American Civil War Game Club (ACWGC)
http://www.wargame.ch/board/acwgc/

Could A/I Consistently Defeat Human Players?
http://www.wargame.ch/board/acwgc/viewtopic.php?f=4&t=23493
Page 1 of 1

Author:  Blake [ Fri Dec 15, 2023 9:44 pm ]
Post subject:  Could A/I Consistently Defeat Human Players?

With the recent advances in A/I programs (not in our games of course, lol), do you think it is possible for someone to design a program to take into consideration all of the movements, variables, and decisions that we make each turn, and then use all that to defeat us with regularity in these games?

If so, would that be a positive thing, or a negative one?

I could say "positive" because it would mean we'd finally have an effective A/I to battle against which would be as dangerous as a human. But it would also be a negative as it might end up replacing the need for human opponents in order to be challenged.

I feel like it could be done and that an A/I player could very well compete with humans. I'd be interested to play against an A/I opponent touted as "equal to" a human player.

Computers can already defeat us in Chess, why not in the WDS games? Though I would argue these games are far more complex than Chess. Although that might statement might start an argument in some circles.

Author:  Josh Jansen [ Sat Dec 16, 2023 8:56 am ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Well, Chess is awesome and it can be very strategic.

Our games certainly have many more variables for sure. Because of the limited number of "counters", I dont think even the best A/I could determine every possible strategy, and even if it could, it wouldnt be able to put the little units in all the right spots, just like we struggle with.

It would be awesome if the A/I could be similar to a human in these games, because I personally prefer to play most games solo. However, a human opponent cant be replaced even if A/I was that great. Humans have two things that A/I doesnt: Personalilty and Irrational Thinking. Emailing a game back and forth and getting to know someone is special. On top of that, humans act irrational. Let me explain. An A/I is going to choose the best course of action almost 100% of the time, humans cant do that and wont most of the time. Sometimes the mistakes we make in these games (and in life) end up working out better. A sensible computer would never frontal assault a hypothetical position by crossing open ground and getting hammered by arty.... but a human might.... irrational, and maybe just maybe, that bold move pays off. I just dont think A/I is willing to do that.

But, hell, now that its been mentioned, it probably will read this and cream me when it gets advanced enough. Better stay on its good side :)

Author:  Quaama [ Sat Dec 16, 2023 3:57 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

The referral to chess is a good one. Consider the amount of effort, and time, it took before a machine could defeat the most capable humans. Chess, although complex has nowhere near the variables a wargame has (and each game/scenario is different with different variables).
A quote (https://www.chessjournal.com/how-many-possible-moves-are-there-in-chess/) about chess [I don't know how accurate it provides a good summary of the situation]:
"There are over 9 million variations after just 3 moves each, 288 billion different possible positions after 4 moves each, and 318,000,000,000 ways to play just the first 4 moves. In a whole Chess game, there are more possible iterations of chess games than there are atoms in the observable universe."

Given the additional variables in wargames I suspect it would be impossible to calculate the numbers, let alone program the machine to then capably predict possibilities and then apply suitable moves 'to win'. Josh refers to human nature and how this can effect things, irrational things can sometimes result in benefits.
'There's always the unexpected, isn't there?' And machines can't deal with that sort of thing.

Author:  Blake [ Sun Dec 17, 2023 2:25 am ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Quaama wrote:
"There are over 9 million variations after just 3 moves each, 288 billion different possible positions after 4 moves each, and 318,000,000,000 ways to play just the first 4 moves. In a whole Chess game, there are more possible iterations of chess games than there are atoms in the observable universe."


Thanks alot, Paul. I was about to go to bed and now my mind is blown from that. I am going to be staring at my ceiling for an hour pondering the insanity of Chess :mrgreen:

Author:  Bill Peters [ Sun Dec 24, 2023 12:55 am ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Oh I am sure that the current AI can give a much better challenge than I have giving my opponents lately. ROFL

But honestly, it always surprises me that over 70% of the customers play AI only. Now that WDS has a public forum its helping to connect them to the clubs or just to pick up a game. I doubt that the percent drops that much but I have played several people over the years that had never played FTF and now can't live without it.

