Hard Light Productions Forums
Hosted Projects - FS2 Required => Blue Planet => Topic started by: GhylTarvoke on October 07, 2015, 05:17:43 pm
-
... or, to make the Blue Planet connection clear, the optimal strategy (http://timkr.home.xs4all.nl/chess2/diary_3.htm) may seem completely alien. (http://timkr.home.xs4all.nl/chess/perfect.htm) I know of no better real-life example than endgame tablebases (https://en.wikipedia.org/wiki/Endgame_tablebase) in chess.
Computers have completely solved 7-man endgames. In other words, given any position with 7 men, these computers will output your best move(s) - the most direct route to checkmate if you're winning, the most delayed route if you're losing, and a move that maintains equilibrium if the position is equal.
These databases were (of course) compiled through brute force. Even with advanced compression, the amount of data involved is mind-boggling. The 5-man database requires 7 GB, the 6-man database requires 1.2 TB, and the 7-man database requires 140 TB. In some positions, the shortest checkmate is in 200+ moves. Suffice it to say that these endgames are beyond human comprehension.
The most fascinating part is how strange the best move can be: in addition to being inexplicable, it can appear utterly nonsensical. This is a realm where human heuristics are useless. From Tim Krabbé:
A grandmaster wouldn't be better at these endgames than someone who had learned chess yesterday. It's a sort of chess that has nothing to do with chess, a chess that we could never have imagined without computers. The Stiller moves are awesome, almost scary, because you know they are the truth, God's Algorithm - it's like being revealed the Meaning of Life, but you don't understand a word.
-
Fascinating! And I agree, very topical.
-
If you wanna beat God at this point you're gonna need one hell of an opening novelty.
Shambala Gambit
1. e4 e5
2. NfU (throw piece at opponent)?? 0-0-0-0 (knock over table)
3. L 8===> (low blow)
1-0
-
The question begged, of course, is how many turns it would take for the computer to reach that endgame table. :shaking:
-
I don't even know if modern chess engines need opening book to crush the top human players anymore, so endgame tables are really just silicon showboating.
-
The point is that even modern chess engines, despite beating top GM with good cpu machines, they still rely on human coded heuristics. You might not beat it, but you still understand it.
This is beyond understanding. These moves completely fail any heuristics test. Anyone performing these kinds of moves live would be laughed in any tournament. Even Komodo or Stockfish would laugh at them if they had been programmed to be able to do so. And the comedy would slowly fade away, for it would be increasingly clear that the player would be getting slowly towards advantage. But without anyone understanding why.
This is truly lovecraftian intellectual territory.
-
I guess I'm just not seeing what's strange here. Given how complicated the positions are even with a handful of pieces, wouldn't it be more surprising if there weren't bizarro mates in 200? Brute force regularly finds resources that are way beyond human depth, and by removing the artificial start and endpoints of needing to evaluate the position, the brute is unleashed in its natural habitat.
These endgames are perfect scenarios for showing off those deep resources and yeah this is way deeper than an engine running on a PC will go, but I don't see anything that isn't just a logical extrapolation of more number crunch = more better at chess. Isn't this exactly what we should expect?
Something weird would be like if with certain combinations of pieces God started playing crazy romantic era chess or otherwise doing something not in keeping with being boring and invincible.
-
It's also interesting that nobody in BP uses this kind of strategy, despite certain entities seemingly having enough computational brute force to do so. Both human factions try to do so but they don't have the resources yet.
-
It's also interesting that nobody in BP uses this kind of strategy, despite certain entities seemingly having enough computational brute force to do so. Both human factions try to do so but they don't have the resources yet.
It'd be a lot harder to brute-force endgames against an opponent that kept changing games.
-
It's also interesting that nobody in BP uses this kind of strategy, despite certain entities seemingly having enough computational brute force to do so. Both human factions try to do so but they don't have the resources yet.
It'd be a lot harder to brute-force endgames against an opponent that kept changing games.
