Here's Hoping They Don't Break Chess Next...

[Matthew Nicholson]

Tl;dw - A new program called AlphaGo Zero needed 3ish days to train itself to become the strongest Go program in the world. Pretty sure chess is on a similar level of complexity as Go... yikes.

https://www.youtube.com/watch?v=tXlM99xPQC8

[Wayne Komer]

Re: Here's Hoping They Don't Break Chess Next...

Here's Hoping They Don't Break Chess Next..

October 25, 2017

AlphaGo Zero was discussed in a forum thread earlier this month at:

http://forum.chesstalk.com/showthread.php?11532-Go-Computers-and-Chess&p=118781&viewfull=1#post118781

[Kerry Liles]

Re: Here's Hoping They Don't Break Chess Next...

Tl;dw - A new program called AlphaGo Zero needed 3ish days to train itself to become the strongest Go program in the world. Pretty sure chess is on a similar level of complexity as Go... yikes.

https://www.youtube.com/watch?v=tXlM99xPQC8

I believe you will find that GO is considerably more complex than chess which makes the achievement even more remarkable.

[Garland Best]

Re: Here's Hoping They Don't Break Chess Next...

Could the method used for AlphaGo Zero be used to come up with a program that could beat Komodo, Houdini, etc in the same amount of time and effort?

[Matthew Nicholson]

Re: Here's Hoping They Don't Break Chess Next...

Could the method used for AlphaGo Zero be used to come up with a program that could beat Komodo, Houdini, etc in the same amount of time and effort?

That's what's so scary! Imagine finding out three days from now that chess is effectively solved!

[Vadim Tsypin]

Re: Here's Hoping They Don't Break Chess Next...

That's what's so scary! Imagine finding out three days from now that chess is effectively solved!

Yep... And within the other three days, AlphaGo Zero will set its sights on us, humans. As the article referred to in the precious thread informs rather casually,

Machine-learning research from Google and others has also shown promise for extracting more ad dollars from consumers.

- the word "ad" will be stricken from there soon.

[Christopher Mallon]

Re: Here's Hoping They Don't Break Chess Next...

Yep... And within the other three days, AlphaGo Zero will set its sights on us, humans. As the article referred to in the precious thread informs rather casually, - the word "ad" will be stricken from there soon.

This was predicted (I'm sure in other places as well) in the Ender's Game universe where an AI takes over managing Ender's money and almost immediately does a better job of it than the humans could have just by examining past behaviour...

[Ed Seedhouse]

Re: Here's Hoping They Don't Break Chess Next...

I believe you will find that GO is considerably more complex than chess which makes the achievement even more remarkable.

"Considerably" is an understatement, and a really big understatement. "Vastly more complex" would be perhaps slightly more accurate.

[Ed Seedhouse]

Re: Here's Hoping They Don't Break Chess Next...

That's what's so scary! Imagine finding out three days from now that chess is effectively solved!

Not going to happen anytime soon. Computers will be many orders of magnitude better than humans at either game long before either is "solved". To merely *record* a solution to chess, the less complex of the two games, would require the use of every atom in the visible universe if we reach the point where each atom can store one bit, which we are not even remotely close to at the moment.

[Matthew Nicholson]

Re: Here's Hoping They Don't Break Chess Next...

Not going to happen anytime soon. Computers will be many orders of magnitude better than humans at either game long before either is "solved". To merely *record* a solution to chess, the less complex of the two games, would require the use of every atom in the visible universe if we reach the point where each atom can store one bit, which we are not even remotely close to at the moment.

Effectively = actually but not officially or explicitly

[Ed Seedhouse]

Re: Here's Hoping They Don't Break Chess Next...

Effectively = actually but not officially or explicitly

Weasel words don't change the facts. Chess will not be "effectively" solved, even in your (incorrect) sense until no one can write a computer program or create hardware that beats the previous best. Using words in the same (silly) way you do I can say that motorcycles "effectively" run at infinite velocity.

[Tom O'Donnell]

Re: Here's Hoping They Don't Break Chess Next...

That's what's so scary! Imagine finding out three days from now that chess is effectively solved!

Even if chess were completely solved someday it wouldn't change much for humans. Not like we would be able to understand or memorize the precise series of moves in all variations that guarantee the win/draw.

I think what could be interesting is to see how long it would take for a computer programmed only with the most basic rules, and playing only games vs itself, to discover many mainline openings, and the order in which it would discover them.

[Matthew Nicholson]

Re: Here's Hoping They Don't Break Chess Next...

Weasel words don't change the facts. Chess will not be "effectively" solved, even in your (incorrect) sense until no one can write a computer program or create hardware that beats the previous best. Using words in the same (silly) way you do I can say that motorcycles "effectively" run at infinite velocity.

Lol "Weasel words" aka Adverbs.

Anyway... Chess basically can't be mathematically solved, but if a computer were to be developed that could beat every other computer on a consistent basis while drawing against an identical version of itself 100% of the time, we've "effectively" solved the game.

I know insulting people is a past time for many chesstalkers, but I'm making a good point here. Computers are already taking some of the 'magic' out of chess, but a self learning AI that supersedes any existing calculation based software would be a huge blow.

[Ed Seedhouse]

Re: Here's Hoping They Don't Break Chess Next...

Lol "Weasel words" aka Adverbs.

Nope, merely an accurate description of what you wrote. Look up the definition.

Anyway... Chess basically can't be mathematically solved

Actually it's easy to write a mathematical function that will solve chess. In practice the solution will take about forever to complete, but that's a practical difficulty, not a mathematical one. Chess is mathematically solvable and the mathematical solution is known. The rest is a boring practical problem that isn't of interest to mathematicians.

but if a computer were to be developed that could beat every other computer on a consistent basis while drawing against an identical version of itself 100% of the time, we've "effectively" solved the game.

But there is no evidence that this will ever happen. Even if it did it's easy to write a program that will always draw with itself so that last criterion is beside the point.

I know insulting people is a past time for many chesstalkers

I didn't insult you, if anything I "insulted" your use of language, but your use of language is not "you".

but I'm making a good point here.

Calling it a good point doesn't make it one. I think you used sloppy and confusing language to make vague and content-less "point" myself. I don't know if that's a normal thing for you.

Computers are already taking some of the 'magic' out of chess

For who? Not for me. And computers have made it possible for more people to take part in competitive chess than ever before. Computers run the internet, have you noticed?

but a self learning AI that supersedes any existing calculation based software would be a huge blow.

Yeah, in just the same way that motorcycles have resulted in all human foot races dying out because they are pointless. Oh.

[Mathieu Cloutier]

Re: Here's Hoping They Don't Break Chess Next...

Actually it's easy to write a mathematical function that will solve chess. In practice the solution will take about forever to complete, but that's a practical difficulty, not a mathematical one. Chess is mathematically solvable and the mathematical solution is known. The rest is a boring practical problem that isn't of interest to mathematicians.

Yeah, sure. But when your 'practical difficulty' becomes something theoretically impossible, I say you have kind of a problem... mathematical or not.