Next-Gen Games Should Ditch Dumb AI And Embrace Complex Systems
by Edward C. Ames, Gaming Editor
Published in Gaming on 20/10/2019
Faster hardware is leading to a revolution in game development, as we saw at the E3 conference this year. A new video game console launch is a rare thing in the industry, but with it comes more powerful technology to allow game creators to make games with more powerful engines and interesting new features. We talked to several developers to learn what they have up their sleeves.
The current generation of games is recognized mostly by their visual improvements. It's also infamous by the awful AI and superficial open-world design present in most AAA games. That should change with next-gen consoles.
The previous generation of consoles brought HD visuals and created a groundwork for the rise of massive open-world games as well as the games-as-a-service model.
This generation was all about those massive open-world titles that pushed visuals even further along with the maturation of GaaS model. The visual scope of AAA games brought increased costs and development time, creating video games with development costs doubled or tripled compared to the most expensive Hollywood blockbusters.
The next generation, with beefy CPUs and super-fast SSD storage (which will be much faster than on an average PC and will be able to play a role of a secondary RAM, but that's not the focus of this piece) will create more cohesive worlds that can be traversed at super speeds with zero loading.
In other words, the next generation will be all about iterative upgrades instead of major breakthroughs. Zero loading screens, high frame rates (in most games) true 4K gaming as a new standard.
These upgrades won't make games even more expensive to make since doing that would launch development costs of the most expensive titles into 1 billion-plus area - some estimations put total development cost of Red Dead Redemption 2 at $950+ million, an insane number that is extremely hard to cover with sales, even if you mar your game with microtransactions and other forms of "supplementary revenue streams." In other words, improving visuals can only go so far before hitting both the monetary and human resource wall. You are limited by the size of your development team, and creating bigger teams doesn't mean increased productivity.
And until AI takes the helm of video game development (which won't happen during the next generation of consoles) we will be limited by money and manpower. In other words, the next generation of games will bring better visuals suited for 4K resolution but don't expect radical changes in graphics. So, instead of "next-gen graphics" next-generation should focus on improving other facets of video games. And one of the most malnourished video game features of the last decade is the enemy AI.
Games like the first Halo, F.E.A.R, and Black and White had AI routines that are better than in 99 percent of modern titles, which is really sad when you think about it especially because those routines were also based on simple rules found in most other games. The only two titles that brought superb AI in recent years that aren't mentioned in the aforementioned piece are Horizon: Zero Dawn with its impressive and varied behavior of robotic enemies and The Division 2 to a lesser degree since enemies are superb in closed environments while they completely break when faced against outside, in the streets of Washington D.C.
Avoiding bad AI
While literally billions are spent into creating huge virtual worlds that get prettier with each passing year the investment in better AI is virtually zero cents. And since next generation of consoles won't create radically better graphics, creating better AI could bring improvements that won't be visible on the surface but that will be capable of making video games of the future massively more enjoyable experiences.
But I don't talk about some advanced AI, like Google's Deep Mind that managed to beat the world champion of the ancient game Go or that gave headaches to the best Starcraft players. No, that kind of enemy AI would be too punishing and too unpredictable for even the best players, let alone the majority that's average at playing games.
And let's not forget that machine learning-based AI is extremely limited at the moment and breaks under its weight with even the most minuscule contextual changes. What I'm talking about is developers ditching dumb AI and embracing enjoyable AI that only seems like it's intelligent (found in games such as F.E.A.R and Halo) and complex AI systems that govern virtual worlds we experience while playing open world video games.
The first task is fairly easy to accomplish. Instead of letting the development of the enemy AI take the backseat during the development of an average game (most times being pushed to the trunk, or even being dragged behind the car) developers should focus on actually creating compelling AI routines. It's sad to see more and more games offering virtually nonexistent AI thanks to the fact that all development money went on fancy visuals or famous voice actors.
For instance, enemies in the Wolfenstein II are dumb as they get, often completely breaking down and taking cover on the side that's faced to the player, standing still and letting the player shoot at them, or just running around in circles. Then we have cheap mechanics with extremely accurate grenade throws seen in so many modern shooter games. Or enemies that simply rush at you without any context, without any advanced strategy like flanking or creating tactical ruses. These practices make video games less enjoyable and once you notice the shortcomings you aren't able to ignore them.
