During a conversation over one particular weekend, our beloved fellow columnist Gregg B announced to me that he had, during one particular Demon’s Souls boss encounter, utilized an artificial intelligence (A.I.) glitch to his advantage. Shame on you brother I said, without inquiring for the full story. However, it did get me thinking about A.I. within videogames…
Since the original Half-Life, I have only ever been impressed with another videogames A.I. once. Half-Life’s marine encounter is a gaming landmark, sadly however, even in Valves future endeavors A.I. has evolved little and at times has felt notably worse. What bothers me though is how critically, A.I. is seldom picked upon. Industry journalists and publications spend very little time discussing A.I. I can’t actually remember the last review I read where A.I. was even mentioned. The prevalent attitude (unless it is notably faulty) seems to be “If it’s functional, it’ll do”. Which to me, just wont do.
STALKER, the much imperfect game that it was, emphasized the possibilities A.I. could have within a videogame; the very nature of the open game world dictating and complimenting such possibilities. Nonetheless, the fact that a studio with very little resource could achieve such A.I. (when it was working) deserved significantly higher praise than was given. Far too many magazines at the time dwelled on the A.I.’s shortcomings, as opposed to praising the steps GSC Game World took in actually trying to breathe new life into the opposition players face.
At times STALKER’s A.I. was brutally brilliant. Being ambushed by a group of soldiers in the Zone was nerve wracking and exhilarating, and for the most part allowed the player to forget they were being attacked by computer opponents. Thankfully, I seldom encountered many of the games well documented issues. I encountered A.I. which I consider the benchmark of what developers should be striving for.
The question I have been asking myself recently though, is when did selecting a games difficulty determine player resources and A.I. hit points, instead of A.I .ability or mission structures?
When I choose a difficulty setting in a videogame I want to be challenged mentally, against foes who from the lowest difficulty to the highest, show a continuous marked improvement in the way they act in a combat situation. I want to be safe in the knowledge that I won’t be cheaply punished by the game for selecting a higher difficulty level. Providing me with less ammunition, increasing enemies health and giving enemies godlike accuracy may be a sure-fire way of making the game harder, but it also undermines a player’s ability. In COD: Modern Warfare 2, giving soldiers pin-point accurate grenades that are infinite in supply leaves the user feeling cheated, and subsequently draws you away from it’s immersion. Inevitably these idle developer methods do little but highlight you are in fact playing against a computer.
To use another example, the Combine in Half-Life 2, irrespective of difficulty, have poor A.I. From Easy to Hard, the only difference is the amount of resources the player is given, the amount of health the Combine have, and the amount of damage they deal. Their path finding, cover tactics, defensive or offensive routines do not change; I have tested this extensively over the last few days, and won’t even begin to discuss the awful sidekick A.I.
Now, is there value in playing at a difficulty which conspires to make the games core framework more challenging, but offers little narrative or cognitive reward in return?
While I understand the financial constraints in developing competent A.I. or adaptive mission structures, we are now regrettably being bombarded with the new Goomba. A series of enemies within varying games that have strictly established, simple A.I. patterns, who serve no other purpose than to act as cannon fodder. From Killzone 2 to Devil May Cry, the only challenge from these dull filler enemies is in their numbers, and not in their tactical ability. While supporting them, are a series of boss encounters that have a similarly limited array of predefined routines and skills that with practice and patience offer little in the way of challenge.
Demon’s Souls is the culmination of this developer approach, where A.I. routines are honed to such monotony that the simplest of boss encounters become treacherous (and torturous) as the game serves to punish you for the tiniest mistakes. Of course, for the most part it is through your own lack of repetition (or practice as some might refer) of the encounter that serves to undermine your personal ability. However, it highlights that instead of placing criticism on a game where A.I. patterns are so fundamentally simple, we instead praise its ingenuity.
Stepping away from A.I. routines and enemy skill sets, Thief was a perfect example of how difficulty should be adopted within a videogame. By making the rules for success more difficult through its mission structure, players were encouraged to adapt the way they played. By increasing the difficulty from Normal to Hard, the number of missions a player had to undertake were increased alongside available loot, and a restriction to killing humans was put in place. This fresh approach did not alter fundamental game mechanics (guard health, player resources, enemy numbers), but allowed for a stricter, more challenging game for those who wanted it. Crysis follows Thief’s approach, with harsher game parameters based on the difficulty you select; notably guards no longer speaking English and radar restrictions put in place.
So why is it then that developers continue to use an archaic difficulty method to prolong a games life span, when making use of many games existing mission structures and locations not only adds more weight to a games content, but serves as an incentive to challenge the player where the A.I. is fair, as opposed to the game creating a false sense of difficulty.
It would be unfair to request such A.I. changes and adaptive structures onto all games, as many do not offer the open world approach that Thief, Crysis or STALKER do. However, where financial resources are swallowed whole by development studios that desire to make their game the most graphically appealing, it is no wonder then that many titles released today feel like they have evolved very little.
Where Bayonetta is the pinnacle of the Ninja Gaiden and Devil May Cry evolution, honing combat to perfection and empowering the player with a wealth of skills and combat abilities that sway the user into believing they are almighty powerful, it is surely undermined by the predictable A.I. whose only purpose is to encourage such a feeling of empowerment. You are in fact a giant amongst ants, and nothing more. To address this balance with genres akin to Bayonetta, development studios should be placing more emphasis on refining what they already have, and strengthening the core experience the user has.
