Putting the 'role' back in role-playing games since 2002.
Donate to Codex
Good Old Games
  • Welcome to rpgcodex.net, a site dedicated to discussing computer based role-playing games in a free and open fashion. We're less strict than other forums, but please refer to the rules.

    "This message is awaiting moderator approval": All new users must pass through our moderation queue before they will be able to post normally. Until your account has "passed" your posts will only be visible to yourself (and moderators) until they are approved. Give us a week to get around to approving / deleting / ignoring your mundane opinion on crap before hassling us about it. Once you have passed the moderation period (think of it as a test), you will be able to post normally, just like all the other retards.

Has Artificial Intelligence in RPGs Stopped Evolving?

Has RPG AI stopped evolving?


  • Total voters
    63

epeli

Arcane
Joined
Aug 17, 2014
Messages
719
The last significant AI improvement in games happened decades ago, when real-time A* pathfiding became commonplace. It's all just the same state machines. Some with slightly better/worse rules than others. There are probably few experimental games with interesting AI developments, but I can't think of any RPGs like that.
 

Ezeekiel

Liturgist
Joined
Dec 19, 2016
Messages
1,783
A.I. has long been a weak point in tactical RPG's... And just about everywhere else. Modders seem to be the only ones doing anything about it.

It's one of the old weak-points that no dev seems to want to actually improve upon. Game makers are more interested in making sure their NPC's have schedules than improving the bread and butter stuff first.
Same goes for 4X games for example.
 

LordofSyn

Scholar
Joined
Nov 8, 2014
Messages
113
A.I. has long been a weak point in tactical RPG's... And just about everywhere else. Modders seem to be the only ones doing anything about it.

It's one of the old weak-points that no dev seems to want to actually improve upon. Game makers are more interested in making sure their NPC's have schedules than improving the bread and butter stuff first.
Same goes for 4X games for example.
Please check out Trese Brothers games on mobile game stores and on Steam. Not only are their games tactical turn based rpgs, but they are continuously tweaking their AI.

Sent from my LGLS996 using Tapatalk
 

Azarkon

Arcane
Joined
Oct 7, 2005
Messages
2,989
AI technology as it currently exists is too much work, for too little gain, in the gaming industry. When most people outside the industry talk about AI in games, they talk about shit like characters that can pass the Turing test, West World style, but the simple fact is that the technology's not even close to achieving any of that. At best, current AI technology, should you invest in it, can give you more optimal decision making on the side of computer opponents = more challenging enemies. But it turns out most people don't want more challenging enemies, and you can achieve the same effect, in any case, just by allowing the computer to cheat or by adding more enemies. So what's the sell for better AI?

I think strategy games are much more in need of better AI; but even there the commitment is low because the market is limited. CRPGs, with their story driven, scripted design, will probably be among the last genres to adopt improved AI.
 
Self-Ejected

IncendiaryDevice

Self-Ejected
Village Idiot
Joined
Nov 3, 2014
Messages
7,407
Oh yeah, one last point I wanted to make in any of my posts on Blackguards1 but always managed to forget is that the AI in Blackguards is actually a lot smarter than one might initially think. Not that AI's are smart, but that the developers have actually given enemies different levels of intelligence with regards to battlefield obstacles.

One of the most unique features of the game is the ability to use environmental objects to your advantage. In the early stages of the game, for example, you can kill an enemy simply by lighting a box on fire, standing behind it, and the enemy will lurch at you, stand on the flaming box while it attacks you, then burn to death without you even needing to lift a finger. By the end of the game, however, the mini-boss-like enemies will actively walk around these burning boxes. Excellent stuff, and stuff that a lot of people might miss.
 

Ebonsword

Arcane
Joined
Mar 7, 2008
Messages
2,326
NPC non-responsiveness will be total shit. Companions disobeying orders mid battle will be terrible and it will just make you play solo.

