Generating information in a neural network playing chess

Complex information arising from random processes: the chess program AlphaZero

quote:

---

A sticking point among a lot of ID/Creationists seems to be whether information can be generated by a non-intelligent process.

​

Computer chess provides a very dramatic example. Previously, chess-playing algorithms were designed by software developers in concert with strong human players. In 2017, however, the AlphaZero program was created by initializing a neural net with the rules of chess, and then having it play millions of games against itself, updating its neural net with each game. After 24 hours of such self-play, AlphaZero won a 100-game match against Stockfish, the currently strongest human-designed algorithm, with the score of 28 wins, 82 draws, no losses. (It was not allowed to update its net based on its games with Stockfish, so did not learn from playing the champion; all it had was its experience playing against itself.)

---

Links and Information, I suppose.

Thread copied here from the Generating information in a neural network playing chess thread in the Proposed New Topics forum.

A bit of a vague, non-committal, and not necessarily true statement. Neverthless, it has no bearing..... because even if you classify what was generated as information, it was generated by an intelligent process. The AlphaZero program and algorithms were made by intelligent minds.

The Alpha Zero program wrote its own "algorithms".

LOL, no it didnt. It was programmed with a MCTS and Alpha-Beta search looking for slightly different outcomes from other chess algorithms. Add in some good memory and compute power and you ended up with the best chess program to date.But it's intelligently designed all the way down.

It wrote its own algorithms.

You mean it used evolved algorithms? Ones that didn’t exist before? That it developed ... through evolutionary process of modification and selection?Enjoy

We don't even know what its algorithms are. They were generated by a neural network; I don't know if that counts as an evolutionary algorithm.

Trial & error, selection of fitter algorithms for next generation, modification & repeat.We don’t know the specific DNA changes either ...Enjoy

No it did't. Did you even look at the paper referenced about the program? The one named "Mastering Chess and Shogi by Self-Play with a
General Reinforcement Learning Algorithm" which describes the general idea of what was developed as the program, what instructions and concepts were programmed in.....by programmers???

No, it didnt "evolve" it's own algorithm. It was programmed with the rules of chess (and other games as it turns out), provided formulas for weighing different moves in the game, given memory and the ability to compare different outcomes so that it could use what it had previously determined and build from there. All the instructions for it to do what it was going to do, the algorithm, was given to it by the programmers. The program did not learn to do any new tasks or abilities that were not given to it.

It was programmed with the rules of evolution, provided formulas for weighing different survival strategies, given memory and the ability to compare survival outcomes so that it could use what it had previously determined and build from there. All the instructions for it to do what it was going to do, the algorithm, was give to it by the programmers. The program did not learn to do any new tasks or abilities that were not given to it.Still sounds like an evolutionary process. No it didn't sudden-macro-evolve a new task, or start playing a different game, but evolution doesn't work that way either.btw -- there are more than one algorithm involved. Just like there are more than one survival strategy.Enjoy

A bit of a generous description, don't you think? After all, there was no mention of the word "evolution" or "evolve", and only one instance of "selection" that had nothing to do with what could be inferred as a "rule of evolution". But hey, if you want to say that, I'll give that to ya.And yet with that sort of description, the results end up being totally within the parameters of a designed system and show nothing akin to the claims of biological evolution being able to accomplish. I'm curious what you think "evolved" in the system.The search space of moves is immense and could not fully be searched in detail with modern technology. And yet that space all falls within the parameters of the rules of chess. The programmers gave the chess program a way to heuristically search through the space, find patterns and apply weights to what it searched to build a statistically more-likely-to-win method to playing the game. It was intelligently designed all the way down, and in the end, no matter how well the program performed it's job (surviving in analogy), it still was just 'playing' chess.... as designed. It didn't learn or develop anything new on it's own.If the whole point was to take the chess algorithm and analogize it to something like natural selection, ok. Nobody, including any ID proponent, is seriously debating that natural selection is a real thing. But I disagree with the title of the PT article, in that there was no complex information created, nor was the process random (claims the paper on the chess program never even hinted at). Any information the program dealt with and the process was programmed in already: ie chess rules, search methods, and algorithm(s) dealing with weighing results. Beyond doing what it was programmed to do in a way that exceeded prior programs, there's nothing there to blow a horn about.Edited by WookieeB, : cleanup

Parody analogy wins again. The winning Strategy. It was not programmed with the winning strategy. So no it was not intelligently designed all the way down it had to develop that strategy through trial and error. It learned how to win using an algorithm that was not part of the programming. Which is meaningless, as (A) this is not a defined biological term, (B) it is not quantifiable by any means I know, and (C) evolution does not have to do anything but evolution.You know ID fails when it makes up sciency sounding word and provides no means to quantify them. That’s pseudoscience.Enjoy

How so? What is your definition of "evolved" then in this instance? What an odd statement! Of course the program was not given the explicit "winning strategy". If the programmers had 'the winning strategy', they wouldn't need to program anything. They did though program how to find a winning strategy. Everything the program did was according to the rules (ie the algorithm) that it was programmed for. Even the result of the weights and best choices for playing the game was a result of the program. That's what is meant by it being "intelligently designed all the way down", a term you apparently had no clue as to how to understand.Your logic makes no sense. You are implying that the results of any search application is the application itself generating information and/or learning to do some task on its own. But that is just not the case.For example, I could develop a computer program to take the addresses of the roughly 327 million people in the USA and then filter out those living in Arizona. Even before I gave it an initial data set (the 327m people), the algorithm is created. Now I feed the program the initial data set and it spits output of about 7.2 million names. By your logic it was the computer program itself that developed how to get a list of Arizona dwellers. That is silly The hell it didn't!!! Put simply, an algorithm is a set of rules to follow to solve some question. The program was basically told how to play chess, and how to make the "best" move in a particular situation. What is defined as "best" was not initially specified, but by further rules provided by the programmers, the program was given the ability to randomly and efficiently search though the chess-game-moves search space, apply weights to various actions, and filter options to get what would (according to the rules) be the most statistically winning results when playing. Put simply, the programmers told the program what would constitute as being able to "win" and the program went out and did it. The program did nothing on its own that wasn't defined prior by a programmer. Funny how you'll quibble about this and not even know the meaning of the words you are using previously - like algorithm. And since we're talking about information in a computer program, and not a biological thing, you're complaint is hollow. Then you end with a tautology? Pot, meet kettle.But if you want to still quibble, drop the word "complex" since you have no understanding of the meaning, use the title as is, and my statement still stands.Edited by WookieeB, : cleanup