Both of aforementioned strategies were present in F.E.A.R, development of which operated with figures that seem minuscule for AAA titles of today. And implementing them was so simple; just make enemies shout their intentions before performing actions like flanking and throwing grenades and create AI routine that won't break mid-fight by simply testing it a lot.
Just create a compelling enemy AI that seems capable and intelligent because once players get saturated with pretty lights and 4K textures they'll start noticing just how enemy AI is lousy and how it makes games less fun and engaging. And that is what will surely happen during the next generation of consoles, sooner or later.
And don't think that the mighty pull of looter shooters gameplay loop that plays with our psychology by constantly activating our dopamine receptors will last forever. Once gamers become overwhelmed with similar experiences - all of them based on the same principles of randomized rewards and operant conditioning - they will crave for something more, and better AI might be the change they'll look for.
Creating complex systems governing virtual worlds that allow emerging gameplay, complex and intimate stories, and tangible consequences for our in-game choices.
The other way to create compelling and engaging gameplay experiences is creating open worlds governed by complex systems. This isn't anything new. Games like Red Dead Redemption 2, The Legend of Zelda: Breath of the Wild, RimWorld, GTA V, The Witcher 3, S.T.A.L.K.E.R titles all have complex open-world systems that create massive levels of enjoinment that fools us, the players, into thinking we're part of a living, breathing world.
The massive complexity of Red Dead Redemption 2 is fascinating. The multifaceted entanglement between daily routines of common folk, consequences of your actions, time-lapsing changes found in the world (like the chopping down of the forest by Appleseed Timber company, or construction of buildings in different towns), random encounters in the world, freeform behavior during combat (like when Arthur stabs enemies with an arrow if they come too close for comfort while wielding the bow, or when enemies fire warning shots, etc.), all create believable world.
Further, you are recognized if you wear the same clothes after robbing a bank - creating a compelling reason for changing clothes instead of making it just a cosmetic thing - and can interact with people in the world with contextual controls (which can sometimes fail miserably). Those people react differently based on Arthur's looks and tone of voice. These and many more features create emerging gameplay that offers something new each time we play the game. They create an illusion of choice, an illusion that the game world is governed by the same rules as our world. This ultimately makes the game extremely enjoyable to play, to spend hours inside its beautifully fabricated world.
The Legend of Zelda: Breath of the Wild, a game that is humble when it comes to the technical side of its visuals charmed millions of people. It did that thanks to its incredible artistic design but also with its complex gameplay systems. You have excellent physics that can make every combat encounter unique even though enemies in the game aren't really smart. Unique powers available to Link create emerging gameplay at every step.
You can have fun trying to cross a river by creating ice platforms, to solve puzzles by stopping time or controlling metallic objects found in the world. The game's unique inventory-expanding system (tiny forest sprites hiding in the world) makes players look for tiny details that stand out, creating an ever-enjoyable exploration loop. Then there are those fascinating gameplay systems like the wind affecting where the fire spreads, or where you'll end up when gliding. The rain that extinguishes fires and makes climbing near impossible. All these systems combined with the massive level of gameplay improvisation offered to the player create emerging gameplay, different from the kind we saw in Red Dead Redemption but equally enjoyable.
Similar things can be said about S.T.A.L.K.E.R games, with their complex and interconnected systems that govern the bleak world of the Zone. You have multiple factions that remember you and your deeds. If you're good with one faction their members will aid you in the wild and will be friendly towards you. Then you have weather systems that affect gameplay (when it rains it's easier to stalk animals and mutants, you have to be inside during blowouts in case you don't want to die from exposure, etc.), and complex economy system that prevents players from misusing it. Its world is filled with the dynamic interaction between bands of roaming stalkers, bandits, and monsters that can fight, cooperate and show all kinds of behaviors towards the player.
Then we have the otherworldly AI storyteller in Rimworld, capable of creating bizarre scenarios that are often better than most stories found in modern video games. These stories make the game so compelling to play.