[khalid]Better part of valor! Better part of valor![\khalid] :D

Seriously, though, maybe focusing on improving companion AI does make more sense than focusing on enemy AI.

Dragon's Dogma is kind of an interesting example. At first, you think that your pawns are just brain dead, but then you slowly realize that they are ignorant, not stupid.

...okay, they may be a little bit stupid. ;)

Still, it is pretty impressive how good a well-trained pawn can be. I've seen Ranger and Strider pawns basically solo Golems because they know how to effectively exploit their weaknesses.
 

fantadomat

Arcane
Edgy Vatnik Wumao
Joined
Jun 2, 2017
Messages
37,087
Location
Bulgaria
NPC non-responsiveness will be total shit. Companions disobeying orders mid battle will be terrible and it will just make you play solo.

[khalid]Better part of valor! Better part of valor![\khalid] :D

Seriously, though, maybe focusing on improving companion AI does make more sense than focusing on enemy AI.

Dragon's Dogma is kind of an interesting example. At first, you think that your pawns are just brain dead, but then you slowly realize that they are ignorant, not stupid.

...okay, they may be a little bit stupid. ;)

Still, it is pretty impressive how good a well-trained pawn can be. I've seen Ranger and Strider pawns basically solo Golems because they know how to effectively exploit their weaknesses.
Most games have the same core AI for all the npcs. Enemies are downgraded companion AI,downgraded because they have less active abilities. It is good to have capable AI.
 
Joined
Sep 18, 2013
Messages
1,258
For AI to have a meaningful impact beyond combat intelligence in RPGs, games will need to start experimenting with emergent narratives with characters/group making and showing data/personality-based decisions and behaviours on the story side. Nothing like complex NPCs free-moving around in a persistent game world as entities and making decisions but rather, dynamic NPC profiles with dynamic story nodes that branch emergently, but otherwise still phase in and out of "existence" as PC interacts with the world.

Current data-mining AI softwares operate on a similar premise. They mine data, create profiles and predict behaviour patterns and flag those that are deemed security threats / marketing opportunities / whatever. They just need to be adapted to PC+NPCs+Narrative paradigm.
 

conan_edw

Arbiter
Patron
Joined
Dec 3, 2017
Messages
846
Grab the Codex by the pussy Pathfinder: Wrath
Why would people with strong understanding of AI stay on video games industry anyway :D
 

Damned Registrations

Furry Weeaboo Nazi Nihilist
Joined
Feb 24, 2007
Messages
14,982
Dragon's Dogma is kind of an interesting example. At first, you think that your pawns are just brain dead, but then you slowly realize that they are ignorant, not stupid.

...okay, they may be a little bit stupid. ;)
Dragon's Dogma is a pretty great example of how people don't understand AI actually. DD actually has some remarkably good AI- but the characters acting on it don't have goals you want them to have. The pawns aren't running around picking up corpses and gathering items in the middle of a dangerous fight because they're stupid, they're doing it because that's their goal- that's what they're SUPPOSED to do. If you train them to do nothing except shoot people in the eyes with flurries arrows and spam consumables to maintain that behaviour while dodging every incoming attack, they're incredibly powerful killing machines way better than the player can be. But the average players wanders around, whiffs shots, and gets hits by things, so the pawns copy the player.
 

Darkzone

Arcane
Joined
Sep 4, 2013
Messages
2,323
Why would people with strong understanding of AI stay on video games industry anyway :D
Money? If there is any. With a good understanding of AI and its application you can nearly only work in a university in computer science if you want to work with / towards AI. And even then it depends where you main specialization is.

There has been games where the is a true AI, like Superpower 2 that learned the specifics of the gameplay of the player. But what here is refered to is not in the understanding a classical AI, which is the strong AI and or weak (narrow) AI .

I think human intelligence in RPG developers has stagnated.
Ohhh you don't even know how right you are. All inteligence that is put into the decision scripts of agents (NPC) is human and is basically not destinguishable from the engine itself in its true classification.

