The creation of more advanced artificial intelligence, or AI, leads to new and exciting possibilities in gaming. But how is AI used in video games today, and why are these advancements such a big deal?
What is AI in Gaming?
Artificial intelligence is programming that allows certain characters in a video game, such as non-playable characters (NPC’S), and enemies, to act in a way that feels as if they were controlled by a human, or were acting with a mind of their own.
AI is used to make games feel more immersive, but what does this mean in practice?
If you’ve ever played the classic game Pacman, then you’ve experienced one of the most famous examples of early AI. As Pacman tries to collect all the dots on the screen, he is ruthlessly pursued by four different colored ghosts.
But they don’t just follow him; when you’re playing they seem to try and ambush the player.
How did they accomplish this?
The ghosts are programmed to know Pacman’s location, but each one is set to act a little differently in response to it.
One of the ghosts is set to wander, one is set to aggressively follow Pacman, one is set to go in the direction where Pacman is going, one is set to move randomly unless another ghost is close to catching Pacman, and the final ghost is set to follow Pacman when he’s far away, but go to a specific location on the map when he’s close.
These four behaviors make these ghosts, even in a game from 1980, appear to have a will of their own. They feel alive. And that is the purpose of AI in video games.
Why Does AI Matter?
Nearly all games use AI to some extent or another. Without it, it would be hard for a game to provide an immersive experience to the player.
The goal of AI is to immerse the player as much as possible, by giving the characters in the game a lifelike quality, even if the game itself is set in a fantasy world.
As AI gets better and more advanced, the options for how it interacts with a player’s experience also change.
Take a game like “Middle Earth: Shadow of War”, and their nemesis system. In this open-world adventure game by Monolith Productions, the enemy orcs that you encounter throughout the game have a wide variety of personalities, traits, and applications. Any orc that you can see in the game can join you, betray you, a spy for you, and more.
But that’s not all; the nemesis system allows the most memorable moments of gameplay to be even more memorable, as by defeating a difficult orc captain, you might be creating an enemy that will come for you in the future, and remember what you did to them in the past.
The system strives to create an entirely new way for players to interact with the NPC’s in the game. When an orc captain kills your favorite orcish ally, it feels personal.
When that difficult enemy that took you ages to defeat returns in the worst possible moment, the game feels much more intense. This experience is catered to the players’ actions and the procedurally generated characters, and so will be somewhat different for every player.
But as advanced as all of that is, it is still made of pre-programmed instructions by the developers.
As the AI uses new technology, a similar game might not just have orcs that seem to plot or befriend the player, but genuinely scheme, and actually feel emotions towards the play. This would make it a game that truly changes based on every action the player takes.
How is Artificial Intelligence in Gaming Evolving?
Up until now, AI in video games has been largely confined to two areas, pathfinding, and finite state machines. Pathfinding is the programming that tells an AI-controlled NPC where it can and cannot go.
From retro-styled 8-bit games to massive open-world RPGs, this is still important. Developers don’t want the villagers in a town they’re working on to walk through walls or get stuck in the ground. They also don’t want them to stand perfectly still.
Finite state machines, on the other hand, allow the AI to change its behavior based on certain conditions. A good example of this in action is the enemy soldiers in the Metal Gear Solid series.
When you enter their field of vision and are spotted, it triggers an alarm, and all of the enemies in the area begin to converge on your location and attack.
You have to leave their field of vision and hide for long enough for the guards to give up the chase, and return to their initial positions.
These are essentially two different states, the “Chase” state, and the “patrol” state.
These may sound like rudimentary ideas, but these two concepts govern most of AI programming today. In the SIMS series, when AI is hungry, they get food. When they are tired, they head to bed.
This mimics real decision making, but it’s actually the state of a SIM changing from “neutral” to “Go to the nearest source of food”, and the pathfinding programming telling them where that nearest source is.
So how is AI evolving? Already it’s changed greatly with the sheer amount of pathfinding and states that developers can give to NPC’S. There’s a world of difference between Pacman and Skyrim after all.
But right now, the same AI technology that’s being used to create self-driving cars and recognize faces is set to change the world of AI in gaming forever. Soon enough, pathfinding might not just be telling an AI where it can go.
AI might create the entire, realistic landscapes from scratch, calculating the walls it can and can’t walk through instantaneously.
What kind of storytelling would be possible in video games if we could give NPC’s actual emotions, with personalities, memories, dreams, ambitions, and an intelligence that’s indistinguishable from humans.
Data scientists have wanted to create real emotions in AI for years, and with recent results from experimental AI at Expressive Intelligence Studio, they are getting closer. It won’t be long after they succeed that we could see these AI in games.
In other words, pathfinding and states, the fundamental principle upon which all AI has been built, may soon be shaken.
Pros and Cons of AI in Video Games?
So what are some of the advantages and disadvantages of AI’s evolving status, and the new technologies that are coming out? Here are just a few of the pros and cons worth thinking about as we enter a new era in gaming.
Artificial Intelligence: Pros
As AI evolves, we can expect faster development cycles as the AI is able to shoulder more and more of the burden. Procedurally generated worlds and characters will become more and more advanced.
You won’t see random NPC’s walking around with only one or two states anymore, they’ll have an entire range of actions they can take to make the games more immersive.
