AlphaZero - Tactactical Abilities

[tsoj]

I think they were very clear in the paper:

Program Chess ...
AlphaZero 80k ...
Stockfish 70,000k ...
...
-----------------------------------------------
Table S4: Evaluation speed (positions/second) of AlphaZero, Stockfish, and Elmo in chess,
shogi and Go.

They meant how fast they could evaluate positions not how fast they were going through the search tree.

[MikeGL]

I think they were very clear in the paper:

Program Chess ...
AlphaZero 80k ...
Stockfish 70,000k ...
...
-----------------------------------------------
Table S4: Evaluation speed (positions/second) of AlphaZero, Stockfish, and Elmo in chess,
shogi and Go.

They meant how fast they could evaluate positions not how fast they were going through the search tree.

But 70 million for SF8 on 64-core looks like nodes per second and not evaluated positions per second.

[shrapnel]

I think they were very clear in the paper:

Program Chess ...
AlphaZero 80k ...
Stockfish 70,000k ...
...
-----------------------------------------------
Table S4: Evaluation speed (positions/second) of AlphaZero, Stockfish, and Elmo in chess,
shogi and Go.

They meant how fast they could evaluate positions not how fast they were going through the search tree.

Exactly ! Everything is clear in the Paper, but the detractors just see what they want to see and ignore the rest.
Or worse, they put their own spins to what is written.

[shrapnel]

People hear that it is a new approach, they hear AI, neural networks, deep learning, that it has only trained for four hours (which is debatable) and so on. And they want to believe. So they read all sorts of things into the very few published games played.

People not only hear, they also SEE . They have HEARD all that you said and also SEEN how the once mighty Stockfish was not just beaten, but completely HUMILIATED !
The Proof of the Pudding is in the eating.
The people who believe are not all fools, as you seem to suggest.
Also, the highlighted words clearly display your bias, in the sense that you clearly disbelieve what is written in the Paper.
When you don't even believe what was written, what are we arguing about ?
You are no different from a Milos or a Tsvetkov and just pretend to be unbiased, but Lady, your Slip is showing !

[Uri Blass]

Maybe AlphaZero just outsearched Stockfish a little bit. Just like Stockfish would against itself on better hardware.

Sorry, but I totally disagree.
It was much more than just a matter of outsearching Stockfish ; it was the whole, overall STYLE of Play that AlphaZero showed that was so impressive and breath-taking.
The Brilliant moves themselves like the famous 19. Re8 played in one of the Games, could be found by other Engines like Komodo and Houdini after suitable Tweaks to the Parameters.
But just finding those moves didn't make the tweaked Komodo or Houdini into a AlphaZero !
What impressed me the most was the almost complete disregard for Material Balance that AlphaZero exhibited.
This, more than any deficiencies in Search and Evaluation, I am convinced , completely unsettled Stockfish which was burdened with conventional ideas of material value.
It was as if AlphaZero, looking at the Board, foresaw ( I can't think of any other word) a Position that was favorable to it, and immediately started making (tactical) moves that would lead to that Position, without caring in the least how much material it threw away in the Process !
It was like putting the Brain of Mikhail Tal/Bobby Fischer into a Robot and calling it AlphaZero.
The Brilliance of the Human was there, but without any human weaknesses.
If you actually play through those Games, you can almost SENSE the discomfort, in a manner of speaking, that Stockfish was experiencing in playing against such an unusual opponent.
All the technical mumbo-jumbo about hardware differences or a lack of an Opening Book, or Time Control not suiting Stockfish or other such rubbish excuses can't hide the FACT that Stockfish was absolutely and comprehensively out-classed by the amazing AlphaZero.
Stockfish simply didn't know what hit it !
The people still having a vested interest in the now obsolete chess engines, fully realize this even if they don't admit it, and are desperately looking for excuses to explain this catastrophe (for them) and are hiding behind fake Analysis, calculations and other Technical mumbo-jumbo.
But Chess is more than just scientific calculations and one doesn't have to be a Programmer, just a reasonably good chess player, to know the Truth.

I only know that we saw a selective part of the games.
We did not see all the draws of alphazero against stockfish.

I do not think that the result as catastrophe for somebody and we simply do not have all the details.

I do not think that the decision of alphazero developers to give only games that alphazero won was a correct decision
and I think that if you look only at games that a player won you can get a biased opinion about the player.

[CheckersGuy]

Maybe AlphaZero just outsearched Stockfish a little bit. Just like Stockfish would against itself on better hardware.

