What to do though when Alpha Zero is everyone's laptops, desktops, tablets and phones. WELCOME TO THE BRAVE NEW AI WORLD. Can you say SKYNET ??!!??!!
Like Morphy and Fischer Alpha Zero is done.
Sort:
RMChess1954 wrote:
How many games did Deep Blue play after it won the match with Kasparov?
Deep Blue played Kasparov in a sponsored public match with plenty of news coverage, not in some Google cleanroom environment with a copied piece of open source software. Deep Blue played against a team that was calling its own shots, where Stockfish was basically hijacked and hamstrung for free publicity, with DeepMind team members running both sides with a vested interest in one winning.
Note that if they had decided to use Houdini or Komodo, which are consumer products, Google would have opened themselves up to a lawsuit by "winning" a backroom match which they essentially cheated on and then publicizing it, hurting revenues for those companies.
If Pepsi took a bunch of Coke products behind closed doors, added to/removed some of Coke's ingredients, did some private taste tests but only using Pepsi employees not only to run the tests, but to be the tasters, and then did a press release proclaiming themselves to have beaten Coke in every test...what do you think would happen?
It's only because Stockfish is free and so essentially suffers no significant monetary damages that Google can safely pull this shenanigan off...they know if it goes to court they won't have to pay jack.
The answer to my question was 0. Not the dog house was green.
Dec 12, 2017
0
#25
many think alpha zero may changes the style of playing chess. it is kind of software which basically prefers position over materials while stockfish mostly calculate material and positions equally. so it may changes the style of playing chess in recent future. 
Dec 12, 2017
0
#26
I would agree that the fact that no-one ever told AlphaZero what a piece was worth, or what positional factors were worth how much, allows it to play beautiful, correct, anti-materialistic chess. This win against the queen's indian was a great example, where it plays a gambit and then sacrifices an exchange, ending up with a positional crush an exchange and two pawns down!
RMChess1954 wrote:
The answer to my question was 0. Not the dog house was green.
Not my fault you are asking the wrong questions through faulty understanding of the situation...but your dodging of points you can't really handle is noted. Smart people will understand that your question was framed by your faulty conclusion in the original post, and will ergo answer in context.
Did you stop beating your wife?
What, you don't answer leading questions with a 1 word answer? Neither do I.
For those who think AlphaZero didn't really beat Stockfish you should watch some of the analysis videos like this one. https://www.youtube.com/watch?v=MTWkX6KZO2A&feature=em-subs_digest
Stockfish didn't know what hit it.
If it's so convincing...just play the match publicly with Stockfish's own team handling the settings for Stockfish and opening books/tablebases included. Easy. AlphaZero will surely win again, so why not?
Your analysis video is based on reported game results from a closed room. That's not science, that's snake oil.
The men didn't walk on the moon, the earth is flat. Put on your aluminum hats. It's all a plot.
Iwinyoulosedontcry wrote:
btickler wrote:
If it's so convincing...just play the match publicly with Stockfish's own team handling the settings for Stockfish and opening books/tablebases included. Easy. AlphaZero will surely win again, so why not?
Your analysis video is based on reported game results from a closed room. That's not science, that's snake oil.
BTickler is just a troll on here with bs stories of dating a girl who works at Google. This is the type of guy who would argue about the weather, just a real dick, speading a toxic waste behind himself because he is clearly a miserable person,trying to contradict everything and everyone for no real reason. Nobody can predict the future of chess and the future of whatever Google has planned out with the chess God pet, this is being contradicted simply to contradict everything, by a miserable person.
Oh look, it's Lawandorderking. You're going to have get over being butthurt and follow someone else around...if this keeps up people might think you have an unhealthy attraction towards me or something.
RMChess1954 wrote:
The men didn't walk on the moon, the earth is flat. Put on your aluminum hats. It's all a plot.
I see...you have no argument for that one. Time to deflect.
Iwinyoulosedontcry wrote:
btickler wrote:
RMChess1954 wrote:
The men didn't walk on the moon, the earth is flat. Put on your aluminum hats. It's all a plot.
