How come AI hasn't advanced very much in the last decade?

[piratecap]

Wouldn't it be awesome if enemies in game could provide some new elements of situational awareness. We've seen the same tired level of AI now for a lot of years. Graphics is cool and all but gameplay would be so much better if enemies could provide something new when in combat. I remember thinking the enemies of Half-life 2 and F.E.A.R were pretty cool, but since then nothing has wowed me in that department.

Is it because it's brutal/expensive to program AI or is it that it's not as easy to market/sell AI rather than flashy graphics?

Do you have any example of recent breakthroughs in AI behaviour? I guess TLOU 2/MGS V has some pretty cool elements to them, but other than that i've yet to see anything impressive in recent years.

Sorry for the bad english, but i hope i get my point across.

 

[Putosaure]

You must save those CPU cycles I guess

 

[Deleted member 2254]

It's not something that comes off as exciting from fixed screenshots and trailers, so devs rarely bother.

 

[DiipuSurotu]

To be fair even natural intelligence hasn't advanced in the last decade

 

[Polk]

Well two things mostly.
Compute power is finite resource. You want better AI you have to sacrifice some other aspect of the game.

Also in most cases smarter enemies wouldn't result in better game. Smarter maybe, but not more fun to play.

 

[sponger]

Prices of experts skyrocketed.

I would say that best went to other branches of IT - I'm talking from personal experience.

 

[LumberPanda]

AAA games are power fantasies for men. If you have good AI then gamers complain that the game is unfair or bad.

 

[RelaxItsJustMe]

A big part is also that multiplayer games that don't really require much if any AI have become so much more prevalent in the last decade.

Why put a ton of effort into next level AI when you can just join all the other pubs and make another BR game!?

 

[Kittens McMittens]

People always cite FEAR as an example of good AI, but it's actually just an example of good design and presentation. I think the AI in FEAR only had like 3 states, one of which was just a specialised form of the other so actually more like two states. It's just the way it was presented to the player made it memorable.

 

[The Masked Mufti]

AAA games are power fantasies for men. If you have good AI then gamers complain that the game is unfair or bad.

Click to shrink...

I think it's this. I read that Riot had an advanced one play against one of the top teams at the time and it totally trashed them (no idea if it's true)

 

[teruterubozu]

What people consider AI is usually just convincing behavior animations.

 

[Fifstar]

It's tough to use ML to create fun ai.

 

[Reinhardt Schneider]

You can't show AI on screenshots

 

[Ewaan]

Jaguar CPUs most likely.

 

[7threst]

Good AI is tricky because games should be a fun experience worth revisiting over again. To have AI implemented in games that cater to that mind set might be a puzzle not yet solved.

 

[KushalaDaora]

I read something along the line on how making enemies too smart might make the game not fun for players.

 

[jsnepo]

Because open world?

I guess it's difficult to do a script that will take advantage of the entire map. FEAR works well because of the level design. It's small, filled with divisions and walls, things AI script can really take advantage of. If you played FEAR2 and FEAR3 which feature bigger and wider levels, the AI isn't as good as the first game.

 

[Galkinator]

Cause games will be too hard I guess

 

[HStallion]

The same reason you haven't seen a large leap in real time destruction physics and other "costly" game features. The general gaming public doesn't expect or demand for it and its not worth the time and effort to get right in most cases.

 

[Akira86]

the money

novelty of a fresh new intelligent AI that can't be used for anything else but the special use cases you put in the game, basically won't fit in any budget.
hard sell for any dev to get a publisher to even sign off to let them make that.

Plus it violates the cheap/easy/good law. It's inherently hard, and expensive, and easily goes bad.

 

[Parcas]

it costs a ton and the money last years has been on multiplayer games so who wants to invest that big on a secondary feature.

 

[Tavernade]

I want better AI for NPCs, personally. Especially in things like Animal Crossing or Open World games. That feels like a selling point more than enemy AI does.

 

[Skittles]

Smart AI frustrates players

 

[Scottoest]

1) AI is CPU-heavy, and the CPUs in the consoles this gen sucked.
2) Better AI isn't necessarily fun to play against. An Assassin's Creed game with actual smart enemies would be a miserable experience.
3) More complex enemy behaviour also probably means a lot more work in other areas - more animations, more ambient VO, etc.

