With all the advances in AI in other fields, why is video game AI still absolute trash?

With all the advances in AI in other fields, why is video game AI still absolute trash?

Attached: 1496154246686.png (1050x1500, 2.61M)

Other urls found in this thread:

youtube.com/watch?v=PFMRDm_H9Sg
twitter.com/SFWRedditGifs

Your computer sucks, and games aren't fun if you can't beat the ai

cause gamedevs are the bottom of the barrel when it comes to programmers, all the brilliant ones work on level stuff

Because consoles suck dick on the processor side and no one makes PC centric games anymore.

because it gives retards like you the illusion of intelligence and helps sell games

Most vidya ai is good enough for what it needs to do. As long as it doesn't react the same way every time that's good enough. No amount of perfect path finding will save an AI from getting stuck on a clip of map that the retarded level designer forgot to flatten out. In short, AI is not the weakest link.

Just cause 4 just got an AI update. Its still a clusterfucj on the ground however I like the air battles

Attached: JustCause4_2019_05_27_18_54_17_557.webm (948x400, 2.76M)

Because video games are literally just digital toys for children.

Why would you develop a high end AI just to impress some pre-teen and teen kids when CoD and Fortnite does it without any hassle?

Attached: 1553386935765.jpg (1024x762, 144K)

what is resident evil 2

>what is difficulty level choice?

Video game AI is actually supposed to mimic decision making.
Other things called "AI" are just automated statistics, or mathematical models generated by automated statistics.

AI is the thing that makes the google botnet know what stairs and street lights are.
Vidya "AI" is just a bunch of code saying
while (this.Dead == false) {
if(canStab(Player)) Stab(Player);
else {
DoPathFindingTo(Player);
Move(Path);
}
}


There haven't been major vidya AI breakthroughs in a while, most of the time developers use random numbers to make it look like the AI fucks up every now and then and make it believable.

Attached: 61c2ef993e435456286adc5cce276900444e5e97.jpg (1838x551, 124K)

This guy is right. The player is so overpowered relative to enemies in every game that gaming is totally retarded, and if that weren't the case, almost no one would play them. Gayming.

Because we're in 2019 and single player game modes are dead

This
Video game AI isn't supposed to be smart, it's supposed to be convincing, similar but not the same, it's not worth the time and effort putting actual good AI in video games
Strategy games could benefit from some strong AI but they're kind of niche anyway and people will buy them even with shit AI so why bother

That's ML specifically. Automated planning can be static and behavior trees are kind of related to it. Some games like FFXV straight up have automated planners to orchestrate their actors. Technically it's sophisticated, doesn't mean it's any good at its job of creating convincing simulations of life.

Believable AI is an art problem right now, not a technical one. And it's going to stay that way until we can emulate the entire human physiology

>Muh AI is too hard to make for a single video game
Why can't brainlet developers use NN to do the work for them?

what work can a NN do for you in developing a video game exactly?

If they can make an AI using NN to destroy pro Starcraft 2 players, they can make an AI for anything

What this user said. You can even have hierarchical planning, where you precompute or hardcode high level policies and let a heuristic planner handle the low level actions. It's great, but I don't think it would be very fun.

Having worked on planning systems, I can say: the AI tends to cheese out any chance you give it. An optimal AI would probably wait forever in a safe corner to shoot you in the head the second you showed up or something like that

>If they can make an AI using NN to destroy pro Starcraft 2 players, they can make an AI for anything
That's completely wrong on every level

>omg, this is literally like dark souls, i want a refund

>brilliant ones work on level stuff
Agree, would work on a level

youtube.com/watch?v=PFMRDm_H9Sg

It's somewhat niche but machine learning is being used for some video game related experiments.

A way of limiting an AI to the point you stupid human brain can defeat them

I usually played video games on their highest difficulty.

Because it isn't a priority. Develop good "AI" for a game would take years. Plus it's a pain in the ass to figure out why good AI is making some decision that the player's find annoying.

>rico! the king of the rebels!
>hey... rico... could i get an autograph, or maybe a parachute ride?
>rico! the king of the rebels!
>rico! the king of the rebels!
>hey... rico... could i get an autograph, or maybe a parachute ride?
>rico! the king of the rebels!

Because the level designer and AI dev need to work hand in hand like they did for FEAR.

heavily scripted encounters are not AI

Because AI is still made wholly by programmers, thus it's limited by what the programmer can conceive of. Emergent behavior can't be counted upon to do this, because it involves far more complexity than any alien grunt with a laser gun needs to have. Only now we're beginning to make AI which can judge and improve itself, and there's a loooooong way to go before they can pass a behavioral Turing test.

Why develop good AI?

Why develop AI at all when the better the AI, the less likely the player will win, and the less likely the player has of winning, the less likely they will get that dopamine reward which makes them like your product and become willing to buy the next 10+ years of sequels and reboots? There is no reason to develop better AI from a business pov. Most gamer tards don't want a challenge, they want to hit the buttons on the controller until the screen gives them a conceptional cookie.

Attached: dab.jpg (800x450, 27K)

>tfw Brendan is being unJUSTed

Attached: D6eUyYjXoAE_fr5.png orig.png (670x528, 215K)

FUCK!!

FUCC