Throwing out a question concerning AI

[Pidesco]

Apart from the AI and the graphics, F.E.A.R. wasn't a particularly good FPS. In fact, compared to Monolith's previous games F.E.A.R. was a big disappointment for me.

 

I'll stop with the offtopicness now.

Edited January 13, 2007 by Pidesco

[steelfiredragon]

well not all games can have "smart AI" like Cortana can they???

[alanschu]

That's exactly how Kasparov was beaten by the IBM Deep Blue; the computer was busy scanning all the possible moves from the current position, so that when Kasparov eventually made his move the computer moved immediately. Part of the psychological strategy of the IBM team. (It is a commonly held belief that Kasparov was the superior player; the IBM team used all sorts of meta-gaming techniques to put him off his best game.)

 

Deep Blue did not make its moves immediately. In the first game it was a bit more mechanistic, not using more than 3 minutes for a move, but as the game came to a close, there were moves that were taking as many as 6 minutes. Some moves took as much as 15 minutes.

 

Kasparov's issue was that Deep Blue wasn't designed to be a good chess player, but rather it was designed specifically to be Kasparov.

 

He also regretted playing the final game, as he wasn't really mentally into it (as evidence by his conceding defeat after only 19 moves). The big advantage Deep Blue had was that it didn't get fatigued. Kasparov spent more time analyzing the games than any other game he had taken part in, and by the end he had lost his fighting spirit.

 

He also admits that his prior experience with computer chess players led him to some false assumptions about Deep Blue. This was part of the reason why he lost Game 2 (it also led to him speculating that Deep Blue was being tampered with, since Game 1 was more in line with how computer chess players tend to play). The fact that it was making moves that he absolutely did not expect shocked him, and added to the pressure. That's what psyched him out.

[Pidesco]

well not all games can have "smart AI" like Cortana can they???

 

AI with boobs?

[metadigital]

That's exactly how Kasparov was beaten by the IBM Deep Blue; the computer was busy scanning all the possible moves from the current position, so that when Kasparov eventually made his move the computer moved immediately. Part of the psychological strategy of the IBM team. (It is a commonly held belief that Kasparov was the superior player; the IBM team used all sorts of meta-gaming techniques to put him off his best game.)

 

Deep Blue did not make its moves immediately. In the first game it was a bit more mechanistic, not using more than 3 minutes for a move, but as the game came to a close, there were moves that were taking as many as 6 minutes. Some moves took as much as 15 minutes.

 

Kasparov's issue was that Deep Blue wasn't designed to be a good chess player, but rather it was designed specifically to be Kasparov.

 

He also regretted playing the final game, as he wasn't really mentally into it (as evidence by his conceding defeat after only 19 moves). The big advantage Deep Blue had was that it didn't get fatigued. Kasparov spent more time analyzing the games than any other game he had taken part in, and by the end he had lost his fighting spirit.

 

He also admits that his prior experience with computer chess players led him to some false assumptions about Deep Blue. This was part of the reason why he lost Game 2 (it also led to him speculating that Deep Blue was being tampered with, since Game 1 was more in line with how computer chess players tend to play). The fact that it was making moves that he absolutely did not expect shocked him, and added to the pressure. That's what psyched him out.

<{POST_SNAPBACK}>

I suspect your expansion on my point was concerning my poetic licence over "immediate" ... I don't know about you, but I've played many computer chess opponents that take hours, days ... even PBEM games.

[SteveThaiBinh]

Out of interest, has a computer beaten a reigning world champion since then? Has there even been an attempt? Or did the world lose interest?

 

Were Deep Blue's moves fast in order to psych Kasparov out or just to stay within the time limit?

[alanschu]

I suspect your expansion on my point was concerning my poetic licence over "immediate" ... I don't know about you, but I've played many computer chess opponents that take hours, days ... even PBEM games.

 

Well, this wouldn't have been allowed.

 

The rules of the match limited the total amount of time, so moves had to be "relatively" quick.