I see...you have no argument for that one. Time to deflect.
Its draining to argue with an Idiot on here, some people dont care enough about you or your opinions and fake stories.
Maybe let people speak for themselves, and quit being some Don Quixote chasing me around on your donkey? It's demeaning for you. Really. Let it go. Be a big boy.
It's amazing that A0 managed to crush SF by playing in romantic era playstyle. Great feat, but they can't really claim they beat SF legitimately. The win shows that A0 is +100 elo over SF with an astrix.
1. SF version was outdated. The new SF 9 Dev is at least 50 ELO stronger from what they say.
2. SF didn't use opening book (50 elo reduction) and Syzygy 6-men tablebase. It's well known that SF is not strong without it.
3. 1GB Hash on a 64 core machine with 1 min/move is a joke. It's critical with many threads because SF uses hash as the main mechanism through which all threads communicate. Another significant ELO reduction.
4. Extreme emphasis was put into making SF understand positions which are critical and which are not. Based on this SF manages its clock very carefully spending significantly more time on some positions. Disabling this feature and forcing SF to use its time based on 1 min/move is another large ELO reduction.
5. Robert Houdart (Houdini creater) said that Houdini with equivalent core as A0 would've rekt SF the same way. No big deal.
[COMMENT DELETED]
Dec 13, 2017
0
#36
Redlynx17 wrote:
It's amazing that A0 managed to crush SF by playing in romantic era playstyle. Great feat, but they can't really claim they beat SF legitimately. The win shows that A0 is +100 elo over SF with an astrix.
1. SF version was outdated. The new SF 9 Dev is at least 50 ELO stronger from what they say. Provide a link. It will be true when it is verified, not when some anonymous source says it, but the current version of Stockfish is version 8. It would be inappropriate to use a version that is still in alpha testing.
2. SF didn't use opening book (50 elo reduction) Guesses are not proofs and Syzygy 6-men tablebase. It's well known that SF is not strong without it. Not strong without a tablebase? That is a real weakness that needs to be fixed. Even we turn-based players on chess.com aren't allowed to use tablebases, because it removes the competitive element in that phase of the game.
3. 1GB Hash on a 64 core machine with 1 min/move is a joke. It's critical with many threads because SF uses hash as the main mechanism through which all threads communicate. Another significant ELO reduction. You claim. It is not clear what the ideal hash table size was or how much difference there would be. Big is not always better for hash tables: sometimes big means slower.
4. Extreme emphasis was put into making SF understand positions which are critical and which are not. Based on this SF manages its clock very carefully spending significantly more time on some positions. Disabling this feature and forcing SF to use its time based on 1 min/move is another large ELO reduction.
[You suggest that Stockfish's strength is highly dependent on it spending a lot more time on some moves. Where is the quantitative evidence for how much effect this has? There is some evidence that it is not very important with the computing resources used, because Elo was changing very slowly with computing time. Reallocating time between moves cannot have as much effect as providing more time on every move!
Note that this is also a way to make AlphaZero stronger: devoting more time to moves when the variance in the MCTS is higher, but it seems the DeepMind team (possibly rightly) saw this as of limited importance.
5. Robert Houdart (Houdini creater) said that Houdini with equivalent core as A0 would've rekt SF the same way. No big deal.
TPUs are not something you can slot in instead of a CPU: they are specialised matrix processors designed for deep neural networks. Stockfish and Houdini would require rewriting for the different hardware, and there is no guarantee this would work at all well. Likewise I don't believe you can run those programs on GPUs. See this discussion.
You rock Elroch!
Again a video that destroys the nay sayers. https://www.facebook.com/chess/videos/10156055408124571/
The fact is A0 is the strongest player on earth.
Do you think Magnus Carlsen would get at least one win against AlphaZero in a 12-game match? 24 game match? How about one draw in a 6-game match? 12-game match?
How many games did Deep Blue play after it won the match with Kasparov?