DeepMind's AlphaZero plays chess like a tornado in the junkyard

[Speedwell]

dp

 

[Chesterton]

it would require creationists to walk back their claims from "the process can't work" to "God created the process."

I'm really not sure biology and electronics are that completely analogous.

 

[Astrophile]

I am curios what take creationist have on this; if randomness cannot create design, no matter what time is given, then what did cause AlphaZero to achieve a superhuman performance in chess in less than four hours time?

I suppose that the answer is that randomness can create design.

 

[Speedwell]

I can't speak for creationists, except to say that even if they work on the same principles, the origin is important.

Perhaps it would require creationists to walk back their claims from "the process can't work" to "God created the process."

 

[Speedwell]

I'm really not sure biology and electronics are that completely analogous.

The math involved is very similar and iron-clad.

 

[In situ]

I can't speak for creationists, except to say that even if they work on the same principles, the origin is important. AlphaZero could not and would not happen naturally, so therefore I'd argue that neither can DNA and consciousness and all the biological stuff.

I refer you to post #81 and post #86 for my arguments why the origin is irrelevant as a distinction.

I just called it that to distinguish it from theistic evolution, which I think would make a difference for what's being discussed in this thread.

And I am telling you it is not a good label since there is no such thing as "atheistic evolution", you could just as well had made the distinction by saying "scientist evolution" or "polar bear evolution". It still does not make much sense to me to single out a specific group which shares properties only superficial correlates with naturalistic evolution.

 

[In situ]

Perhaps it would require creationists to walk back their claims from "the process can't work" to "God created the process."

Which is theistic evolution, therefore a creationist cannot claim this to be the distinction. I.e. there must exists a distinction separate from this claim. I am not saying a creationist have to name the distinction, I am just saying it must exists a distinction for creationist to be correct about the claim that evolution does not work.

 

[In situ]

I suppose that the answer is that randomness can create design.

Whether designed by natural process or and intelligent agent all designed things has a purpose and meaning. In chaos there exist no meaning (or an infinite number of meanings, dependent on how you see it) so design cannot be acheived from it.

 

[In situ]

I'm really not sure biology and electronics are that completely analogous.

I just had some electrons with my coffee. They tasted good.

 

[Chesterton]

I refer you to post 81 and post #86 for my argument why the origin is irrelevant as a distinction.

And I am telling you it is not a good label since there is no such thing as "atheistic evolution", you could just as well had made the distinction by saying "scientist evolution" or "polar bear evolution". It still does not make much sense to me to single out a specific group which shares properties only superficial correlates with naturalistic evolution.

I hadn't actually read anything on AlphaZero. Before we get carried away it would be good to make sure the science is good. This guy raises some questions in this article, at one point noting this:

Self-play. Does AlphaZero completely learn from self-play? This seems to be true according to the details provided in the paper, but with two important nuances: the rules and the typical number of moves have to be taught to the system before starting playing with itself. The first nuance, although looking obvious, is not as trivial as it seems. A lot of work has to be dedicated to find a suitable neural network architecture on which these rules are encoded, as also explained in the AlphaZero paper. The initial architecture based on convolutional neural networks used in AlphaGo was suitable for Go, but not for other games. For instance, unlike Go, chess and shogi are asymmetric and some pieces behave differently depending on their position. In the newest AlphaZero, a more generic version of the AlphaGo algorithm was introduced, englobing games like chess and Shogi. The second nuance (i.e. the typical number moves was given to AlphaZero to “scale the exploration noise”) also requires some prior knowledge of the game. Also, the games that exceeded a maximum number of steps were terminated with a draw outcome (this maximum number of steps is not provided). It is not clear whether this heuristic was also used in the games against Stockfish or only during training.​

 

[Astrophile]

Whether designed by natural process or and intelligent agent all designed things has a purpose and meaning. In chaos there exist no meaning (or an infinite number of meanings, dependent on how you see it) so design cannot be acheived from it.

Perhaps I should have said that randomness can create the appearance of design. However, I don't agree with you. For one thing, randomness and chaos are not the same thing. For another, meteorology is chaotic, since the weather can't be predicted accurately for more than a few days, and yet snowflakes, clouds and tornadoes give the appearance of having been designed. I suspect that we mean different things by the words 'design', 'randomness' and 'chaos'.

 

[In situ]

I can't speak for creationists, except to say that even if they work on the same principles, the origin is important. AlphaZero could not and would not happen naturally, so therefore I'd argue that neither can DNA and consciousness and all the biological stuff.

Beside the objection I already made there is another objection to the above and that is the null position. The null position is that you cannot claim the above to be either true or false. If the null position is valid then your argument is invalid.

This is why:

One cannot assumes that order is a degenerated state of design and that order cannot create design, since we know design can emerge from order and the other way around. Order was given to the world of AlphaZero in the form of the chess rules it been provided with. Since there now exists order in the mini world of AlphaZero, then it implies that design can, in principle, emerge from this order. Which is precisely what happen (see post #81 and post #86). In a similar fashion, the laws of nature impose order on the universe. This order is, in principle, sufficient to explain "DNA and consciousness and all the biological stuff".

To know if the null position is valid we first need to ask: was the the order in the universe imposed on the universe by an agent? This is a question on whether or not order can emerge from chaos. The answer is; it can.

Hence the null position is valid.

Therefore we can just as well assume the universe emerge from chaos. This implies AlphaZero can emerged from chaos and therefore your claim that "AlphaZero could not and would not happen naturally" is invalid.

