares for pieces.  More actively thinking now: "what would AZ do here" and ," how can I give up material to gain positional advantage."



  
Back to top
 
IP Logged
 
barnaby
Senior Member
****
Offline


The night is dark and
full of terrors.

Posts: 345
Joined: 01/09/12
Gender: Female
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #15 - 02/07/19 at 05:07:31
Post Tools
  
Back to top
 
IP Logged
 
edgy
Junior Member
**
Offline


lacking the master touch

Posts: 77
Location: New York
Joined: 08/09/04
Re: New book "Game Changer" on Deepmind's AlphaZero
Reply #14 - 02/06/19 at 15:01:03
Post Tools
RdC wrote on 02/06/19 at 01:43:14:
Jonathan Tait wrote on 02/05/19 at 23:39:20:
Well, perhaps I'll push my h-pawn down the board even more than I usually do Smiley


It used to be something of a Larsen speciality to prod the rook pawns down the board. But like Alpha Zero, he was self-taught, mostly at least.


I recall he said that he wanted to call his selected games collection _The Strategic Significance of the Rook Pawns_ but the publisher wouldn't let him.
  

Caissa have mercy on a miserable patzer: http://altergoniff.blogspot.com
Back to top
WWW  
IP Logged
 
Page Index Toggle Pages: [1] 2 
Topic Tools
Bookmarks: del.icio.us Digg Facebook Google Google+ Linked in reddit StumbleUpon Twitter Yahoo