You don't experience just the game; you also experience a number of personal and interconnected stories that are lives of the colonists, which are unique every time you play. The Witcher 3 doesn't feature the emergent gameplay of Red Dead Redemption 2 or complex systems of Breath of the Wild but its world is filled with emotion and it reflects our world so good. World filled with gray morale, hard decisions that have grave consequences, compelling, emotional and intimate stories that just wait for the player to uncover them. This emerging and lifelike storytelling are what makes these two games so enjoyable to play.
These open-world games governed by complex system are the exception to the rule. For each open-world game that features complex systems allowing for emerging gameplay and storytelling we have a dozen of open-world games with a superficial design that feature zero interactivity and game worlds filled with activities that mask the blandness of their design. As beautiful as Assassins' Creed Odyssey is its world if fairly simple in design. You don't have random encounters, nor advanced physics, nor emotional storytelling with overlapping consequences.
You have zero emerging gameplay. The magic of that game is the constant feeling of discovery thanks to its huge world and pretty visuals along with a ton of busywork that functions as a driving force for many players. For instance, when we see a quest marker, we feel bad for leaving it out there; we will finish that quest not because we found enjoyment in playing it, but to avoid the bad feeling that will stay with us in case we skip it. We don't do it for the prize, we do it to avoid punishment.
I enjoyed that game and spend 100 hours in it but I know that enjoyment was way below the levels experienced in games mentioned above. I also know I won't play any other Ubisoft open-world game for some time because I need the rest from this form of open world design that puts busywork front and center while at the same time ignores complex AI systems and fails at offering the experience of emerging gameplay.
While we're talking about Ubisoft open world games it's worth mentioning that one of these emerging gameplay systems will be the star of the show in the next Watch Dogs game, Watch Dogs Legion. The ability to recruit and play with literally any NPC found in the game world is the exactly the kind of complex AI system offering emerging gameplay that I'm talking about today, and that's great. It seems people at Ubisoft are familiar with problems tied to the simplistic, busywork-focused world design found in countless modern games and are trying to come up with something new and exciting.
Because millions of people enjoy games like AC: Odyssey but will they continue to do so once these games reach their visual peak, once these amazing graphics become dime a dozen, once they are served with one after another huge open-world game featuring bland and superficial design that doesn't go beyond scratching our itch for discovery and taking advantage of our human weaknesses? I don't think so. I think that gamers will become overburdened with these superficially designed video games and dumb enemy AI found in shooters at some point during the next generation of consoles, once the novelty evaporates and they realize that visuals won't get any better.
Once that happens pretty visuals and huge worlds won't be enough. Developers will have to offer compelling AI and implement complex gameplay systems into open-world games. If they fail to accomplish that AAA video games will become stagnant and revenue will start to fall down. The best thing is that creating compelling AI and advanced open-world systems doesn't cost too much; it only takes a bit of focused work and lots of testing.
I just hope that, once that inevitable downslope revenue trend becomes apparent executives won't blame it on the gamers and "their shift in preferences" as it happened with the recent emergence of GaaS titles and will realize that an enjoyable game is more than a pretty face and the exploitation of basic human weaknesses. Creating enjoyable AI and complex worlds filled with emerging gameplay is the best way to offer progress without increasing development costs. It's certainly better than stuffing games with "surprise mechanics" and other exploitative practices to substitute ailing revenue.
NEXT GEN POSSIBILITIES
These all add-up to games with never-before-seen combinations of graphics and gameplay that result in impressive experiences. Mark Rubin, executive producer of the latest title in the bestselling Call of Duty franchise, says, “We have a Call of Duty that you’ve never seen before, with new engine, new tech, new story, new subplot. It’s going to be the best Call of Duty you’ve seen. It’s going to be the best looking Call of Duty you’ve seen ever, by far. It’s the title that if players are going to buy a new console, they are going to buy Call of Duty: Ghosts.” And this is just the beginning, as more developers create next gen games and the new hardware is really put through the paces. Sullivan says, “Next generation has a lot of horsepower to do great visuals, but it’s not just about visuals. It’s about what’s the next gen thinking behind the game. It’s going to be interesting to see which teams around the world are embracing the next generation way of designing a game. What’s that next generation experience? That’s what people are expecting.”