 

[In situ]

Perhaps I should have said that randomness can create the appearance of design.

You can at most achieve order, but not design. From order you can then create design. However, just because A implies B and B implies C does not mean A implies C, i.e. you cannot just assume the transitive law is true for every relation, that is you are not entitled to just skip one phase in how design can be achieved without actually showing this to be the case. And to my knowledge nobody has done that yet.

However, I don't agree with you.

Feel free to disagree. That is the best way to straighten out once own thoughts.

For one thing, randomness and chaos are not the same thing.

You can claim that, but it does not make your claim true.

For another, meteorology is chaotic, since the weather can't be predicted accurately for more than a few days, and yet snowflakes, clouds and tornadoes give the appearance of having been designed.

Beside the non sequitur, chaos and chaotic system is not the same thing. A chaotic system is simply put a nonlinear system. Nonlinear system posses order, i.e. the trajectory of the system. Here is an example of a chaotic trajectory:

Does the picture above look like the picture below to you?

A chaotic system is not the same as chaos, in fact chaos is not a system at all because it is random - which is the complete opposite to a system.

I suspect that we mean different things by the words 'design', 'randomness' and 'chaos'.

I suspect you are not clear on the definitions of things and therefore you make incorrect conclusion. I suspect the reason a tornado emerge is because of the order which exists in a weather system, i.e. it is the order in chaotic system which gives rise to design not the randomness in the chaotic regions.

And yes, if you think cloud looks designed, then we definitely mean different things with design.

 

[In situ]

I hadn't actually read anything on AlphaZero. Before we get carried away it would be good to make sure the science is good. This guy raises some questions in this article, at one point noting this:

Thank you! I will read his article later and see what he has to say.

 

[In situ]

Self-play. Does AlphaZero completely learn from self-play? This seems to be true according to the details provided in the paper, but with two important nuances: the rules and the typical number of moves have to be taught to the system before starting playing with itself.

At a fast glance at this quote it seems the author might have conflated different concepts. See post #86 and maybe post #81 to understand my argument against the objection that AlphaZero did not learned something from the self play and why the rules of chess must be given as domain knowledge.

 

[In situ]

I hadn't actually read anything on AlphaZero. Before we get carried away it would be good to make sure the science is good. This guy raises some questions in this article, at one point noting this:

This article raised a red flags in my mind already at the start. The author starts with an appeal to authority - this is a needle in my eyes right at the beginning. He tries to smooth this out thou. Interestingly enough nothing what the author claims in the remaining article seems to be actually related to his claimed expertise so I don't see why he needed to mention it to start with.

While continue reading, more red flags started to raise and now about several conflicting claims. He claims things which does not resonate with what I read elsewhere, e.g. he claims the allocation of resources was unfair, however I read claims the setting was fair. It also seams like he actively trying to make it look more bad for AlphaZero than actually was the case, e.g. claiming AlphaZero won by 64 to 36 thus giving the impression Stockfish had won 36 games while in fact the author split the 72 draws as 36 wins to the both of them (AlphaZero won with 28-0). Why one would like to present the result in this way is beyond my understanding and I can only speculate in why.

Anyhow, I can go on and on and rant about what I found strange, but sufficient to say is that I had a weird feeling from start until end when reading this article. This feeling was confirmed when afterwards I had a fast glance at the comment section. One comment claimed he was cleaving hairs just to make a point and had missed the real claims of the research paper (and I agree), another comment basically asked if the author was stupid or not when pointing out if the author actually thought AlphaZero was supposed to reinvent the rules of chess. I also started to wonder if the author really had all brain faculties in order when he in essence replied "yes I said that but I did not say that" to some other criticism.

My impression of this this guy is not good. My feeling is that he is not to bright, self-occupied, and an attention seeker. His article just feels wrong from start to end in a multitude of ways. What he claims does not resonate with his assertion to have a "reasoned opinion on the subject as constructive as I could". The author only present one side of the story, done in a way that it looks as bad as possible for AlphaZero. And consider his article is published the 11th December, just less than a week after the result was announced, it seams like he was more in a hurry to get in print than to actually be "as constructive as [he] could".

 

[FrumiousBandersnatch]

As a side note. There exists devices which generates true random number series. These numbers are generated by utilizing radioactive decays and are fairly easy to built into the electronics of a computer.

Yes, I mentioned that.

 

[FrumiousBandersnatch]

This article raised a red flags in my mind already at the start.

I thought it was fair comment - the author seemed mainly concerned with the level of hype around a single paper with rather limited detail and no means yet for replication. It's reasonable skepticism, which is good science - and it should be possible to assuage all his doubts with more information and public demonstration of the claims, running the competing systems on comparable computing platforms.

 

[Larniavc]

Interesting OP. (I love chess.) I agree that we get many interesting variations from randomness. However, I don't know if the Alpha Zero program was truly only random. As you stated it was "given the rules of chess". That seems to be like a big leap up from assembly a Boeing 747 in a junkyard. Unless the program was given detailed designs of a Boeing and then just created the jet from the parts available. This seems different to me than just randomness.

You could argue that the tornado was given the rules of nature.

 

[durangodawood]

Please reread the OP. I already stated telos in that form is irrelevant .

Well its relevant now, because you stated AZ didnt have a telos, and if so, that makes its decision-to-learn a stunning breakthrough of deep consequence.