When a genre has arguably reached its potential and is trapped beneath the glass ceiling (racing games currently finding itself in such a position) why are developers still so quick to improve everything else besides the fundamentals of what makes a game enjoyable in the first place? As developers tinker away with subtle refinements that have little to no impact, when are they going to realise that in order to bridge the gap between player interaction and immersion, they need to place greater emphasis on autonomous A.I. encounters in order to challenge the player and create “wow!” moments that are talked about for generations, instead of relying on spit and polish to cover the cracks.
Email the author of this post at lewisb@tap-repeatedly.com.
My first thought when reading the title went to A.I. War the videogame.
http://www.arcengames.com/aiwar_features.php
@Hanamigi- AI War: Fleet Command is great and I intend on doing a first impressions when I get more time to settle into it. Incidentally, I’ve got Solium Infernum now so I’ll shoot you an email at some point.
@Lewis: it’s interesting you mention racing games because as a fanatic of F-Zero GX I can honestly say the AI was some of the most impressive I’ve seen in that genre. Certain characters were particularly good drivers and were usually hanging around the front of the pack while others were simply aggressive bastards so overtaking them became an additional challenge. Motorstorm: Pacific Rift, while not possessing particularly advanced AI, had suicidally aggressive drivers which made races far more interesting and tense.
I also recall the paramites in Abe’s Oddysee and Exoddus being particularly fun little critters to watch and pay attention to. They were very timid on their own but when they were cornered or in a pack they were fearless much like those ‘dog eye’ things in the original Half-life. I enjoyed Bioshock’s and Dark Messiah’s AI because enemies fled. I know that’s a very basic thing but it certainly gives enemies another dimension.
More than anything it’s making enemies have behaviours that make sense and are different; hiding, fleeing, coming back in numbers, waiting for you to come out and not running at you head on. I remember in Dark Messiah those little goblins running away, hiding and getting down on their knees, begging for mercy when you’d given their friends a good hiding. It made them really interesting and fun adversaries.
@Lewis: “why are developers still so quick to improve everything else besides the fundamentals of what makes a game enjoyable in the first place?” Come on, because the mon€y$ that casual and new gamers give them. I repeat myself, but Indies are and always will be the future, and with indie I mean a mentality that decides to break free from greedy logics by risking a lot in order to refresh, renew and re-evolve a genre.
Amen, brother! That’s been bugging me for years. I could give a fuck about improving the graphics any further–they already look great, thanks. Instead, lets have developers channel all that energy into better A.I. and better game design.
@ Gregg, Motorstorm: Pacific Rift is probably the only racing game I’ve played where you genuinely feel challenged, but not cheated by the other drivers. It has a level of competitiveness that doesn’t see you coasting miles ahead of the pack, but then you don’t feel that your opposition is being helped by driver “catch-up”.
Did you know the original Houndeye in Half Life was intended to accompany the player and help them? But during original player testing people just killed them, so Valve implemented the cute/follow me characteristic before the shockwave that would hurt the player. Fleeing enemies is still very basic and has been around since Half-life’s first marine encounter, but it certainly helps.
@ Hanamigi, I understand your sentiments (and being assaulted by Gregg to look into indie games more) its evident that they honestly are pushing the boundaries of innovation within video games, however the money that casual gamers give developers should be invested into their games A.I. routines. There is really no excuse. If developers can waste countless quantities of money continuing to rev up graphics within video games, and add gimmicky extra’s in an effort to extend a games life span (God of War 3 allows you to get new outfits for the protagonist once you have completed the game! Woohoo!) why can they not place more emphasis on A.I.?
Although its probably because within the industry and with the general buying public, simply don’t want more complex A.I.
@Lewis: You got it, even if I wasn’t very clear in my statement, which is that most of the gamers are now casual and-or “newfags” which have grown with an industry that puts the focus on graphics and coolness instead of actual gameplay. This illusion makes people feel satisfied with their fu*king eyecandy stuff, but always wanting more and more graphical performance.
I believe that this happens because eyecandy and s*it can only temporarily satisfy someone, then after two hours the user gets bored but, instead of desiring better gameplay, he craves for shiny stuff, bump mapping and hi-res because that’s what he has been taught.
It’s the ideal of individualism and maximised earning, where you must feel instant satisfaction which can be achieved by crappy AI that gets easily killed and shiny HD, while publisherd kinda force developers to use the Copy/Paste technique with “successful” but dated gameplay and focus on these two mildly useless aspects.
Great article, Lews. (Thanks rotating banner thing!)
I don’t think A.I. has ever surpassed the level it reached in Half-Life, frankly. That game can spank every CoD combined in the intensity department.
I also love the mention of difficulty in Thief modifying the player’s objectives and tactics, rather than the amount of hit points their opponents have. That needs to happen more in games.
Half-Life 2 is also surprising in that the Combine A.I. didn’t improve upon that of the marines in HL1. Not even close, actually. You could never just kill a marine by firing off 4 pistol rounds into their face at point blank range. Combine, however . . .