/sci/ - Science & Math

Do I think?

>>5717436
How about non-timed chess?

>>5717422
Never in speed chess, doesn't look good for the human. Chess is dead, so is Bobby Fischer.

>>5717440
You can easily google for this

>>5717422
>>5717440
Still no. Even if chess is never solved (which I seriously doubt), computers can calculate future paths quicker than any human and choose the optimal one. Whether speed chess or no, a computer playing at its best (ie, optimally) will always beat a human.

Oh cool black people can play chess, too!

>look a little closer
>that pawn structure
>hes mising two bishops, his queen, and his king

wat

>>5717464
>he's missing his king
My muffled guffaws could be heard down the block.

>>5717464
That's Usain Bolt playing checkers

>>5717422
that chance died, the day Kasparov lost the match against Deep Blue.

>>5717527
>lost
>implying there weren't human players in the IBM room helping the machine

>>5717464
>>5717467
Is it seriously not obvious to you that they're just playing checkers with a chess set?

It's a pretty common thing to do.

>>5717543
Suuuuure, anon. Suuuure.
And there are little men inside your computer that help your computer kick your ass at chess too

>>5717464
http://www.youtube.com/watch?v=zyGeR2mPnzY

>>5717570
Kasparov said it himself.
>Kasparov and the rest of the Chess Community were seriously suspecting human intervention in Game 2 due to this fact. It is very possible that there was a relatively strong human (not necessarily a grandmaster) watching the screen, and Deep Blue needed permission from this person to make a move. If the human noted that Deep Blue's contemplated move would place its king in danger, the human would drop the main log, and revert to the second-best, safer move. It is effectively impossible for a computer, especially one like Deep Blue, to miss a perpetual check, like Deep Blue did in Game 2. However, it is rather plausible that a human might, especially when playing under pressure (as was Kasparov).

>>5717572
>The guy who lost claimed the other guy cheated
Possibly the least reliable person to ask

Computers are as good or better than us right now. They have much more room for improvement than us. The answer's no, and I'd say it's a pretty dumb question.

Just write a really poor chess AI.
The real question is whether we'll be able to beat a GOOD chess program.

>>5717572
That line of reasoning doesn't make sense. The theory they came up with says that Deep Blue must either make the best or the second best available move. However, it is said that the perpetual check outcome should have been obvious to Deep Blue. Therefore any move which had any probability of ending in a win for Deep Blue ought to have been ranked higher than the perpetual check. Only if there was exactly one move which had even a possibility of ending in a Deep Blue victory does their theory make any sense, and that is unlikely.

[looking up the game]

It appears Deep Blue had three alternatives, not one. Chess grandmasters confirmed for sore losers.

>>5717601
You know what he meant you pedantic shitface
To the OP: Not without cognitive augmentations.

>>5717543
Even if that were true, he would not stand a chance against modern computers in his prime.

This thread is now about board games where a computer will NEVER beat a human pro

Pic related

>>5718182

What about a team of grandmasters against a good Chess AI program? I know Kasparov did some kind of "Kasporov vs. The World", with people from around the world using the internet to connect. But that was kind of bogus since plenty of people don't know whats up and their votes were probably just messing up the game.
What I'm talking about is the top 10 in the world get together in some conference room(in person) and actually play the computer in real time....

>>5718190
Never is a long time...

>>5718190
Not a comp sci student, but Go looks like a game that can be logically 'solved' in the future, maybe not completely but at least partially like chess so comps can match/outdo humans in various cases.

>>5718192

I'm pretty sure they could beat the best Chess program around right now!!!

>>5718192
The world game had several grandmasters/prodigies guiding the team - it wasn't your street bum facing Kasparov.
And that would be nice, top 10 chess players playing the best chess comp irl. Sadly it won't happen.

>>5718190
Arimaa
http://en.wikipedia.org/wiki/Arimaa

>>5718201

I'm not so sure, Go moves are very subjective and moves that seem good at the time may be meaningless a couple turns later; not that it's different from chess but the sheer amount of pieces and turns in Go would be incredibly difficult to calculate efficiently. Additionally, unlike chess, the one with more pieces doesn't equal an advantage, like chess. I'm very curious to see if it is "solvable" though

>>5718216

Oops, didn't mean to type that "like chess" part in the 2nd last sentence

>>5718190
Keep in mind that modern computers are amazingly "dumb" compared to humans. If a computer was limited to the computational resources of a human brain, good chess players would murder it. Bottom line: computers are vastly superior when it comes to hardware; humans are vastly superior when it comes to "software", roughly speaking. What we call "AI" is still a total joke. All we have is the illusion of intelligence created by brute-force search.

