Learning about programming

why not simply have shit tier graphics but good tier AI, music, story, art style and so on.

because video games like making money, and shiny graphics are the first thing people notice and the easiest to advertise

user this is a mistake many wannabee programmers make when reading on AI.

If you make the perfect AI, it doesn't translate to a fun game.

It's actually extremely easy to make an unbeatable AI, it's the easiest thing you can do, the true goal is to make an AI that is beatable but interesting enough to fight against.

It's a lot harder to achieve the right balance, any code monkey can make an AI that can never be beaten at chess and stupid shit like that, your mind is still too small to understand the actual problem here.

You'll get it in time.

Because no one wants to bother working on a game that no one would ever play anyway.

You ever heard of a Japanese game with no waifu potential before? Yeah, not happening here.

You realized that games have shit tier AI just now? What the fuck have you been doing all these years

he probably was busy being born and breastfed.

I'm talking about interesting AI, not unbeatable one.

like an RTS army with a director you can waifu and has emotions.

except minecraft, sims, AoEII, WoW are the biggest money makers.

hell, even nintendogs sold like hot shit.

...

aoe 2 had good graphics for the time, wow's was better than the rest of the mmo field and minecraft didn't run ads

There could be games that benefit from advanced AI, but as an opponent it would be bad. Maybe as a follower, or a judge/referree npc.

...

Smh tbh fam

there's no need for it to be an advance enemy, most of AI could be enhanced by emulating emotions and group behaviours.

imagine a beat em up where the enemies have flocking behaviours or some rpg where the enemies call hear each other.

there's no need to be overpowered, simply be well designed.

do you remember pacman, right?

If you need to ask the question, you won't understand the answer.

AI won't be a thing in videogames until you are long dead.

The way to make good AI in videogames using current tech is to make very simplistic AI and have a very good awareness of all the situations that might come up as well as have the AI be able to make those simplistic responses to subtle things. The Nips are about a million times better at this than anyone else in the world, as evidenced by games like DMC1.

Those already exist.

Well no shit that games have shit AI. Imagine the computer trying to simulate 50 different NPCs all interacting with the environment simultaneously with fuckhuge walls of text specific to every single NPC.

top kek

what is pacman?

there's enough cycles if you have something like Mario 64 graphics.

realistic garbage.

top kek

Try F.E.A.R.

there's no need to make a superhuman skynet AI.

game AI is about design interesting enemies, not making academic AI.

We already have good tier AI since Doom and Pacman, great examples.

...

Designer has to define what information is, where it is, how it can be perceived and what an AI can do with it. Then every AI needs a record of what it knows, what it wants and what it does, creating a dynamic behavior.

This is all very complicated. So developers won't be putting that effort into their games as long as they can get away with cheap scripted garbage and basic if-then behavior with no memory or any level of simulated awareness of the virtual environment.

The way the industry goes, it doesn't even have the talent left to even try such an AI development.

only the dead can know peace from this pedantry

Whatever it takes to derail a thread.

Well yeah but you're not really going to see anyone try to do that just so they could achieve it, at least not in practice.

...

You have understood exactly zero of what I wrote. Congratulations. Does your handler type your posts? You sure are too stupid to be literate yourself.

you literally said that designing interesting characters.

congratulations, y'all retards who think game AI is the same as academic AI.

designing characters is too hard *

Doesn't matter if you're pronouncing it or not, it's the grammatically correct method.

Yes, yes. Go back to binging on Extra Credits. At least they know what they're talking about from professional experience.

lmao

nigga, I was reading about game design and game AI since 10 years ago, before those faggots came about.

do you even know who is chris crawford?

Do you even how to capitalize?

this thread is bazinga tier

you seem to worry too much about some taco english in a chan.

fuzzy logic is god tier user.
cheap as fuck to implement and you can do crazy shit with it.

what are you, 12? everyone with two neurons knows this.

There is nothing particularly special about fuzzy logic. It's obvious that you barely understand the topic and felt the need to namedrop shit in order to seem smart. Like I said, bazinga tier.

do you think game AI needs to be complicated?
holy kek

Do you have reading comprehension or do you usually just shit random words into the reply box?

I'm sure you had realized now that I'm not a native speaker, so maybe there's some missunderstanding.

look.
my point is that good game AI doesn't need to be complex.

is not a problem of complexity, but lazyness from developers who focus on the graphics alone.

I'm sure a new game using simple graphics like Mario 64 could have god tier AI.

Most games don't even have real AI to begin with.

You also have to account there are multiple instances of the AI running at a single time, they purposely make it simple so the cpu can actually handle it.

I don't think it's necessarily the developers who waste money/time on art assets so much as they have incompetent programming teams. They don't know how to make something fun without being braindead or balls to the wall extreme. Not to mention AAA games apparently have an expectation for difficulty levels, which means they'd need to develop multiple levels of intelligence, which is where we got artificial difficulty from: the AI isn't actually doing something smarter, it just does more damage or takes more damage.