#3 is the big one for me, and one I'd like to see focused on next-generation. There's nothing more depressing than hearing the same exclamation from a guard in a game for the 50th time. One of the things that people love about Bethesda's Fallout and Elder Scrolls games, is that the world actually has a clockwork to it, as janky and limited as it can be. People get up, go to work, go to the tavern, and go to bed. I want the next-level of this kind of clockwork, from next-generation games.

Make worlds feel like worlds, and not a tableau with a bunch of animatronics whose life revolves around the player character.

 

[Magic Mushroom]

Also in most cases smarter enemies wouldn't result in better game. Smarter maybe, but not more fun to play.

Click to shrink...

This.

 

[Ferrs]

People want better and more realistic until they play it and don't want it anymore. Also it's hard, it's not just about the AI, lots of other things surrounding the work needed for a better AI enemy.

 

[ps3ud0]

Prices of experts skyrocketed.

I would say that best went to other branches of IT - I'm talking from personal experience.

Click to shrink...

Pretty much it...

AI talent isn't attracted to the games industry.

ps3ud0 8)

 

[dean_rcg]

Graphics sell

 

[Parham]

The average player cares more about a game's visual presentation and gameplay mechanics than the AI, so it's hard to justify putting more time and resources into designing and implementing a robust AI system.

 

[Starlatine]

Because giving games good AI would kickstart the machine revolution that will enslave us all. Devs makes enemies dumb for the sake of all humanity.

On a serious note, smarter AI almost invariably means harder AI and people overall want less and less challenge these days. Chasing it is just not a priority if you want to maximize profits

 

[Stinkinmushroom]

There's some people out there that don't like TLOU2 out there and i actually think that the AI is already too advanced for them. People like to learn patterns and see weaknesses easily.

 

[TSM]

The explanation that I've always heard is that people think they want smarter enemy AI, but what they really want is more interesting enemy behavior. Making the enemy AI smarter tends to just make the game more frustrating to play.

 

[Ghostwalker]

People say they want a smart realist AI but they don't.

Think of it this way, pick your favorite multiplayer game and play it solo against a team of 3-5 if not more players just as good if not better than you are and know how to work together as a team and understand how to fully use the game's mechanics.

That is what playing against a smart realist AI would like.

Players want an AI that maintains the fantasy of the game.

 

[Aegus]

When 99% of people state that all they want from next-gen is higher res and improved textures then it's obvious that developers won't focus on AI.

It's a shame as proper AI would be insane in an open world game. Thousands of NPCs with their own AI running (think simple Westworld) with their own vague motivations in the game world. And the player could theoretically get involved with any one of those storylines with true branching decisions that the NPCs react too.

My dream game is one where the main enemy is running around the world in the same way as the player and none of this bullshit "I'm going to spend 100hrs doing sidequests before advancing the main plot".

 

[HStallion]

There's some people out there that don't like TLOU2 out there and i actually think that the AI is already too advanced for them. People like to learn patterns and see weaknesses easily.

Click to shrink...

I'll be incredibly vague but there is an encounter in TLOU2 that has an incredibly capable and smart enemy. One who is constantly on the move but checking choke points and looking into blind spots as they do, laying traps in their wake and making sure they're always on their guard so its very very hard to sneak up on them or use the basic techniques you have most of the game. Its an incredible combat scenario and I've actually wanted a difficulty mode where the regular enemies have similar AI and abilities. It would be a brutal experience with such capable enemies but it would be something really special in my opinion.

 

[Bartis]

Not sure if it qualifies as AI, but more cause and consequence would be a great start. For example Watch Dogs 2, where an ambulance came after a car accident. I'm thinking of AI that finds a corpse and keeps searching thorough instead of leaving it be after 10 seconds. People calling cops after witnessing a crime, whatever... Rockstar is probably best in this regard at the moment

 

[Vinx]

I think it's this. I read that Riot had an advanced one play against one of the top teams at the time and it totally trashed them (no idea if it's true)

Click to shrink...

It was Dota 2 best players in the world vs Elon Musk's "Open AI", and the players got stomped into the ground.