The last significant AI improvement in games happened decades ago, when real-time A* pathfiding became commonplace. It's all just the same state machines. Some with slightly better/worse rules than others. There are probably few experimental games with interesting AI developments, but I can't think of any RPGs like that.
Yes, but... Dijkstra algorithm is the most cost effective search for the shortest path in a graph, but it is not a necessary basis. A* requires specified nodes and weighted edges, therefore "potential field" path planning are used when such a graph does not exist and cannot be made.

I always cringe at the use of the word AI in this topic, because of its distorting meaning of what is understood as AI in the classical sense. AI was always associated with a learning process (machine learning).
A harder definition states that intelligence is the ability to introduce and expunge known facts or decision rules in the database of an agent about its environment. Reasoning is the part where the intelligence is in the AI. RPGs have only action- reaction scripts and that it. And this is only required for a cRPG, everything else is an unnecessary expenditure of resources.
If someone here is really interested in the AI then i recommend to learn it with Prolog and Ivan Bratko's "PROLOG Programming for Artificial Intelligence" as a basis for understaning of facts and rules in a logic based system.
 

Zanzoken

Arcane
Joined
Dec 16, 2014
Messages
3,559
The problem is that enemies are viewed as "things for the player to kill" instead of independent agents in the world who have their own goals and motives. I think if you are going to place something in the game world, it should have a purpose -- not just to the player experience, but within the world itself.

AI is a component of that, along with things like encounter design. The question devs should be asking is, given what we've presented about this enemy, how would we expect it to behave?

However in RPGs this behavior is typically uniform across all encounters, which is obviously wrong. Consider the difference between say, a mindless zombie whose only desire is to ravenously feed itself, versus a bandit who is motivated by coin but also keenly interested in self-preservation. You would expect the zombie to keep attacking until destroyed, but the bandit would more likely than not give up and beg for mercy when defeat became inevitable.

Anyway not to get too long winded here -- you probably get the point. The takeaway is, spend as much time thinking about how the bad guys see the world as you do the good guys, and everything else will flow logically from that.
 

huskarls

Scholar
Joined
Aug 7, 2016
Messages
108
Most of the problem is a lack of playtesting and hand placed encounters. I don't even know what a 'dynamic' AI is and a lot of people here are demanding better AI without specific behaviors listed such as pathfinding or targeting. Each encounter/area just needs its own quirk, like there's red stuff you can't stand, or poison, insanity spells, or whatever that the player has to figure out then move on. The AI just needs to do what it is supposed to do, and I imagine a great deal of time has been wasted from some project heads saying they want their AI to be 'alive' which results in conversations about mudcrabs. When I played pillars/wasteland if you played through one fight, you basically saw all the fights. It didn't matter that the AI was more sophisticated in determining attacks of opportunity with its path finding or able to identify and use cover. The AI in interplay games was pretty stupid, friendly fire, pathfinding issues, using inferior abilities and items, and most of your nostalgia from these games is from how these enemies were placed with what abilities, all done by people and not sentient machines.
 
Unwanted

Micormic

Unwanted
Joined
Mar 25, 2009
Messages
939
If anything I feel like AI in games has gotten substantially worse in the last decade or so.
 

Silva

Arcane
Joined
Jul 17, 2005
Messages
4,778
Location
Rio de Janeiro, Brasil
Is there an RPG where the player competes against other agents with similar objectives , needs and behaviours that could finish the game in his place? That would be a formula worth of developing AI for. (The nearest I've seen a game get to it is Stalker with it's ALife but even so it's limited in a number of ways.)

The truth is that the traditional "format" of tabletop RPGs where a group of players form an adventuring party of superfriends expected to go through the GM story and save the day at the end, is a pretty crap source of inspiration for producing an emergent and dynamic electronic analog. A better model would be a competitive one where each player assumes a different character against each other. Or, at least, one where GM and player compete against each other with fair rules, and there is no set expectation that the story must have a happy ending and a player succeed.

