November 13, 2019
If you have ever played a video game, you’ve surely come across a specific situation that made you fly off the handle. Maybe it was an NPC in, say, the GTA series that got inexplicably stuck in a tree preventing you from going through with a mission. Perhaps it was an infuriatingly hard final boss, like Absolute Virtue in Final Fantasy XI. It could be that you didn’t stand how frustratingly boring an open-world could become, like in No Man’s Sky.
When all that happens, most players like to blame video game developers and their faulty use of artificial intelligence (AI). How come that big companies, seasoned coders, and QA services can’t see what’s happening with the supposed intelligence they are using? Why do so many games still have these many issues? Isn’t the AI used in them, you know, “intelligent” enough to keep those things from happening?
Well, it may come as a surprise, but what you think is AI in action in a lot of the video games you play isn’t artificial intelligence at all. In fact, the logic of most modern games is programmed by developers through complex algorithms. What you see in Dead Cells and its procedural level design uses basically the same principles as some oldies, like Dwarf Fortress, Diablo, and Rogue.
So, if AI isn’t really powering the video games we all play today, what could happen when it starts being used? That’s the question a lot of people in the video game industry are asking themselves. And the answers are starting to appear. Here’s what we can expect from AI and its use in video games.
Why we are NOT using AI extensively today
It might come as a surprise that games that came decades ago use the same logic to create their worlds than its more modern counterparts. However, for game developers, this makes a lot of sense. To understand that, you have to take a look at some basic concepts about video game design.
The first and most important one is the goal of creating video games in the first place. Basically, we can safely assume that all video game developers are out to create an enjoyable and fun experience for players. The objective surely is to bring a challenging game that can keep players hooked and interested but not so hard that it becomes unbeatable. We’ll come back to this in short but first, it’s also important to understand the logic used in video games.
What we all think of as AI in video games is nothing more than a combination of logical components at a base level. When a boss anticipates your movements to prevent you from striking it, when a tactical team ambushes you in a hallway, when a defender tackles off the ball from your forward—all of that is in some sort programmed into the video games themselves. How could that be?
Though it’s a lot more complicated than that, there are two main components that make this happen. On one hand, there’s an element called pathfinding, that allows things in video games to move from point A to point B as required. On the other hand, there are finite state machines, which can be seen as the different states a non-playable character exists at any given time, sort of the possible options for actions in which that NPC moves around.
The combination of these components and their variations is what makes the most of the intelligence you see in games today. The only thing that’s changed since the Pac-Man and Pong days? The fact that developers are using these concepts at bigger scales and pushing their limits thanks to the increased computation power we all enjoy today.
This means that the seemingly intelligent decisions taken by your foes, allies, and NPCs in a lot of video games are more determined by programming than taken on the spot. So, why not enrich those video games with true AI algorithms, capable of making their own decisions and creating a more lifelike experience for players? Wouldn’t this prevent the frustrating issues we’ve mentioned above? Certainly not.
Machine and self-learning techniques used in modern AI would break the experience video games want to offer you. Remember the games’ main objective? Developers couldn’t create an enjoyable and fun experience if they let AI decide to do whatever it wants. That’s because the AI itself (through the characters or the scenarios) could disrupt the story in unpredictable ways.
Think of it like this. What would happen to a director shooting a movie where all the actors are doing and saying whatever they want? First, the director would go insane. But, most importantly, the story could never be told, as all actors involved would do and say what their intelligence tells them to do. That’s precisely what would happen in a video game.
If a video game, regardless of its genre, would run on AI algorithms, it could lead to an unplayable beast. That’s because the AI would adapt itself to the players to learn how to better defeat them, to the point where it would become impossible for the players to win. Additionally, the game would change scenarios, NPC behavior, and even the story to fit whatever the AI sees fit, throwing the whole experience into the territory of the unpredictable and even incomprehensible.
In short, the AI could present certain interesting variations on a video game basic mechanics (like new smart ways to team up against you in an FPS or dynamic strategics in an RTS) but the unpredictability would make it highly frustrating for players. That’s because the players could possibly develop ways to beat a machine that’s always learning how to get better and better.
That’s why the use of AI in modern gaming is reserved for other aspects. What we call artificial intelligence in today’s video games is more of an illusion. In other words, the logic programmed underneath is complex enough to feel intelligent but it’s limited to fit a narrative that won’t disrupt the gaming experience.
So, if we’re not truly seeing AI in games today, will we see it someday? Of course, we will! Perhaps it’s not how you are expecting it, but there are a couple of uses in test today that can illustrate how AI will redefine future gaming.
AI in video game development
If artificial intelligence isn’t a good match for a video game’s narrative or characters (yet), then there’s definitive promise in two different fields. On one hand, AI can be used to take the procedural design generation to new levels. On the other hand, AI can be of great aid to develop games, from the fundamentals all the way to QA development tasks.
When it comes to the time to mention some of the best open-world games out there, it’s highly likely that titles like Red Dead Redemption 2 or The Legend of Zelda: Breath of the Wild come to mind. It’s hard to argue that both are some of the finest examples of world-building in the current gaming scene. Yet, through AI’s aid, such worlds could be taken even further.
By limiting its application to world-building, AI could power environments that are as closest as the real world as we’re ever going to get. Breath of the Wild’s underlying systems are very complex and very close to real life (in the sense that there are rules of physics and thermodynamics at play). But imagine that those rules change over time, bringing modifications to the game itself. Seasons could change. Constructions could age, decay, and come crashing down. The things you do could bring permanent consequences to the world around you. The possibilities are endless.
In this sense, AI could mean the dawn of the adaptable and ever-evolving game that would take a procedural generation to heights we’ve never seen before. In fact, the ultimate goal would be for machine learning algorithms to be always on the move, bringing change after change every time you play the game, making it feel fresh with every new access.
As for AI used in gaming development, this could be a serious game-changer (pun intended). Through the use of deep neural networks and other AI-powered software, computational systems could create complete games from the ground up with minimal or no human intervention. Case in point: a system called Angelina that has already created some games you can check out at itch.io.
Of course, these kinds of systems are far from being marketable in their current state. AI systems to create games are in an experimental phase that’s restricted to researchers or indie developers. However, such efforts are leading the way to a new way of understanding gaming and could have a significant impact on the whole industry.
One particular department where it could have the most effect is in testing. You can expect to see AI systems help developers by providing them with QA development services to find bugs and improve the overall gameplay. Additionally, these systems would be embedded in the games themselves, allowing developers to gather important data and insights that could lead to further improvements over time.
Some final words
The days where game studios, freelance designers, and QA development outsourcing teams are replaced by AI-powered software are far from us. Though the technology is taking huge strides in the gaming development world, chances are that we’ll see them collaborate with human teams in the short and midterm rather than replace them—which is a fantastic thing!
AI can bring new insights, suggestions, and directions to game creation. It can aid in bringing more complex and sophisticated worlds, and help in releasing complete games that leave the current, painfully bloated and buggy game releases behind for good.
All of this could eventually lead to a new way of implementing AI into the game itself. But if we leave the video game’s narrative and inner workings in the hands of artificial intelligence, what will we find once we power up our consoles? Maybe, the introduction of AI at such a deep level will require us to redefine what we understand video games as entertainment experiences. However, seeing that developers themselves aren’t taking this route, you’ll be fine if you leave those wonderings to sci-fi stories