 

They had a maximum of 2 hours for the first 40 moves, 1 hour for the next 20 moves, and 30 minutes for all remaining moves. If you had time remaining from one area, that time carried over. So if you only used 1.5 hours for your first 40 moves, you had 1.5 hours to play your next 20.

 

Basically, 40 moves must be played by the 2nd hour, 60 moves must be played by the 3rd hour, and the rest of the moves must be done at the 3.5 hour mark. 3.5 hours was the longest possible game.

Edited January 13, 2007 by alanschu

[metadigital]

Deep Blue was also a supercomputer, so it could do pretty well in those moments.

[alanschu]

That it was.

 

I would like to have seen a move by move analysis of how long each player took, but I seem unable to find anything of the sort

[metadigital]

CONSPIRACY

[Calax]

well not all games can have "smart AI" like Cortana can they???

<{POST_SNAPBACK}>

Rule 34?

[Watchman]

I think the originator of this thread has the answer he was looking for:

 

1. AI casts $ to develop that most companies don't think will pay off

 

2. AI is processor intensive cutting into the ability of the game to expand in other fields

 

3. Games are limited simulations of enviornments on which only the limits of the game can be worked with. AI can know all variables in the game. Only enough is needed to give the player a challange. Beyond that it could flatten players and become a problem. Most of what people comment on when talking about AI is beter enviornment simulation that must be build into a game before the AI can use it.

 

I believe that is small list of what has come up so far.

Edited January 14, 2007 by Watchman

[Checkpoint]

Er... Thanks. But how is your post anything more than a summary that couldn't be gathered by reading a relatively short thread? We went off topic right after the first page...

 

Also, I think I'll change my username to "The Originator of This Thread."

 

The Destroyer of Worlds.

 

The Annihilator of Injustices.

 

The Originator of This Thread.

 

Yeah. That rocks.

[Walsingham]

I can't cope with the AI in company of heroes because it can multi-task, whereas I can't. It can launch two or three well-resourced combined arms attacks. Impressive, but practically impossible to nail for my simple brain.

 

For me, I want an AI that has a personality. Ultimately I don't believe it is enjoyable beating someone by playing better numbers. It is enjoyable to beat something by mentally seeing their weakspots and exploiting them. A superb AI has few weak spots.

 

As I say, I do wanta smart AI, but I don't want a supersmart hypercapable AI. Because I don't want to have to become an accountant to beat them. I want to enjoy beating the otehr guy. I want to be Patton, not bloody Monty, if you get the reference.

[metadigital]

I want to enjoy beating the otehr guy. I want to be Patton, not bloody Monty, if you get the reference.

<{POST_SNAPBACK}>

Certainly a prima donna whichever way you go ... )

[Walsingham]

We already knew that.

[Plano Skywalker]

I can't cope with the AI in company of heroes because it can multi-task, whereas I can't. It can launch two or three well-resourced combined arms attacks. Impressive, but practically impossible to nail for my simple brain.

 

For me, I want an AI that has a personality. Ultimately I don't believe it is enjoyable beating someone by playing better numbers. It is enjoyable to beat something by mentally seeing their weakspots and exploiting them. A superb AI has few weak spots.

 

As I say, I do wanta smart AI, but I don't want a supersmart hypercapable AI. Because I don't want to have to become an accountant to beat them. I want to enjoy beating the otehr guy. I want to be Patton, not bloody Monty, if you get the reference.

<{POST_SNAPBACK}>

 

absolutely right, "good AI" is simply that which mimicks a good human player, not an entity that makes no mistakes and has an unfair advantage (i.e. goes for all the "tech combos" early on that you would have to read a strategy guide to even know about).

[Calax]

The AI in COH does cheat... (tanks randomly chase after hidden snipers) however it also doesn't use its' extra abilities that much (never seen the expert AI fire off smoke or get veterancy)

 

Really you shut it down rather quickly if you just use your building unit to cap points while building at your base (I generally do two rovers one building squad) then you deny them the resources needed to produce units... except for manpower, thats where fighting Blitzkreig and Airborne doctrines get sticky (able to produce stuff thats good against anything for just manpower... airborne, how I loathe thee)

 

 

oh, and for an old time easter egg, hit ctrl ~ and on the console type in ee_bigheadmode(1)

 

it's funny.