Brute-force search is much more problematic in Go than in chess, but if/when actual AI comes along, computers will trounce us. And then maybe take over the world.

>>5718190
>>5718201
>>5718216
>>5718218

Can't you fuckers even do some research?

4x4 and 5x5 Go is solved already.
The only thing needed for solving 6x6, 7x7, ... ,19x19 is computer power (which will eventually increase enough to do so)

>>5718221
>What we call "AI" is still a total joke. All we have is the illusion of intelligence created by brute-force search.

If this is the state of AI in
>2013
then perhaps it was possible that Kasparov was playing another human or humans during the Deep Blue match. Might have been a group of Grandmasters who had the assistance of the "brute force" of a computer.

>>5718225
>4x4 go

>>5718229

I honestly believe that Garry Kasparov was cheated out of his win when he faced Deep Blue. Too much money and too many people worked on that project for IBM to just let Kasparov win.
Shortly after Kasparov lost, IBM stocks boomed.
The only reason Kasparov decides to speak out about it and call it out, is that he takes too much pride in his game.
Russians take playing chess too seriously and Chess players in general do too. just look at Bobby Fischer.

I bet Usain Bolt is better at Chess than anyone here. Probably faster than all of you and with a bigger knob as well.

>>5718190
I think I could beat my laptop at mouse trap

>>5718229
Perhaps? PERHAPS?!?

I HAVE DOCUMENTS PROVING IT

>>5718255
I would win so hard over my laptop at twister

>>5718258

I happen to be a big fan of Mr. Alexander Emrick Jones, I listen to his show everyday from 9-12 Pacific. Don't see why he's relevant though?

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

Explain to me how he was able to win the first match?
Even in the second match, he won a game.
He is quoted as saying, it "played unlike a computer"

>>5718251
Even if you're right, it's not really relevant anymore. Computers routinely beat grandmasters. It started with Kasparov-Deep Blue (assuming no cheating) and has continued ever since.

>>5718190
Finite state space, trivially solved. Step up, plebs.

Pic related, computers will never be creative enough to keep up in ESPORTS..

Chess player here just saying this now: Just because you are good at math/science whatever does not make you able to discuss chess in a non-retarded manner. Right now it's like you watching high school fedora wearers debating string theory because of some show they watched on discovery. Goodnight.

>>5718225

>implying the next board sizes up dont exponentially increase in possible moves in the next couple turns

>>5718221
This guy's got it, we need "actual" AI, which may or may not be feasible to create artificially, in order to win at something like Go consistently; Chess is too "easy" with brute force searching

>>5718763
*difficulty

Yes if I set the difficult to easy.

>>5718190
No computer will ever win a game that deals primarily with the meta-game of player interaction. Werewolves, Avalon, and Diplomacy are the first three I think of although there are many more.

>>5718772
>Avalon

I was thinking of the expansion, this should be The Resistance

>>5718772
>>5718775
And since even with all the geeks here, I bet few of you actually play board games, a brief overview:

Werewolves - Each player is secretly assigned a role - Werewolf, Villager, or Seer (a special Villager). There is also a Moderator who controls the flow of the game.

The game alternates between night and day phases. At night, the Werewolves secretly choose a Villager to kill. Also, the Seer (if still alive) asks whether another player is a Werewolf or not. During the day, the Villager who was killed is revealed and is out of the game. The remaining Villagers then vote on the player they suspect is a Werewolf. That player reveals his/her role and is out of the game.

Werewolves win when there are an equal number of Villagers and Werewolves. Villagers win when they have killed all Werewolves.

>>5718779
The Resistance - Players are either Resistance Operatives or Imperial Spies. For three to five rounds, they must depend on each other to carry out missions against the Empire. At the same time, they must try to deduce the other players’ identities and gain their trust. Each round begins with discussion. When ready, the Leader entrusts sets of Plans to a certain number of players (possibly including himself/herself). Everyone votes on whether or not to approve the assignment. Once an assignment passes, the chosen players secretly decide to Support or Sabotage the mission. Based on the results, the mission succeeds (Resistance win) or fails (Empire win). When a team wins three missions, they have won the game.