It's also pretty hard to do something unique when a publisher is breathing down your neck to get a product out as fast as possible, so just slap some easy to do bullshit and get it done.

There's also age and seniority; I hear stories about interns being more competent at programming than the programmers themselves.

FEAR has simple AI though. I hear they use various tricks to make it seem smarter than it really is.

Pretty much this. Think about a shooter game. Halo, in particular, was focus tested using a variety of different ai strategies. Whenever they made them more accurate, players felt like it was bullshit and they had no way to react.

In fact, the only thing that felt "fair" to players was simply increasing the health of enemies.

tl;dr– AI in video games aren't stupid, they're meant to conform to the player's expectations.

That's true of any workplace environment though, especially in development. Contractors are worked harder than full-time employees. Juniors/mid-levels are hired on because they understand the new technology better than seniors. And so forth.

Game studios in particular are very selective about who they work with. So it wouldn't surprise me if this normal effect were magnified. They're not going to hire retards, even at the intern level.

w-what happened to those waifus, op?

Civilization: Call to Power comes to mind.

For FPS maybe. For something like 4X current AIs generally cheat or are pushovers, usually both.

filename

Why do they call them AIs if they're basically a bunch of ifs and elses?

I read it but it doesn't tell me much.
Did you build an ai to composite some waifus into procedural waifus?

Civ V initially had a very good internal AI. It also took upwards of fifteen minutes per turn per AI in the Renaissance area and could easily sprawl out to hours by Atomic as each player individually calculated the best path to victory and updated that action plan every turn.

The average computer isn't strong enough to permit good AI to run in a timely fashion so they make up for it by cheating or giving it hella shortcuts.

The one genre of strategy (and I use the term loosely these days) where AI have a real chance to compete is probably moba honestly. Stuff like Berkeley Overmind shows that computers have a natural bonus in micromanagment.

is not mine.
some user programmed a waifu generator using deep learning or some shit.

this tbh
people who say FEAR has good AI are wrong by definition

you mean rts, right?

Good micro gets you basically nowhere in Supreme Commander and is only about 1/2 of Starcraft. Can't paint RTS with a single brush.

but where the fuck do you find micromanagement in mobas
they're 1 character games, with maybe 2 or slightly more controllable minions for some characters

I think you mean rts. Hundreds of units with concentrated fire calculated to deliver exactly the right amount of damage to foes to maximize use of dps.
I remember watching some video of an AoE2 game where two highest difficulty ai went at each other with almost nothing but monks.

That's deep dude, you should become a game developer.

It should be really, really, really obvious that when someone talks about making AI that's not retarded, what they mean very specifically isn't taking a retarded AI algorithm and changing its "accuracy" and "reactionTime" settings to "1" and "0.01", user. This sort of shit should be game developer logic 101 - remedial game developer logic 101.

...

good game AI is about designing interesting characters that work in harmony to create complex output.

Like the ghost in pacman, that are basic as fuck but they're interesting and working in harmony make them complicated.

Explain the STALKER games. There human enemies aren't bullet sponges even on the highest difficulty.

It wasted most of it's time shifting every worker around for every tile. That was like 60% of the bloated processing time of early CiV.

A good AI and a unbeatable AI are two different things, you retard.

I can give any AI controlled enemy GodMode. That doesn't make it good AI.

the human enemies are bullet sponges on everything but the highest difficulty. Just the mutants are bullet sponges on master. Stalker's got god awful difficulty design.

Simcity 4 comes to mind, at least it seemed that way. You had this happiness indicator that would make your city grow at a rate. It's probably a RNG in the back but I could be implemented as a fuzzy logic system.

Actually game AI and real world AI are basically the same, the main difference is that in games you have the full state space and in the real world you have as far as the machine can "see".

The creator of alphaGo was also the lead AI programmer for black and white.

This is probably the closest to a fuzzy system but no, it's actually a probabilistic logic system.

They are almost the same but with PLS you have the actual probabilities of outcomes and in FS you have the "kinda, not gonna happen and it's happening" outcomes.

It has a bunch of scripted options (e.g. predefined spots where it can vault over an obstacle, or whatever) and some unscripted ones (tossing grenades, suppressing fire etc.), that it uses as building blocks when formulating a plan to kill the player. Most AI just react to the current situation with predefined behavior, e.g. tossing a grenade if the player is in cover, unless it's out of grenades, in which case it should try to get around the cover, but not if someone else tossed a grenade just now or there's some harmful obstacle in the way, etc. etc., basically traditional enemy AI can't think very far ahead because it needs to account for all kinds of circumstances.
FEARs AI could in theory be extended e.g. with a hint about red barrels exploding and damaging the player if shot or damaged, and it would integrate with the other options and it would start doing shit like starting chain reactions to get at a player that was hiding behind a wall.

There's two interpretations of the phrase. One is the scientific "man-made intelligence," the other is the more game specific "appearance of intelligence."