Oh, and for it to be worth any effort, it must be in an environment where failure is accepted and there is a myriad ways to succeed. If save-scumming continues being a thing, I don't see why anyone would have genuine interest in improving AI.
 
Last edited:

gaussgunner

Arcane
Joined
Jul 22, 2015
Messages
6,151
Location
ХУДШИЕ США
The problem is that enemies are viewed as "things for the player to kill" instead of independent agents in the world who have their own goals and motives. I think if you are going to place something in the game world, it should have a purpose -- not just to the player experience, but within the world itself.

That can be done with scripting alone, and it's still incline. AI wouldn't necessarily improve it; AI would just be navigating a graph of dialogue and waypoints and shit to reach the NPC's nearest goal. Sounds hard to quantify. Might be easier to implement than scripting every possible outcome. Not as easy as railroading ofc.
 

Pizzashoes

Scholar
Joined
Oct 31, 2017
Messages
444
Sword Coast Stratagems evolved Baldur's Gate 2's combat more than a decade ago. But that was made by a computer scientist in his spare time. The AI cheats but it mostly uses more intelligent programming. Playing through the game makes you feel like you're up against maybe 16 intelligence wizards instead of 10 or 9. But I imagine that video game companies like Obsidian do not want to pay top dollar for good programmers.

The potential to create truly reactive AI that scales, matches, and pushes players' skills to their limits exists. In RPGs, shooters, strategy, whatever. We'll see it eventually.
 

Silva

Arcane
Joined
Jul 17, 2005
Messages
4,778
Location
Rio de Janeiro, Brasil
That can be done with scripting alone, and it's still incline. AI wouldn't necessarily improve it; AI would just be navigating a graph of dialogue and waypoints and shit to reach the NPC's nearest goal. Sounds hard to quantify. Might be easier to implement than scripting every possible outcome. Not as easy as railroading ofc.
There, you just hit it in the head. The premise that informed CRPGs was a railroad, for good or ill. And that's the reason we don't see AI in this genre. The expected environment is a frozen world waiting for players to come into viewing distance to wake up. What should an AI do in this? Meditate on the bleakness of life as a mook to be repeatedly slain by heroes?

:hmmm:
 

Iznaliu

Arbiter
Joined
Apr 28, 2016
Messages
3,686
After the bullshit that was "3D printing" (wonder why we aren't hearing about that anymore?)

3D printing doesn't have that much to do with an AI; sure, optimisation techniques to prevent printing deficiencies might be somewhat AI-like. The main reasons why 3D printing has failed to expand outside the hobbyist market are because prices have failed to come down sufficiently, there are too many options out there, causing customer confusion, and the software/modelling side of things is still unintuitive for the average person.
 

laclongquan

Arcane
Joined
Jan 10, 2007
Messages
1,870,144
Location
Searching for my kidnapped sister
Is there an RPG where the player competes against other agents with similar objectives , needs and behaviours that could finish the game in his place? That would be a formula worth of developing AI for. (The nearest I've seen a game get to it is Stalker with it's ALife but even so it's limited in a number of ways.)

The truth is that the traditional "format" of tabletop RPGs where a group of players form an adventuring party of superfriends expected to go through the GM story and save the day at the end, is a pretty crap source of inspiration for producing an emergent and dynamic electronic analog. A better model would be a competitive one where each player assumes a different character against each other. Or, at least, one where GM and player compete against each other with fair rules, and there is no set expectation that the story must have a happy ending and a player succeed.

Oh, and for it to be worth any effort, it must be in an environment where failure is accepted and there is a myriad ways to succeed. If save-scumming continues being a thing, I don't see why anyone would have genuine interest in improving AI.