>>5718780
Diplomacy - In the game, players represent one of the seven "Great Powers of Europe" (Great Britain, France, Austria, Germany, Italy, Russia or Turkey) in the years prior to World War I. Play begins in the Spring of 1901, and players make both Spring and Autumn moves each year. There are only two kinds of military units: armies and fleets. On any given turn, each of your military units has limited options: they can move into an adjoining territory, support an allied unit in an attack on an adjoining territory, support an allied unit in defending an adjoining territory, or hold their position. Players instruct each of their units by writing a set of "orders." The outcome of each turn is determined by the rules of the game. There are no dice rolls or other elements of chance. With its incredibly simplistic movement mechanics fused to a significant negotiation element, this system is highly respected by many a gamer.

This summary doesn't do the negotiation part of this game justice. This game is ENTIRELY about convincing others to do what best helps you, that everyone else is betraying them and you will help them (until you betray them).

Is that Norwegian kid better than Kasparov or Fischer yet? His ELO is higher but I suppose that doesnt make him better per se?

>>5718759
>may or not be feasible to create artificially
as if something incredibly dumber didn't create intelligence in the first place.
Intelligence is a much better selection process than evolution; evolution just had a massive time advantage in AI research
It still did a pisspoor job

>>5718796

Why isn't this guy studying at a university?!

>>5718796
He drew Kasparov at 13. What do you think?

>>5718833
Because chess ability is poorly correlated with intelligence.

>>5718799
Did it?

>>5718861
He's a moron, don't listen to him.

>>5718849

Well that just isn't true, Kasparov has an IQ of 190.
Chess people consistently rank as someof the people with highest IQs.

>>5718845

>spectators at a chess game

that must be absolutely riveting to watch.

i hope it's at least speed chess.

>>5718898

>>5718474
I know it wouldn't make them win, but imagine the micro...

>>5717572
>yfw chessmaster gets bested by your average everyday scientist
Only on /sci/fi
>>>/x/

>>5718890
His IQ is actually given as 135 as professionally measured to confirm or deny unsupported suggestions of an IQ between 185 and 190.

>>5718890
>ipmlying IQ and intelligence are the same thing.

>>5717422

lol, nigga captured his own king


As for the answer of dumb question:

http://www.chessgames.com/perl/chessgame?gid=1497429

And that's a minute game against (at the time) strongest computer program. The thing with computers is that they are highly predictable. Give a player access to program version and he'll have it figured so much that he'll embarrass the hell out of it.

>>5719114
Lol'd at game.

>>5717422

and he has his board upside down

>>5719114
>computers are predictable

not really, if you make them choose the move they calculated as best, they are, if you make them randomly choose from the top 3 or set up a really simply bluff/trap algorithm to try to bait the enemy champion running on top of the top 5 choices for your move, it will stop being predictable and start being nasty.

>>5719148
>so many conjectures

>>5718845

http://www.youtube.com/watch?v=zlfbpC3dQWw

>nigger playing chess

my fucking sides

>>5719174
That kid is good.

>>5719185
>nigger
That's Usain Bolt you dolt. He achieved more than you ever will in your life.

>>5719199
i guess if you call running slightly faster than a house cat an achievement

>>5719199

>Running fast in a straight line

Amazing achievement, great contribution to the human race. Give this shit skin a Nobel Prize already.

>>5719205
>>5719207
>His achievements in sprinting have earned him the media nickname "Lightning Bolt",[15] and awards including the IAAF World Athlete of the Year, Track & Field Athlete of the Year, and Laureus Sportsman of the Year (three times). He is the highest paid athlete ever in track and field.[16] He has been called the world’s most marketable athlete[17] and the greatest athlete ever.[18][19][20]
Yeah, I wonder what you two have achieved so far.

>>5718225
not at all.
Go on small board is radically different from 19x19 and would need a completly new algorithm.
Go is heavily based on pattern recognition and intuition, something computers sucks at, now and for at least the next century.

>>5719210
wow the world would truly not be the same place without him, u got me

>>5719217
How is being 12 working out for ya?