"Gonna demolish that Activate AI button on toolbar some day" LOL

Author:  L.P. Smith [ Sun Jan 07, 2024 9:26 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Blake wrote:
With the recent advances in A/I programs (not in our games of course, lol), do you think it is possible for someone to design a program to take into consideration all of the movements, variables, and decisions that we make each turn, and then use all that to defeat us with regularity in these games?

If so, would that be a positive thing, or a negative one?

I could say "positive" because it would mean we'd finally have an effective A/I to battle against which would be as dangerous as a human. But it would also be a negative as it might end up replacing the need for human opponents in order to be challenged.

I feel like it could be done and that an A/I player could very well compete with humans. I'd be interested to play against an A/I opponent touted as "equal to" a human player.

Computers can already defeat us in Chess, why not in the WDS games? Though I would argue these games are far more complex than Chess. Although that might statement might start an argument in some circles.


It's coming. Would love to play against a competent AI.

Author:  Blake [ Sun Jan 07, 2024 10:22 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

I don't know.

I've been playing Ultimate General and Take Command: Second Manassas this week. The A/I in those games is... interesting? They don't seem to think very much and with ultimate general it feels like they just mass men and fire in all directions. Is it challenging? Yes. But not because you have to think about it, because you just have to endure a mob of armed men firing on you. When you move your mob closer, it just gets confusing and bloody. Strategy? I don't feel that it is. Yet the reviews for the game rave about the A/I. Is the A/I really thinking, or are they just programed to defend flags on the map? It feel like they just clump together on map points and that's it.

Take Command is very old but still very good. The A/I uses infantry well but they can't use their cavalry or artillery to save their lives. But even all these years later, man, that is a fun game. The infantry spreads out and they do seem to know enough to go for your infantry flanks. Of course, all you do is put guns on your flank and you shred them.

Is Grand Tactician any better? I don't know yet. I haven't played it.

But these games are all realtime games and the A/I still feels like it is just aiming for points on a map and that's about it. They will fight you if you run into them but that's it.

I am sure the A/I will eventually be a great challenge in games like ours. Maybe not any time soon though and maybe not with WDS. I don't see them pouring the money and resources into A/I development to make it as good as a human. Like Paul mentioned before, the calculations for our games are astronomical. Could a supercomputer do all of that and think intelligently enough to go "off-script"? Maybe someday. Humans never follow the script which makes us very hard to anticipate or defeat.

Author:  Ashdoll Ren [ Mon Jan 08, 2024 8:59 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

I believe AI will finally go beyond us in many areas, including wargames, no matter whether turn-based or real-time. Besides serving as a better opponent discussed above, I am also interested in that AI can serve as better subordinate officers in the OOB. So we can just send orders to our AI officers instead of ordering every single regiment or unit in real-time (through a kind of mystical telepathy capacity). Many games have provided such a function, but the current AI officers are underperforming.

-

Maybe Go is a better example than Chess to understand the method. Despite its relatively simple rules, Go is extremely complex. Compared to chess, Go has both a larger board with more scope for play and longer games and, on average, many more alternatives to consider per move. IBM's Deep Blue defeated the Chess World Champion in 1997. 19 years later, Google's AlphaGo defeated the Go World Champion in 2016. Only 1 year later, AlphaGo Zero in 2017 can learn to defeat Humans without learning from any Human experience but just by playing against itself.

AI wasn't based on the calculation by exhaustion to defeat Human Go players. And, personally, I suspect the computational complexity of a specific wargame scenario is less than Go.

Author:  Karl McEntegart [ Mon Jan 08, 2024 10:38 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Damn and blast it Sir, now I'm going to have to go out and buy Go............. :shock:

Author:  Josh Jansen [ Thu Jan 11, 2024 5:33 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

Karl McEntegart wrote:
Damn and blast it Sir, now I'm going to have to go out and buy Go............. :shock:


Its very difficult, maybe you will be better than I. Seems simple, but I dont understand it. Good luck!

Author:  Blake [ Sat Jan 13, 2024 12:33 pm ]
Post subject:  Re: Could A/I Consistently Defeat Human Players?

I'd love a A/I that could pose a real threat.

But getting beat by the A/I in our games?

Image

Page 1 of 1 All times are UTC - 5 hours
Powered by phpBB® Forum Software © phpBB Group
https://www.phpbb.com/