[Walsingham]

Thanks for the ideas. I do sometimes wonder if I'm just too damn old to beat it. But I am itching to whallop a human player. And teh single player missions are the best EVER. Ai as a cinematic tool.

[Kalfear]

  Thanks for the ideas. I do sometimes wonder if I'm just too damn old to beat it. But I am itching to whallop a human player. And teh single player missions are the best EVER. Ai as a cinematic tool.

<{POST_SNAPBACK}>

 

Wal, what you describe (to me anyways) in your previous post is not so much AI as it is TB vrs RT topic.

 

I understand exactly what you are saying though. Take Civ 4 for example. In the turn based game I can beat the computer each and every time. But if I change the setting to RT I lose on cheiftain because a human being is limited and can not realistically multitask the game as much as the computer (assuming you have multiple fronts of course).

 

Thats always been my complain about RT, so what if I cant point and click as fast as the next guy, my stratagy is what should matter, not my clicking speed.

 

I am one of the worst RT players on the planet (im sure of this as I like to think and RT punishes you for thinking). But set me up in a turn based game against same opponent and I never lose.

 

So anyways, I think you mistaking one issue for the other to be honest. I dont actually disagree with you, just dont think AI is the real issue you speaking about because if the game RT, it doesnt matter really if its AI or Real Player, he who clicks fastest wins in those games. Simple as that.

[Checkpoint]

Thats always been my complain about RT, so what if I cant point and click as fast as the next guy, my stratagy is what should matter, not my clicking speed.

<{POST_SNAPBACK}>

Yes. Nothing as stressing as RT; I passionately hate not being able to sit down and have a good look at my options and act accordingly. Thank God for the pause button. And never play multi-player.

[alanschu]

im sure of this as I like to think and RT punishes you for thinking

 

Right.

[LadyCrimson]

I've dreamed of a good RTS AI opponent that didn't rely on computer decision speed vs. human decision speed. I have yet to experience one. I can understand money/time being a large issue, but over the years it's pretty disappointing that it hasn't become much better, even so.

 

There have been plenty of strategy games that were crazily difficult to beat/weren't easy, where myself and perhaps others would want them toned down for being too unfair - but that doesn't neccesarily make them good AI. Typically the AI opponents I've liked best were very defensive rather than offensive, simply because it's less of a speed-game.

[Plano Skywalker]

I've dreamed of a good RTS AI opponent that didn't rely on computer decision speed vs. human decision speed. I have yet to experience one.  I can understand money/time being a large issue, but over the years it's pretty disappointing that it hasn't become much better, even so.

 

There have been plenty of strategy games that were crazily difficult to beat/weren't easy, where myself and perhaps others would want them toned down for being too unfair - but that doesn't neccesarily make them good AI. Typically the AI opponents I've liked best were very defensive rather than offensive, simply because it's less of a speed-game.

<{POST_SNAPBACK}>

 

right, when I play a strategy game on "hard", it is hard because the computer is making decisions faster than I am, exploiting tech combos and the like faster than I am.

 

what we need is to see actually personalities emerge in the AI and with the AI only going for the game lock combos when it sees we are about to do the same.

 

such an AI player would just be a lot more fun to play against.

[LadyCrimson]

I'm not sure what is meant by "game lock combos" but there have been games with different AI types - of course, they still weren't dynamic in response to player-actions, they'd just use different "plans", and thus were predictable before long.

 

Dynamic reactions to human actions would be nice, yes...assuming, of course, that the programming made it feel random enough in some way. If it only had one reaction to any action - where I knew if I didn't want it to build a certain unit I just had to avoid doing a certain thing to cause the AI reaction - that would be just as predictable, I think.