on 10.21.2013 (server time)Artificial Unintelligence
Reply via cblogs
I'm an avowed Total War nut. There's few things I find more entertaining than forming a vast horde to roll across Europe, dispatching every enemy in my path. Even now, some thirteen years on from Shogun, developer Creative Assembly's first release, no other series really scratches that itch for large scale, opening-scene-from-Gladiator-style warfare. You can imagine then, how much I was looking forward to Rome II, the sequel to Total War's finest hour- apart from that TV show with Richard Hammond in it.
Then I played it.
It's a mess, in many different ways. It wasn't the variety of technical issues or the half-hearted new gameplay mechanics that stood out to me, however, but the frequently idiotic moves made by my computer controlled foes. I've been playing Total War games for about ten years or so now, and I'm used to Creative Assembly's familiar habit of leaving the issue of battle and strategic AI to the last possible minute, to the extent that titles like Medieval II actually relied on fans to do most of the work making the enemies, if not whip smart, then less desperately stupid.
It's not that I'm entirely unsympathetic. Creating any video game is a mind-bogglingly huge task, let alone one the size of Rome II, which tries to depict the rise and fall of the Roman Empire in its entirety. I can totally understand how concentrating on fancy graphics, buzzwords like 'emotion' and neat-sounding mechanics must seem an easier sell than a man droning on about how good the computer is at moving their armies about. Of course, it's exactly this unglamorous aspect of the game that experienced Total War fans are so vocal in their desperation to hear more about, which makes it all the more frustrating when a new release rolls around with the same AI problems the series has had since its inception.
Priorities, people. This is a game about war. To make a game about war, it's generally advisable to have enemies who don't respond to every threat by charging around in circles like a dog with concussion. Rome II is symptomatic of a recent trend towards vacantly pretty games populated by heavily armed morons, that mask their lack of complexity behind a smokescreen of special effects, marketing jargon and bloated production budgets. Given the opportunities afforded by modern processing technology, I'm continuously baffled by the lack of concern given by developers towards developing competent, adaptable opponents.
I've been thinking about the subject a lot recently, especially in light of the upcoming generation of consoles, and the priority developers seem to place on it. At one point it was all the rage. Half Life prided itself on its brigade of hard as nails marines, who would take it in turns to pepper you with assault rifle rounds and grenades, while you wept and prayed for those halcyon days where all you had to worry about was a nightmarish alien organism wearing your face for a hat. Games like the Metal Gear Solid series and F.E.A.R were hailed for their realistic and adaptive enemy behaviour.
Metal Gear Solid 2's search-and-clear squads in particular were a revelation, hunting you down relentlessly through the corridors of the Big Shell facility. More than the impressive graphics or the ludicrous storyline, the moment I remember most from MGS2 was crouching panicked in a closet while an armoured response team slowly cleared locker after locker, hoping desperately that they'd be called away before they found me. Here was an enemy force that had a plan, a routine that forced you into improvisation; there were no easy outs when the alarm sounded, very few blind spots to exploit. They checked under crates, opened lockers with guns held ready and recognised when they were being picked off. They felt real.
One reason F.E.A.R's achievements in this area were so noticeable was that skirmishes in that game typically involved smaller groups of enemies in more open, navigable environments. It wasn't a non-linear game by any means, but most combat situations had some kind of verticality or separate paths to approaching a gunfight, options which the enemy would utilise against you as often as the other way round. It's easy to cover up short-comings in enemy intelligence when you're wandering down a set corridor machine-gunning eight hundred disposable goons into Swiss cheese.
Call it the Call of Duty approach, where the constant forward motion of the scripted thrill-ride spectacle distracts you from the stupidity of your enemies, who typically either kneel behind a box and refuse to move even if you're clearly creeping round to shoot them in the head, or otherwise simply charge straight at you. Call of Duty's resolute adherence to a rigid gameplay formula that prioritises scripted set-pieces and overwhelming enemy numbers over smart and adaptable AI gives the illusion of challenge, the player pushed slowly forwards via trial and error, buttressed by generous use of checkpoints.
