|
|
Post by Cory Trese on Feb 13, 2017 16:05:59 GMT -5
Well, an incredible amount of time has been put into developing AI that play chess. The reason for this is that chess is a problem particularly well suited to an AI opponent. The nature of the rules and the strategy lends itself very well to using large lists of possible moves and advantage weights to plan moves.
The problem set being solved by a chess AI isn't the broadest or deepest out there, but it is very well understood and well researched. Are they the "smartest" game playing AI? I'm not sure about that, but they are the ones that have had the most R&D dollars spent on them!
|
|
poryg
Templar
[ Star Traders 2 Supporter ]
Posts: 1,723 
|
Post by poryg on Feb 13, 2017 18:07:56 GMT -5
Chess AI isn't smart in the sense of being smart. It has a position and calculates the best variation based on equations. Anybody can count equations  Computers are just better in it, because they are capable of so many operations per second and they dn't make mistakes. So no, chess AI isn't really the smartest... But it's really far developed, that's true. However... Still, chess computers lack human elements. They cannot always determine drawn positions... Or lost positions. Once there was even a case of a position where it was a clear mate in 2. The computer saw it... But completely forgot that it was a mate and calculated 20 moves further  But anyway, when computers begin searching for the most fighting move instead of the objectively best ones, that will be the day when we can talk about artificial intelligence in chess Just like when the world Go champion got finally beaten by a computer... This computer had a neural network, but... It still only worked with data. It chose the best moves by playing games against itself and where it won the most games, the move was the winner. |
|
|
|
Post by fallen on Feb 13, 2017 19:54:59 GMT -5
Hi. Since I am kind of uneducated in computer gamesLiar. Hehehehehehe, consider the on-going discussion on this thread, this is still cracking me up. |
|
|
|
Post by stratego on Feb 13, 2017 23:54:45 GMT -5
Well, an incredible amount of time has been put into developing AI that play chess. The reason for this is that chess is a problem particularly well suited to an AI opponent. The nature of the rules and the strategy lends itself very well to using large lists of possible moves and advantage weights to plan moves. The problem set being solved by a chess AI isn't the broadest or deepest out there, but it is very well understood and well researched. Are they the "smartest" game playing AI? I'm not sure about that, but they are the ones that have had the most R&D dollars spent on them! Well the R&D spent on them has made them the smartest AI newest chess AI can easily beat world champios in chess, I can't say that to any other AI currently made |
|
|
|
Post by wascalwywabbit on Feb 14, 2017 0:01:13 GMT -5
Hehehehehehe, consider the on-going discussion on this thread, this is still cracking me up. He's not lying, he just has high standards for what constitutes 'educated in computer games'. I mean I watch extra credits on YouTube, read yours and Cory's posts about game design and mechanics regularly, play a handful of digital word, puzzle and board games, plus one more serious computer game (HoS atm, obviously), read a book on game programming when I was 14, keep abreast of gamefromscratch.com etc., but aside from that I'm totally ignorant... Couldn't afford console games growing up (my brother bought a few though from his summer job money), played a few shareware and later commercial games when I got my first computer, but stopped playing all games for years after I got physically and mentally sick with the attendant depression. Only started again after tablets became a thing cause I can play them lying down ... sometimes I'm just way too tired and sore to sit up at my computer proper or even the couch to watch a show... So there's obviously gaps possible in all sorts of ways. Computer games is such a broad field of interest now, that if you haven't spent all the waking hours of a team of healthy persons' lives reading about and playing them since the 1960s, there's going to be gaps in knowledge for sure... 'I am wise in that I know I do not know.' -attributed to Socrates |
|
|
|
Post by stratego on Feb 14, 2017 0:01:13 GMT -5
Chess AI isn't smart in the sense of being smart. It has a position and calculates the best variation based on equations. Anybody can count equations Computers are just better in it, because they are capable of so many operations per second and they dn't make mistakes. So no, chess AI isn't really the smartest... But it's really far developed, that's true. However... Still, chess computers lack human elements. They cannot always determine drawn positions... Or lost positions. Once there was even a case of a position where it was a clear mate in 2. The computer saw it... But completely forgot that it was a mate and calculated 20 moves further But anyway, when computers begin searching for the most fighting move instead of the objectively best ones, that will be the day when we can talk about artificial intelligence in chess Just like when the world Go champion got finally beaten by a computer... This computer had a neural network, but... It still only worked with data. It chose the best moves by playing games against itself and where it won the most games, the move was the winner. Nobody makes calculating AI since Kasparov won that match. Nowadays they use AI that adjust to any situation not simply calculate it. |
|
|
|
Post by amongstshadows on Feb 14, 2017 3:30:06 GMT -5
I can direct everyone to a youtube video that shows how to make an AI that will ALWAYS win in tic-tac-toe. That's considered strategy, right?
