The AI would be for implementing each of those behaviors convincingly, I guess.

Each of those behaviors already has been implemented convincingly. It's just a matter of cribbing from somebody else's work.

I thought the whole point of this thread was that they hadn't.

In the GTA IV example, there is not an AI making a decision as to how to beat the player. Rather the AI is following a script, with inherent probabilities.

I'll us GTA IV as the example: You pull out your gun in the wrong neighborhood. 1/2 the people are armed the other half (unarmed) are scripted to run. Do the Armed persons just sit there and shoot at you regardless of if they have cover? For easy level yes, for hard no they are scripted to move then fire from cover. Really here we are talking about the AI following a IF condition X is present then I take action Y.


This is not the issue at hand, rather than does the AI have limits in resources, if no then it really can just "cheat" to win the scenario by sending wave after wave of mooks. Otherwise, does it use some mooks to pin the player behind his car? While the remaining try to flank. Do they attack piecmeal, or all at once depending on the situation?

My original question, and many others seem to think as well AI development has kind of been poor compared to the development of graphics. And in many cases sequal have worse AI's than their predecessors.

Thought about it a little bit. If SWAT 2 were re done with an emphasis on multiple high quality AIs it would be a hell of a game.

It would also need an interface where you can actually understand what's going on and issue orders to your unit.

Soldier of Forture 2 did have some issues with realism, mostly in the form of the teleporting keycard squads. You go through a cargo ship killing everyone you can find, which is more crew members than the ship could possibly support, even if they slept three to a bed, and then you flip a switch and there are suddenly dozens more hostile crew members for no apparent reason. Ships like that really aren't designed to support more than thirty or so crew members. But the game just keeps spawning more and more of them for you to kill.

So yes, more realism would have made it significantly easier. But it isn't just enemy AI, it's organizational composition.
Video game hostile organizations tend to have absurd numbers of faceless mooks thanks to unrealistic spawing. You end up with smalltime street gangs who apparently outnumber the US Army and other things of that nature. Terrorist cells that should have five or ten people tops somehow throw hundreds of cannon fodder soldiers at you, all armed with impossible-to-obtain fully automatic weapons. In a worse case scenario it doesn't matter how many millions of people you kill, even though the game supposedly takes place in the real world.

This is the best post I've read in a while.

It makes me think there could be potential in a first person adventure type game in a realistic setting which supports reasonably good combat gameplay, but at the same time doesn't focus exclusively on it, so that you could just make an interesting immersive story about, say, a regular guy who stumbles on a terror cell in his city. Instead of mooks, each bad guy would be an actual character.

People playing a first person shooter like to first person shoot. That requires targets. They like to feel awesome and roflstomp the game, and that requires targets that don't murder-face the player on default difficulty. The end result is wave after wave of faceless mook. The more action-movie FPS games get, the better they seem to market (compare the newer Rainbow Six games to the originals), so ultimately it's our fault. The lack of realism is just gamers getting what gamers voted for (with their wallets).

Well you don't see any new Rainbow Six games in my damn collection, that's for sure.

I need to find and post my screenshot of Raven Shield end mission screen where I completed it after being hit 32 times (but to be fair, it was mostly shotguns going through doors or near misses).

Gods I miss Raven Shield again, I want to play it when I get home. My roommate and I would do co-op elite tango hunts (and hostage rescue as well) with just the two of us pretty effectively. It was a lot harder if you didn't abuse smoke. I can't even remember how many people got pissed at us during hostage rescues when we'd go in with our 12ga shotguns and raise hell, but thanks to somewhat shoddy AI, as long as there was a door in the way of my loud gunshot, they wouldn't shoot the hostage. (always enjoyed the FN-FAL with a silencer for long range shots as well as picking off tangos through doors)

That said, on elite, the AI cheats and knows where you are most of the time. Ghost cam sometime and watch them start tracing people behind walls as they walk up to doorways.

While I do tend to suffer from mook fatigue when mook waves are overused, I don't object to the concept in its entirely. It certainly has its place in a certain style of game. The problem is that it can get totally ridiculous when overused, especially in supposedly realistic games. This is especially true when combined with poor level design and unrealistic spawning.

Oh, yeah. I'm right there with ya, trust me, just saying it's kind of our own fault (as gamers, we get what we pay for). I remember when I was first playing Call of Duty: Modern Warfare I kept thinking "Man, there's an awful lot of these terrorist fellas, huh?" There are certainly times it gets awful silly.