>>5719114
>>chessgames.com

Why don't those fuckers announce which player is which color? That's annoying as fuck.

>>5719246
I didn't know either. It's obvious from the endgame that black is human.

>>5719248
Yeah, obviously. White's ridiculous opening (I haven't followed chess in a long time, and when I did I didn't stumble upon it) also hinted at that. I'd have preferred to know before playing through the game, though.

>>5718474
>>5718970

Some guy made an AI that had over 20 000 APM (actions per minute), I think it was somewhere between 25k and 50k APM. In comparison, most progamers lie in the 200-300 area, with some dipping to as low as 120 and some going as high as 400-500.

If we (humanity) wanted to, we'd create a SC2 AI that could never, ever be beaten by any human.

>>5718190
Eh, we've got bots using Monte Carlo evaluation now that have reached 6d, which is higher than the vast majority of amateurs will ever get.

The problem with go AIs is that there's no incentive to spend a lot of resources making a really strong AI, like there was for chess. Don't misunderstand, I'm not saying it isn't much harder to program a strong go AI than it is to program a strong chess AI, because it is.

>>5717422
>Do you think humans will ever be able to beat computers in chess?

Yes, but only in the past.

>>5719270
>>Some guy made an AI that had over 20 000 APM (actions per minute), I think it was somewhere between 25k and 50k APM. In comparison, most progamers lie in the 200-300 area, with some dipping to as low as 120 and some going as high as 400-500.

>>If we (humanity) wanted to, we'd create a SC2 AI that could never, ever be beaten by any human.

I've actually worked on SC:BW AI and the problem is never getting the correct micro down, it's trivial for an AI to control units way better than a human ever could (my AI focused on microing reavers and shuttles in combat for example). The problem is that the overarching strategy of the game is extremely hard to get right, and the game will probably never be practically solved due to being realtime rather than turn-based.

Less strategy and more execution and the game becomes trivial for a computer, you can make bots that stomp the shit out of people because of perfect reactiontime and aim in an FPS, but good RTS AI doesn't exist yet.

>>5719246
>>5719248

In chess white is always written first. So Rybka (Computer) vs Hikaru Nakamura. Means Rybka white, Nkamura black.

>>5718208
>http://en.wikipedia.org/wiki/Arimaa
Trying to imagine some arriima chess
1-White player select where all his pieces will when the board start, putting them on first and second row (pawns has double move on last row).
2-Then black player choose where his pieces will be on his first and second row, he can not select a position that will make his king be into check.
3-Castling is possible if king not moved and is on d or e file and the rook is at corner at the same rank.

the 4 moves it would not be possible because would make player be able to win on the first turn

>>5719270
>If we (humanity) wanted to, we'd create a SC2 AI that could never, ever be beaten by any human.
sc2 is too complex, the amount of stuff you can do per turn (yes real time games have turns) is extreme and the amount of turns per second is extreme

>>5719341
You don't need to completely solve the game, just play it better than a korean zergbot.

>>5719341
SC2 is strategically shallow (i.e. severely limited when it comes to strategical branching), which means it's "just" an issue of implementing satisfactory build, strategy, scouting and tactical algorithms.

I briefly considered which race would hold the advantage under optimal play, and figured it'd be Protoss due to the longevity of the units, but it would obviously also be map-dependent.

I'd love to see like 15 Warp Prisms lifting individual zealots and stuff, now that would be awesome (T here).

>>5719537

You'd be surprised how much metagame there is. It's not 'just' an issue of implementing good strategies because strategies change with updates and players finding their way around them. You have to learn new tactics every four months or so to keep up, assuming you're not in High Diamond or higher. Protoss is also the weakest race in terms of the players because Terran has far stronger units and is more efficient.

Protoss player here.

>>5719548
>You'd be surprised how much metagame there is. It's not 'just' an issue of implementing good strategies because strategies change with updates and players finding their way around them. You have to learn new tactics every four months or so to keep up, assuming you're not in High Diamond or higher.

Meh. Eventually everything will be cookie-cutter, just like in BW and WoL.

>Protoss is also the weakest race in terms of the players because Terran has far stronger units and is more efficient.

I don't understand what you mean by "in terms of the players". It is, however, self-evident to me that it is easier to save P units than it is to save Z and T units by the virtue of the high HP (and shields) of P units as compared to the other races.