|
|
poryg
Templar
[ Star Traders 2 Supporter ]
Posts: 1,723
|
Post by poryg on Feb 14, 2017 3:49:22 GMT -5
You cannot force a win in tic-tac-toe. So I don't believe you amongstshadows |
|
|
|
Post by wascalwywabbit on Feb 14, 2017 4:09:04 GMT -5
I can direct everyone to a youtube video that shows how to make an AI that will ALWAYS win in tic-tac-toe. That's considered strategy, right? Do you mean play a 'perfect' game? Two perfect game players of tic tac toe will always TIE, not have a winner... A perfect play in that game only wins if the other player makes a mistake. It's not that hard for an intelligent, attentive, human to play that game perfectly too ... even this dumb bunny has done it against a perfect play ai many times... |
|
|
|
Post by amongstshadows on Feb 14, 2017 4:10:09 GMT -5
You cannot force a win in tic-tac-toe. So I don't believe you amongstshadows Channel is called makinggameswithben. Will be in the advanced c++ tutorials playlist |
|
|
|
Post by grävling on Feb 14, 2017 6:41:31 GMT -5
Yeah, but if you are playing with a 5 by 5 board and with 3 players, I don't think you are playing what I call 'noughts and crosses' and is called 'tick-tack-toe' in other parts of the world.
|
|
poryg
Templar
[ Star Traders 2 Supporter ]
Posts: 1,723
|
Post by poryg on Feb 14, 2017 6:47:30 GMT -5
amongstshadows Thanks. But still, creating an AI for tic-tac-toe lacks a practical value. There are no more than 362,800 combinations in Tic-tac-toe. So even a 400 MHz computer would be able to calculate all possible combinations in less than 1 second. And computers nowadays have so much power that they don't even need an AI. Creating a database of all possible continuations is something even a beginner is able to do. Determining which lines are winning or losing isn't that hard either. Determining which variations can win should not be that hard And the todays' computers have so much power that they can play in a blink of an eye without intelligence.
|
|
poryg
Templar
[ Star Traders 2 Supporter ]
Posts: 1,723
|
Post by poryg on Feb 14, 2017 7:11:12 GMT -5
Chess AI isn't smart in the sense of being smart. It has a position and calculates the best variation based on equations. Anybody can count equations Computers are just better in it, because they are capable of so many operations per second and they dn't make mistakes. So no, chess AI isn't really the smartest... But it's really far developed, that's true. However... Still, chess computers lack human elements. They cannot always determine drawn positions... Or lost positions. Once there was even a case of a position where it was a clear mate in 2. The computer saw it... But completely forgot that it was a mate and calculated 20 moves further But anyway, when computers begin searching for the most fighting move instead of the objectively best ones, that will be the day when we can talk about artificial intelligence in chess Just like when the world Go champion got finally beaten by a computer... This computer had a neural network, but... It still only worked with data. It chose the best moves by playing games against itself and where it won the most games, the move was the winner. Nobody makes calculating AI since Kasparov won that match. Nowadays they use AI that adjust to any situation not simply calculate it. Actually, that's not necessarily true, and neither is your previous post. The computer AI takes the piece placement. Then through implemented methods it excludes the immediately bad moves (in other words moves that lose instantly). Then the computer starts calculating moves through the same method... And after 20 moves or so, it chooses the "best" move, in other words the most sensible line according to the algorithms given to it. True, the computers are able to beat grandmasters and world champion. But that is because they don't make mistakes and they calculate moves based on precise algorithms. Humans, on the other hand, aren't capable of millions operations per second. They aren't capable of processing the position so quickly. They use their knowledge too, but other than that they use intuition, feeling and skill in how they choose their moves. While the computers process things so quickly that they never miss anything... The humans often overlook even basic things. This is what makes computers better than humans. However, if you had a perfect human vs. a perfect computer... The human would win, because he has deeper understanting of the position than any computer will ever have. That's because every computer is limited by the algorithms. So computers have worse understanding of flanks for example. And this is where humans are better. We aren't bound by any algorithms and we are adaptable. We just don't have the computer-like capacity to look 20 moves deep every move. Also, depending on their algorithm, various computers have various uses. For example, the world champion among computers is Komodo. It is very strong and very resillient. That is because it is closer to human than any other computer and plays very positionally... However, it needs more time to calculate, which makes it unsuitable for blitz. On the other hand, the world blitz champion is Stockfish. It's something you would say to be of tactical nature in chess player terminology. However, Komodo vs. Stockfish, even though they are of similar playing strength, always depends on time control. Long games Komodo, due to it's focus on not losing... Short games Stockfish, due to it's ability to calculate fast and attack aggressively. |
|
|
|
Post by stratego on Feb 14, 2017 15:22:52 GMT -5
amongstshadows Thanks. But still, creating an AI for tic-tac-toe lacks a practical value. There are no more than 362,800 combinations in Tic-tac-toe. So even a 400 MHz computer would be able to calculate all possible combinations in less than 1 second. And computers nowadays have so much power that they don't even need an AI. Creating a database of all possible continuations is something even a beginner is able to do. Determining which lines are winning or losing isn't that hard either. Determining which variations can win should not be that hard And the todays' computers have so much power that they can play in a blink of an eye without intelligence. Wrong, it would take 2 full cycles to calculate that as there is OS too and we don't measure that in seconds |
|
sheff
Star Hero
[ Star Traders 2 Supporter ]
Posts: 503
|
Post by sheff on Feb 16, 2017 22:38:52 GMT -5
Hehehehehehe, consider the on-going discussion on this thread, this is still cracking me up. I think we're being ignored though, which makes it even funnier. |
|
Computers are just better in it, because they are capable of so many operations per second and they dn't make mistakes.
But anyway, when computers begin searching for the most fighting move instead of the objectively best ones, that will be the day when we can talk about artificial intelligence in chess 