My favorite FPS with mooks remains Damage Incorporated by Richard Rouse. Since you were clearing survivalist compounds, the kind of mass slaughter portrayed in the game wasn't too jarring in terms of the numbers. And in the ending the last boss comments how a few hundred slaughtered aren't enough for you.

Necto verratas nictu....?
No.
Nepto verratas nictu?
Close enough.


Just started my 3rd game of Civ 5. I'm not sure of it but I think they forgot the program the AI to do Amphibious assaults. There has yet to be one....I take that back, there was on launched against an allied City state.

But of note in the game design, difficulty is not so much the improvement of the AI, but bonuses given to the AI

Klaatu... Verada... Nic*coughcough*. Okay, I said the words!

(side note - the words Klaatu Verada Nicto are from "The Day the Earth Stood Still (original) as the words that would stop Gort.)

My guess is that a lot of publishers yell at the game designers:

"MORE EXPLOSIONS!! YEESSSSS BLOW UP THAT BUILDING, OH AND THAT OLD LADY, NOW BLOW EVERYTHING UP !!" followed by "NooOOooOOooo, Our players don't need any challenge! We want the game to be easy as pie so that we can sell it to total noobs and people never played any video game before"

.... well that's how I imagine a board meeting between a dev studio and the publishers

Which is why I'm looking forward to what DoubleFine is going to put out next! No Publishers!!!

Neither of these things are true. At least not mostly. I don't doubt SOME publisher has yelled that at SOME studio, but that's not how it usually goes. Mostly it's the publisher yelling, "What do you mean you're behind schedule? What do we have to cut to get you to the release date on time?" Publishers have some say in how explosive a game is, but it's mostly in the pitch phase. If you say, "I'm gonna make a slow-paced, thoughtful adventure game about brewing coffee," they won't step in and demand more explosions - they just won't give you any money to begin with.

And Doublefine does have a publisher. Self-publishing is still publishing - look at the Dead Kennedys. It means your success or failure rides only on yourself.

OK, how about advertizing? I can think of two excellent games that were killed by ads: Callahan's Crosstime Saloon (Which was billed as a WESTERN!), and Brutal Legend.

What part of advertising? In-game, or like... magazine ads?

Magazine.

In-Game Advertizing, the worst I've seen was Alan Wake. Luckily it didn't detract from the game.

Magazines and magazine ads are, as far as I know, always terrible. But I don't know how Brutal Legend was killed by advertising. All the ads I saw were pretty normal.

I think he's referring to the fact that Brütal Legend was marketed as a third-person action game, when in fact it's closer to being an RTS.

Fixed that.

That is why the AI development gets cut. I have not seen a challenging AI in a game since X-com. And even then it was only a challenge till the old Jedi mind trick came into play.

Hopefully the new X-Com will have challenging AIs.

And the money being pumped into Wasteland 2 should make for some really interesting options for AIs I hope!

That's also not generally true. If AI development gets cut - IF - it's almost ALWAYS because it needs to be simplified in order to work in the context of the game. If there's a need to cut something due to a time crunch, it's levels/zones, player options, or game features. That's because if you straight up cut a level, game mode, customization option, etc, people will often be none the wiser. If you cut AI development, you still need to implement AI of SOME kind.

AI is VASTLY better than the days of XCom, your perceptions of it not withstanding. Xcom's AI is not particularly intelligent; on higher difficulties, enemies are simply more accurate, and cheat more outrageously. Enemies were still highly predictable even at Supergenius.

Do not mistake game difficulty for AI quality.

I don't. that has been one of my complaints with HOI3, Empire Total War, Shogun Total War II (was looking forward till my friend got it an saw the AI was no better than Empires), and now CIV V. But the Aliens in X-com did seem to do some intelligent things, but 15 years later my memory of it might be more a fond remembrance.

Have you tried Frozen Synapse, by chance? Some very, very clever computer work in there. It uses all the usual AI tricks, but it really doesn't cheat, as far as I can tell. Which is wonderful.

Oh yeah, Frozen Synapse is something wonderful - got distracted by its multiplayer and didn't finish the single player. There are some really amazing tricks in tactics you can pull off in that game, like stand duck to get the attention of a gunner so your shooter has the time to take them out from a different angle.