And then there's also the fact that we don't actually know that none of the involved parties haven't computationally bruteforced a solution to at least some endgame.
-
I guess I'm just not seeing what's strange here. Given how complicated the positions are even with a handful of pieces, wouldn't it be more surprising if there weren't bizarro mates in 200? Brute force regularly finds resources that are way beyond human depth, and by removing the artificial start and endpoints of needing to evaluate the position, the brute is unleashed in its natural habitat.
These endgames are perfect scenarios for showing off those deep resources and yeah this is way deeper than an engine running on a PC will go, but I don't see anything that isn't just a logical extrapolation of more number crunch = more better at chess. Isn't this exactly what we should expect?
Something weird would be like if with certain combinations of pieces God started playing crazy romantic era chess or otherwise doing something not in keeping with being boring and invincible.
It's strange because most of those moves run counter to every heuristic that grandmaster chess players have developed through their careers. It's just not human. To repeat myself, it even goes against machine-level heuristics that chess engines have. Chess engines that display ELOs of way over 3300 with simple PCs I might add. These things just seem irrational by themselves. Only by looking at the bigger (200 moves) picture will they perhaps start to make sense, however it's a kind of sense that is beyond human comprehension.
The analogy would be if an army would start to behave in irrational terms, going back and forward, making strange perpendicular moves, seemingly stupid decisions that would compromise local positions, mind****ing every local commander that would mistake the opponent for a very stupid, callous adversary. And then, slowly, it would bizarrely advance on you, gaining ground, despite all the terrible local decisions that you see them doing. And despite all the best decisions you think you're making, you're still losing ground. And you're wondering, what the hell is going on? How come am I losing to this irrational opponent? And the Lovecraftian aspect is, you won't even ever understand how you lost. You just know you did.
It's also interesting that nobody in BP uses this kind of strategy, despite certain entities seemingly having enough computational brute force to do so. Both human factions try to do so but they don't have the resources yet.
It'd be a lot harder to brute-force endgames against an opponent that kept changing games.
I have to agree with Admiral and I question Battuta's phrase there. The game might "change", but the board is the same. An opponent that keeps "changing" their objectives within a game of chess will lose faster than one that doesn't. Every "changing of the game" itself are merely tools of the grander game. It's still just a tactic. Even what grandmasters call "strategies" are just higher level scale tactics of a sort.
I know, I know, "But the point is that they are changing the game itself", which points to unforeseen territory even for godlike species like the shivans or the vishnans. It means they themselves are within new "endgame" territory. Which means, they are unaware of the future. Which is a weird thing for a species that is beyond "time" but hey DS9 also made that mistake so who cares.
-
Luis described the strangeness better than I can. In the OP, I said that I could think of no better real-life example. Actually, I can think of no other real-life example, though I'm sure computers will eventually find more.
Now that I think about it - most of you probably beat me to the punch - the strangeness must be a property of chess itself. It has nothing to do with computers, because the "perfect moves" have always existed.
-
I think it's a property of every system that is sufficiently complex. A richer game than chess (with more varied pieces, bigger boards, etc.) will have "brute forced endgame" solutions much weirder than these, I believe / infer, etc.
"Real life" problems are much stranger though, for you cannot reliably compute every single atom's future path within an entire region of the galaxy. No matter how amazing shivan computation power is, "Force brute" calculation is just silly. Less silly is the idea that they behave like weirdly designed "chess engines", which have their own heuristics. IOW, shivans and vishnans also have heuristics, they are not "God". But, from the human point of view, these calculations might be so good (the chess equivalent of a player of 50 000 Elo, or something that astronomical - note: a 200 Elo difference means that one will have 76% likelihood of winning the other, no matter how high or low the Elo's are - hat they would be effectively as "godlike" as this brute force demo we have here.
It would seem irrational too. It would have heuristics that would be far beyond human comprehension. And it would obviously always defeat you.