Also, excitingly, if NPC’s have realistic emotions, then it fundamentally changes the way that players may interact with them.
You can learn to truly care about the citizens of a town you’re protecting, or hate the villainous enemy that always stays one step ahead of you until you finally defeat them. There are plenty of opportunities presented with ever-evolving AI, but there are also some problems.
Artificial Intelligence: Cons
As AI gets more advanced, game makers have less control over them. This means we might miss out on some of the carefully crafted worlds and levels we’ve come to expect, in favor of something that might be easier but more…robotic.
It may be a similar situation to how players can often tell when a game was made using stock assets from Unity. It affects the overall experience.
It also means we may get some clunkier games when the AI programming goes awry. For example, in Aliens: Colonial Marines, the AI enemies acted in strange and peculiar ways, destroying the immersion.
If AI is essentially programming itself, will we eliminate that problem altogether or make it easier to happen? It’s hard to say.
Finally, there’s a chance that as AI is able to handle more of the game programming on its own, it may affect the jobs of many game creators working in the industry right now.
These are all things worth considering as we move forward into uncharted waters.
Using AI in Game Development
In a few short years, we might begin to see AI take a larger and larger role not just in a game itself, during the development of games. Experiments with deep learning technology have recently allowed AI to memorize a series of images or text, and use what it’s learned to mimic the experience.
There are AI being used right now to generate works of art similar to Picasso, or emails that seem to be written by a human. This same technology is currently being adapted to game development, wherein an experiment, an AI was able to program a playable level of a game using nothing but photographs.
As this technology becomes more reliable, large open-world games could be easily generated by AI, and then edited by the developers and designers, speeding up the development process.
But that’s not all, there is also the advent of facial recognition software and deep fake technology that looks like it may play a big role in future development cycles. Deep fake technology lets an AI recognize and use different faces that it has scanned.
While it’s in its infancy, impressively realistic 3D models have already been made using the faces that this kind of AI can scan. Now imagine if this same technology was used to generate a building or a landscape. This would save countless hours of time for developers.
Thinking even bigger, it’s entirely possible that soon enough, an AI might be able to use a combination of these technologies to build an entire game from the ground up, without any developers needed whatsoever.
They may even be able to create these games from scratch using the players’ habits and likes as a guideline, creating unique personal experiences for the player.
It’s hard to say if it will get this far anytime soon, but it would be no surprise if we start to see AI taking a bigger role in game development in the next few years.
The Future of Artificial Intelligence in Gaming
While AI technology is constantly being experimented on and improved, this is largely being done by robotics and software engineers, more so than by game developers. The reason for this is that using AI in such unprecedented ways for games is a risk.
Right now, developers have control over the world they’re creating. Leaving their games in the hands of hyper-advanced intelligent AI might result in unexpected glitches, bugs, or behaviors.
If the possibilities for how an AI character can react to a player are infinite depending on how the player interacts with the world, then that means the developers can’t playtest every conceivable action such an AI might do.
While the technology and the prospects it offers are exciting, big studios are a little slow to jump right into it. So how long before they do?
It’s inevitable, but hard to say.
At some point, the technology may be well enough understood that a studio is willing to take that risk. But more likely, we will see ambitious indie developers make the first push in the next couple of years that gets the ball rolling.
With how fast technology is progressing, it’s very possible that we will have everything we always dreamed AI could by the end of the decade.
Will AI Make Games Too Hard?
There is an old saying that still holds up; be careful what you wish for. It is entirely possible that as we begin to implement more advanced AI into our games, we may run into some problems.
If, for example, the enemy AI knows how the player operates to such an extent that it can always win against them, it sucks the fun out of a game. Already there are chess-playing programs that humans have proved unable to beat.
If a similarly difficult AI-controlled every aspect of a videogame from the ground up, the results could be very unfair and broken. If NPC’s in a game develop real, human-like personalities and intelligence, then maybe playing a game begins to feel a bit too overwhelming, as players are forced to juggle social responsibilities in both the real and virtual world.
There are also some issues with our relentless pursuit of realism in video games. Is making a game where an AI reacts to every decision a player makes really more fun, or annoying?
Imagine a Grand Theft Auto game where every NPC reacts to your chaotic actions in a realistic way, rather than the satirical or crass way that they react now.
Would we really enjoy a game where the power fantasy side of video games is ripped from us, and committing one crime is enough to spend years of real-time in a prison?
Do we want game characters to bully us if we wear silly costumes?
Plus, there’s a big question as to how expensive the technology required for these advanced AI systems will be.
If they become the new industry standard, do AAA games drift further and further away from being possible for Indie developers? Or will advanced AI be available to everyone, and actually equalize the playing field a bit?
The truth of the matter is, these are hurdles that will have to be faced, but that doesn’t mean they will damage gaming forever.
Even failed experiments give you information. If we build AI that can develop entire games, have realistic emotions, or change the entire game world based on any decision a player makes, and it ends up not being fun–so what?
We’ll either stop using them or learn from our mistakes and make the next superpowered AI games more entertaining.
Remember, VR started with the Nintendo Virtual Boy, a notorious failure. It took a while, but the concept of VR survived and has improved. So expect a few hiccups as these advanced AI are implemented, but you can also be sure that we’ll get past them in time.