In contrast, fights in F.E.A.R stuck with me because they gave me time to think and plan, and space for the enemy to completely screw me with a well-placed grenade from two storeys up. Combined with the lack of regenerating health and smart use of bullet time, F.E.A.R rewarded both quick-thinking and player skill, while heavily scripted series like Battlefield, Call of Duty and the like play more like wars of attrition. Especially when seven hundred things are running at you and quick get on that machine gun and defend the APC and oh no you're dead from grenades try again.
It feels like an unwillingness to truly test the player, born from a fear of frustrating them and therefore risking the colossal audience that these series currently enjoy. Developers have astonishing budgets for actors, special effects and musical scores, and by the God Emperor they want everybody to see them. Thus, the single-player campaign becomes a theme park ride with the requisite explosions, slow-motion breaching sequences and patriotic dick-stiffening, and it's multiplayer where you're directed if you want to test yourself.
Shooters aren't the only culprits when it comes to cutting corners with AI. Company of Heroes 2 is technically accomplished game, but I was disappointed by the willingness of the computer, particularly in the single-player campaign, to sit back and wait for you to regroup rather than pressing an advantage. It felt like the game was taking it easy on me. Civilization V takes its single player challenge far more seriously, and has been markedly improved by its recent expansion, but even there you can see the seams, the moments when a lack of computer intelligence is covered up by the game granting enemy civilisations an unearned bonus. Dragon Age II abandoned its predecessor's more strategic fights in favour of continuous waves of suicidal minions hurling themselves at you.
The problem with creating a rigorous AI, and especially with marketing such a feature, is that there's rarely a quantifiable visible improvement. Instead you need to experience it yourself to get a feel for it, whereas if you're just looking for shinier graphics or fancy environmental destruction, you can whack that in a screen-shot and people know exactly what you're selling. Hence also the current trend of prioritising multiplayer, where AI performance becomes a non-issue.
It's a worrying trend to see, going forward into a new console generation. Check the wide range of videos publicising the next generation's first batch of big titles- few of them talk explicitly about improvements in artificial intelligence, certainly compared to the time spent extolling the virtues of a persistent social space, big explodey building physics or improbable combat dogs. There are developers, like Naughty Dog and *gasp* Ubisoft, making the right noises, but then we've heard that before. Remember Bethesda lying about how ground-breaking Oblivion's 'radiant AI' would be? Then you played it to find that the computer intelligence amounted to an orc with a face like the Stay-Puft man dissolving through a door every six hours.
Playing a video game without competently programmed competition is a frustratingly empty experience. If you're not challenged by a game's system (and challenged in a fair way, I'm looking at you and your instantly swarming hordes, Shogun II) then I'd argue you're being deprived of much of what makes video games a uniquely engaging medium. That's why my Rome II campaign felt so sterile and boring as I half-heartedly marched across Europe. There was not a single story worth the telling, despite the sweeping landscapes and cinematic battles. Technology continues to advance at such a dizzying speed, bringing new advances in what studios can achieve, yet there's a definite stagnation in this core area of game design.
Which is why I can't help but sigh a little whenever a new shooter has a snazzy trailer with lots of shots of destructible environments, or when Dragon Age: Inquisition happily announces the use of the next Frostbite engine as a major selling point, rather than a focus on creating a more challenging, intelligent combat system. It's not that I'm against these pyrotechnic displays, far from it, but rather that I increasingly get the feeling developers are getting carried away with the cool, shiny things they can build rather than getting down to the often unsexy work of creating realistic, believable opponents to outsmart.
Please report harassment, spam, and hate speech to our comment moderatorsCan't see comments?
Anti-virus apps like Avast
or some browser extensions
can cause this. Easy fix: Add
to your security software's whitelist.