-
I know, I know, "But the point is that they are changing the game itself", which points to unforeseen territory even for godlike species like the shivans or the vishnans. It means they themselves are within new "endgame" territory. Which means, they are unaware of the future. Which is a weird thing for a species that is beyond "time" but hey DS9 also made that mistake so who cares.
Deva
-
Well that's a relief to know. I'm with you, for this meme of "species beyond time" to coexist and solve "problems" with us "time bound" species should completely die out. It's just contradictory stuff.
-
I was under the impression they could "see the future" to a highly limited extent.
-
The Vishnans seem to be able to see the future very capably, but only within a conspicuously limited scope.
-
The present branches into infinite potential futures, just as infinite pasts converge into a present.
My hypothesis is that Vishnans (and probably Shivans) can break down the borders between adjacent "presents" if they are close enough to each other, but connecting a present to a future or to a past is a bit more difficult (or straight up impossible). And there are probably limits - the more "similar" the two presents are, the less effort it takes to connect two jump nodes in different realities.
So here's the question: Do Vishnans have precognitive abilities... or do they pick and choose the realities that proceed according to the scenario they have in mind, and simply migrate to the one that provides the best platform for accomplishing their goals?
Why are they silent? No one has heard a whisper in more than a year. What does it mean?
DISCLAIMER: Not to be considered Word Of God, just idle speculation from my part...
-
@ Luis
I'm just not feeling the sense of strangeness that you're describing when I look at these examples. Of course we can't calculate out the reasoning behind many of the moves past a few steps back from checkmate, we suck at chess, but there's nothing here that's out of keeping with what we would expect, is there? Shouldn't it be completely natural that the most accurate play will include a bunch of artificial computery looking moves we would never find that serve as prophylaxis against something the other side might like to do way down the line or whatever?
And if the reason you have this sense of alienness or whatever is seeing the individual counterintuitive moves themselves, you can just play against the table armed computer and get utterly crushed by good old pins and forks and skewers, no 400 move Ctulu knights needed! The table might be very conservative in making sure the opponent gets as little chance to counter as possible, but I'm guessing that the really wacky looking stuff is a product of both sides being omniscient, and that if you can't hang with it, it will just brutalize you in the traditional style of slapping your king around until you "resign" in an angry fit at the last second by quitting the program.
-
It is a big and cool theme in BP that contests between godlike powers may be full of moves that feel locally suboptimal.
-
@ swashmebuckle, that's fine. Not everyone has the same sensibilities for the same things, you are permitted not to feel in awe about this stuff, I hereby give you this allowance :D.
-
While not quite as strange as this, the conversation here reminds me of Derelict.
It took the whole of the GTVA's top minds all the way until the 11th hour to figure out what the Shivans' goal was, and even then, it was only because of the intuition of a random fighter pilot on the front lines who happened to have been at the right place at the right time to make a scan of the right thing.
Almost makes me wish they were in the same universe.
---
Putting that aside, this also reminds me of the conversation at the end of WiH Act 3. The "explanation" of why the Shivans did not just bring out their strongest forces right away and instead chose to slowly evolve their strength to stay only a few steps ahead of the GTVA until they felt there was an appropriate time to bring out the big guns, so to speak. Even then, they just blew up a star for no reason that Command could fathom.
-
If you send in crappy ships and make suboptimal moves then you can generally pretty successfully avoid revealing anything about yourself, while rapidly learning about the opponent, if done correctly.
-
That's the point, though, QuakeIV. They're not suboptimal. They are by the definition of the word the most optimal move for the situation - the opponent simply can't comprehend why because they lack the base, raw computing power to see that many moves ahead in perfect clarity.
It's an excellent analogue for the Shivan's combat doctrine, which often doesn't seem like a doctrine much at all. Aeshma's suck, and Basilisks are not good, and any number of Shivan craft are not designed for a human recognized optimal solution. But they're perfect for the Shivans, because they're playing a game on a scale that humanity simply fails to comprehend.