Sorry, but I totally disagree.
It was much more than just a matter of outsearching Stockfish ; it was the whole, overall STYLE of Play that AlphaZero showed that was so impressive and breath-taking.
The Brilliant moves themselves like the famous 19. Re8 played in one of the Games, could be found by other Engines like Komodo and Houdini after suitable Tweaks to the Parameters.
But just finding those moves didn't make the tweaked Komodo or Houdini into a AlphaZero !
What impressed me the most was the almost complete disregard for Material Balance that AlphaZero exhibited.
This, more than any deficiencies in Search and Evaluation, I am convinced , completely unsettled Stockfish which was burdened with conventional ideas of material value.
It was as if AlphaZero, looking at the Board, foresaw ( I can't think of any other word) a Position that was favorable to it, and immediately started making (tactical) moves that would lead to that Position, without caring in the least how much material it threw away in the Process !
It was like putting the Brain of Mikhail Tal/Bobby Fischer into a Robot and calling it AlphaZero.
The Brilliance of the Human was there, but without any human weaknesses.
If you actually play through those Games, you can almost SENSE the discomfort, in a manner of speaking, that Stockfish was experiencing in playing against such an unusual opponent.
All the technical mumbo-jumbo about hardware differences or a lack of an Opening Book, or Time Control not suiting Stockfish or other such rubbish excuses can't hide the FACT that Stockfish was absolutely and comprehensively out-classed by the amazing AlphaZero.
Stockfish simply didn't know what hit it !
The people still having a vested interest in the now obsolete chess engines, fully realize this even if they don't admit it, and are desperately looking for excuses to explain this catastrophe (for them) and are hiding behind fake Analysis, calculations and other Technical mumbo-jumbo.
But Chess is more than just scientific calculations and one doesn't have to be a Programmer, just a reasonably good chess player, to know the Truth.

I only know that we saw a selective part of the games.
We did not see all the draws of alphazero against stockfish.

I do not think that the result as catastrophe for somebody and we simply do not have all the details.

I do not think that the decision of alphazero developers to give only games that alphazero won was a correct decision
and I think that if you look only at games that a player won you can get a biased opinion about the player.

You do know that this was just a preprint and not the entire paper ? Why can't people wait until DeepMind has published the entire paper before accusing DeepMind of not doing a good job.

I bet that we will see many many more games published by DeepMind so watch out

[pinkfloydhomer]

You do know that this was just a preprint and not the entire paper ? Why can't people wait until DeepMind has published the entire paper before accusing DeepMind of not doing a good job.

I bet that we will see many many more games published by DeepMind so watch out

No one is accusing AlphaZero of not doing a good job?
It is very telling how some people choose to read the points given in this thread.

My main point is that we don't know very much at this stage and that people are jumping to conclusions and are extrapolating groundlessly.

AlphaZero might be positionally godlike or it might just be a tiny bit better tactically for much more mundane reasons. We don't know at this point.
We also don't know how strong it really is and we can't conclude anything about it yet, even though a lot of people do. It might be just a little bit stronger than the version of Stockfish it played against or it might be much, much better so that even if/when we test it against stronger and stronger opponents, we wouldn't hit it's peak strength until much later. People seem to assume this blindly, but we simply don't know at this point. We have seen it a million times in computer chess: An engine which is a little bit better, especially tactically, can dominate it's opponent and seem to have a much better positional understanding.

Imagine playing my old engine Chezzz against Stockfish at a hardware disadvantage over 100 games. The hardware disadvantage in this case handpicked so that the result ends up something like 72 wins/28 draws/0 losses for Chezzz. This is easy to do. It might mean that Chezzz runs at a modern PC and that Stockfish runs on my old HTC Desire phone or even worse. Chezzz most certainly doesn't have superior positional understanding compared to Stockfish (quite the opposite). Not in general and not in this scenario. What it does have in this scenario is a tactical advantage. You can even calculate how much better Chezzz is in this scenario in terms of rating (in a given rating pool, ratings are not fixed quantities as some people seem to think).

We have _no_ objective, scientific way of testing at this point how strong AlphaZero actually is or if it is primarily for tactical reasons (how it deals with the search tree) or for positional reasons (how the neural network evaluates positions). And I know that these two are related and that there isn't a clear line between tactical and positional features, but you get the idea.

[Milos]

I bet that we will see many many more games published by DeepMind so watch out

Are you working in DeepMind?
What would you like to bet, I'm interested?

[CheckersGuy]

I bet that we will see many many more games published by DeepMind so watch out

Are you working in DeepMind?
What would you like to bet, I'm interested?

No but Demis Hassabis twittered that they are going to publish the full paper. I bet that they publish more games along with that paper

[hgm]

I am a bit puzzled by this whole paper business. Normally a preprint is exactly what you submitted for publishing. Only if the referees give you a hard time you would change things to avoid rejection, and you will not be looking for a hard time.

I addition, journals like Nature or Science are very picky about novelty. If discoveries are already floating around publicly, they usually do not want to touch them. Shouting around results before they lift the embargo on submitted manuscripts is usually a sure way to get rejected.