Elon Musk's 'Dota 2' Experiment is Disrupting Esports in a Big Way - No Playing Field

Elon Musk's artificial intelligence research company, OpenAI, is developing a self-learning bot for one of the most complex esports titles: 'Dota 2.' It has ...

www.youtube.com

People always think that advanced AI costs a ton of money or is too difficult to program but from most of the developer comments Ive seen they purposely dont pursue advanced AI because video games are still just games and games are supposed to be fun and be beaten.

 

[Aangster]

GMTK's video on the topic is worth a watch, especially on perceptions of smarter AI and what's simply 'smart' behaviours.

I'm more excited to see AI improvements to RTS like Age of Empires and strategy games such as Civ or Total War.

I actually think UI design/navigation/performance have stagnated this generation.

 

[Serein]

Just going to give a quick plug to YouTube channel AI and Games

 

[Bansai]

I'll be incredibly vague but there is an encounter in TLOU2 that has an incredibly capable and smart enemy. One who is constantly on the move but checking choke points and looking into blind spots as they do, laying traps in their wake and making sure they're always on their guard so its very very hard to sneak up on them or use the basic techniques you have most of the game. Its an incredible combat scenario and I've actually wanted a difficulty mode where the regular enemies have similar AI and abilities. It would be a brutal experience with such capable enemies but it would be something really special in my opinion.

Click to shrink...

If enemies were smart like the one in this encounter, games wouldn't be fun for a looooot of people.

Same with TLOU2, which is why most enemies are almost as blind as clickers.

 

[HStallion]

If enemies were smart like the one in this encounter, games wouldn't be fun for a looooot of people.

Same with TLOU2, which is why most enemies are almost as blind as clickers.

Click to shrink...

That's why I said I was hoping for a difficulty setting that specifically makes the enemies this smart, not the base game. Consider it something above even the Grounded mode difficulty aka the hardest setting in the game.

 

[Bansai]

That's why I said I was hoping for a difficulty setting that specifically makes the enemies this smart, not the base game. Consider it something above even the Grounded mode difficulty aka the hardest setting in the game.

Click to shrink...

I can certainly see that, but ultimately I think the cost of developing that level of AI working throughout the entire game vs the people who will actually play it that way is not viable to even bring it up for consideration.

 

[wild_one]

Let's extrapolate this to a logical extreme. Say a player says the game has fun and challenging foes and whatnot. Who's to say the next one says the same thing? Guess were we end up on that discussion. Everything is a cost/benefit analysis, so the cost to implement something that completely frustrates a userbase due to how unique the foes respond isn't worth the loss in revenue due to no one wanting to have to dedicate the time to learn the varied counters.

 

[Deleted member 8468]

I know this isn't what you're saying OP, but DLSS is an incredible example of advanced AI and what it can do.

Mimicking real human behavior is way more difficult than what we're capable with AI right now. Games are still using the same type of closed loops for AI they've used since the 90s. Most of the 'complex, good AI' is just writing out as many possibilities, as well as restrictions, in the code. It really boils down to 'how many possibilities exist for this AI" vs "how smart is this AI".

 

[Deleted member 27751]

AI was and has always been advanced in nature, its just that game design needs to take into account fairness to the player by not driving a wedge between enjoyment and difficulty. You can't have the enemy constantly outrunning and outgunning you without some give to the playability. Path mapping has helped a lot in organising AI diversion if obstacles are in the way or paths change, but otherwise its pretty standard as is. All depends on the designer.

 

[Cheesetriangles]

Basically you don't want games to have good AI because it wouldn't be any fun to play against.

 

[entremet]

There's a misconception that the majority of gamers want harder games.

 

[Dogui]

AI probably evolved a lot in the last 10 years, videogames just didn't use it.

Gamers definitely don't want smarter enemies.

 

[Servbot24]

Enemies having a degree of predictability makes games more fun.

AI probably evolved a lot in the last 10 years, videogames just didn't use it.

Gamers definitely don't want smarter enemies.

Click to shrink...

This.

 

[werezompire]

I remember with one of the Civilization games, the developers were really proud at how the AI acted more like a human player and players thought the game's AI was broken when other countries would break treaties and backstab them for personal gain. Players don't actually want advanced AI enemies - they want smoke & mirrors that makes it seem like the enemies are smart while still allowing the player to easily beat them (like in F.E.A.R).