There is a game close to what you are describing , Depths of Peril

You as player appear in a world. There's 4 guild of adventurers for you to choose. Once you choose one, the other become AI-driven.
The AI guild send out adventurers to complete quests in the area, competing with you on the way.
When they get to certain levels thank to quest reward, they attack each other.
Once you complete the main game or only guild left, you win that area and export to the the next world.
 

almondblight

Arcane
Joined
Aug 10, 2004
Messages
2,549
Knights of the Chalice was the last cRPG that reflected some awareness among enemies and adjusting their tactics.

Yep. One of the few games where the enemies wouldn't mindlessly run through a chokepoint and get mowed down one by one.

It's worth pointing out that a lot of games make the AI bad on purpose. I think there was an interview with the BG1 team a while back where they said they originally had kobolds in the mines targeting your mages with fire arrows, but people got angry that their mages where getting killed in one hit so they had the AI go after the tanks. Battletech is another example of this - the modders who have been tinkering with things have said that the base game AI is designed to be bad on purpose.
 

Latro

Arcane
Joined
Jun 5, 2013
Messages
7,345
Location
Vita umbratilis
Knights of the Chalice was the last cRPG that reflected some awareness among enemies and adjusting their tactics.

Yep. One of the few games where the enemies wouldn't mindlessly run through a chokepoint and get mowed down one by one.

It's worth pointing out that a lot of games make the AI bad on purpose. I think there was an interview with the BG1 team a while back where they said they originally had kobolds in the mines targeting your mages with fire arrows, but people got angry that their mages where getting killed in one hit so they had the AI go after the tanks. Battletech is another example of this - the modders who have been tinkering with things have said that the base game AI is designed to be bad on purpose.
that's less to do with AI and more to do with ranged being generally OP in turn based strategy games
 

T. Reich

Arcane
Joined
Apr 15, 2013
Messages
2,714
Location
not even close
Doesn't Space Rangers 2 have various factions with competing agendas acting independently of you and each other? To the point of you losing the game (or "losing" because and AI faction "won" it) if you get sidetracked for too long?
 

Raghar

Arcane
Vatnik
Joined
Jul 16, 2009
Messages
22,506
The last significant AI improvement in games happened decades ago, when real-time A* pathfiding became commonplace. It's all just the same state machines. Some with slightly better/worse rules than others. There are probably few experimental games with interesting AI developments, but I can't think of any RPGs like that.
LOL I did stuff like psychomodeling, personality based fuzzy decisions. Stuff like that. It would be boring when every opponent use the same optimal tactics.



A* fails when all entities are able to move simultaneously as you can see in Pillars of Eternity when a paladin is running around enemies in circles instead of attacking closest.
 

The Bishop

Cipher
Joined
Oct 18, 2012
Messages
359
Chess computers show that computers should be able to handle things on a tactical level.
The big problem is AI on a strategic level. The best I've seen there is the old Master of Orion.

In computer games the majority of players prefers an entertaining AI over a purely competitive AI tough.
But that's the thing though - MoO was probably the most fun I had in a 4X game, maybe even in any game. And I don't even think AI was incredibly competitive, but it responded sensibly to challenges it faced, and created that organic environment that was so fascinating to just observe, but even more to participate and tamper with.

I think it all changed when games in general shifted towards checklist gameplay, requiring players to perform action X and repeat it N times, then move to action Y and repeat it M times. Games like that don't require whole lot of AI. Good AI in this style of game will only make it more difficult, and you want your games to be beatable with sensible effort. So the real state of the art AI programming more or less disappeared from the industry. AI part has become more of a routine task within a game development project, no creative breakthrough expected. And so, this is where we are right now.

Chess computers show that computers should be able to handle things on a tactical level.
The big problem is AI on a strategic level.
And Go computers show that AI is already pretty damn capable on strategic level as well. Fun fact - the guy behind Alpha Go was making video games at some point. But then he realized that AI is a hard sell in video game industry...
 

As an Amazon Associate, rpgcodex.net earns from qualifying purchases.
Back
Top Bottom