GatewayRedbirds.com

A Message Board Dedicated to Discussing St. Louis Cardinals Baseball!
It is currently January 20 19, 6:59 am

All times are UTC - 6 hours




Post new topic Reply to topic  [ 14 posts ]  Go to page Previous  1, 2
Author Message
PostPosted: December 27 18, 8:03 am 
Offline
"I could totally eat a person if it were a life/death situation"
User avatar

Joined: August 5 08, 11:24 am
Posts: 22731
Location: Thinking of the Children
Also, not to be lost in my chess rambling, is how DeepMind performed in the game Go which has many (exponentially) more possible positions than chess although to Schlich's point it is still well-defined inputs and outputs with simple rules. But, this was a game that many thought computers wouldn't be able to master due to the larger number of possibilities.

After learning to play the game for 30 some odd hours against itself, it beat the top player in the world in I think 5 of 6 games. In the one game it lost, it seemingly got confused and started spitting out random/non-sense moves. Then, in the next game, it was spitting out moves that the developers were very concerned were non-sense, but in the end it won the game by a narrow margin.

So, yeah, it wasn't as dominant at Go as it was at Chess, but Go is a more complex game and DeepMind managed to become better than any human in a relatively minuscule amount of time.


Top
 Profile  
 
PostPosted: December 27 18, 8:09 am 
Offline
"I could totally eat a person if it were a life/death situation"
User avatar

Joined: August 5 08, 11:24 am
Posts: 22731
Location: Thinking of the Children
MAGA wrote:
Great post AW. It’s why I really think Chess960 may offer a future for human players and removes so much automaton like play.

Again, not that I know much, but I always figured if Chess960 was adopted as the preferred game, after a period of time, it would get to the point that Chess is today where with any given set of pieces, players would be able to memorize openings and go from there. As it is today, there's only one setup but the openings have become standard theory. Given time, maybe a lot of time, standard theory would be obtained for all of the 960 set ups and of those 960, while there would be nuances, a lot of the setups would be similar.

That said, and I don't think I'm alone when I say this, I much prefer watching old games between the likes of Tal, Ivanchuk, Fischer, Nezhmetdinov, Kasparov, Alekhine, etc where they were figuring almost everything out over the board and taking risks that escape the best players today largely because most risks have been assessed and figured out.


Top
 Profile  
 
PostPosted: December 27 18, 8:59 pm 
Offline
Hall Of Famer
User avatar

Joined: July 15 08, 8:24 pm
Posts: 18541
Location: Low A Minors
This whole thing creeps the hell out of me. First off, and probably least important, is that this kind of thing just kind of eliminates any "art" to playing chess, or anything else - baseball too. This research implies that there may be One Best Way to play chess and all others are inferior. This may be correct, but it's boring as hell. It kind of reminds me of this Harper's article I read recently, which argues that the art of writing novels is even being reduced to a "conservatory" art - like orchestra. Meaning all the important stuff has already been done, and the important thing is to just preserve it. The best you can ever hope to accomplish is to mimic one of the old masters (or in the case of chess, mimic the best AI).

To Schlich's point about how this kind of AI really only works for closed systems with well-defined rules....my fear is it will be too tempting to apply this kind of AI on those messier problems, anyway, and there will be all kinds of unintended consequences that we won't understand until it's too late. It's probably already happening, like in the stock market for example. Companies are using algorithms to control their buy/sell decisions and we've already had massive, instantaneous drops in the market. The stock market is complicated, but even it is not as complicated as like, cancer or weather or quantum physics or whatever.

Imagine a world where every product, song, film, sport, ad campaign, news article, etc has been designed to meet some ultimate "Best Way" according to some AI. I think it would be boring as hell and we would only get a glimpse of the "code" every so often when there is some spectacular failure. We are already far down this road. Everything is just a risk to be eliminated. Every movie nowadays is a superhero movie sequel, songs are written by a small group of Scandinavians, and all ballplayers will be Paul DeJong clones except for the most elite. Obviously, I think there would be a backlash against this optimization of everything. But then what if the AI is smart enough to foresee that backlash and account for it? After all, backlash is a big risk that must be eliminated. We're gonna live in The Matrix guys.


Top
 Profile  
 
PostPosted: December 28 18, 9:36 am 
Offline
"I could totally eat a person if it were a life/death situation"
User avatar

Joined: August 5 08, 11:24 am
Posts: 22731
Location: Thinking of the Children
Kind of following along with the chess thing, because really that's what I understand best, it's not so much that there is a right/wrong way to play. And, no one that I know of is saying that DeepMind is the end all be all master, rather at this point in time it is the best 'player' human or computer in the world. The chess community, I think, has started to take notice that the Stockfish/engine method is being refuted every time DeepMind plays a move that is inconsistent with the materialistic/incremental bludgeoning to death of opponents. Now, will a human ever be able to replicate the computing power that either engines or AI have? Obviously not. They're computing millions of positions in seconds. That's far beyond the capability of the human brain.

What I find alarming is how rapidly AI systems progress. Like I noted in my previous post, human created chess engines have been under development since the 60s or 70s. And, even given superior computing ability to DeepMind, DeepMind is still superior and became superior in 4 hours. Or to put it another way, an AI chess playing program took 4 hours to surpass decades worth of human programming efforts. That's absolutely terrifying.

If an AI program like DeepMind was turned against security for banks, governments, etc, how long would it take a program to hack into any of those systems?


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 14 posts ]  Go to page Previous  1, 2

All times are UTC - 6 hours


Who is online

Users browsing this forum: No registered users and 22 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum

Jump to:  
Powered by phpBB® Forum Software © phpBB Group