AI and the future of gaming for game devs
Artificial intelligence has been a major part of video game development since the industry’s inception. The first examples of AI in gaming date back to 1951 with the mathematical strategy game Nim, where players had to compete against an in-game AI. Today, AI doesn’t just power in-game opponents; it’s used to populate entire digital worlds filled with engaging non-playable characters, as showcased in titles such as Red Dead Redemption 2 and Grand Theft Auto V.
With the advancements in generative AI, many believe that these new forms of artificial intelligence will revolutionize video game development. Based on recent news, that revolution seems to have already begun. For example, Roblox announced they’re building a platform that will, “allow every user to be a creator,” using generative AI tools. Unity just launched a new AI marketplace featuring generative AI and behavioral AI solutions alongside several new AI-powered tools that help developers enhance their gameplay.
Meanwhile, Ubisoft’s R&D has designed a new in-house AI tool called Ghostwriter, which generates the first drafts of “barks” - phrases or sounds made by NPCs during a triggered event - to give its scriptwriters more time to polish the main narrative. These recent examples of AI in gaming are only the beginning of a new wave of innovation. Generative AI will help developers build more extensive and immersive worlds by automating much of the legwork, enabling them to focus on designing creative new mechanics and features.
Below, we dive deeper into how AI is being used in game development right now, and the best examples of AI in gaming.
What type of AI is used in gaming?
In the past, AI in gaming has typically come in one of two forms: deterministic or nondeterministic. Today, we also have a third player on the field: generative AI.
Deterministic AI behavior follows a specified, predictable algorithm. For example, a character might be programmed to move toward a specific location, such as how the Goombas in The Super Mario Bros. games walk along a defined route.
Nondeterministic AI behavior is the opposite of deterministic AI and follows varying degrees of uncertainty, depending on the sophistication of the deployed AI.
One of the most famous applications and a great example of AI in gaming is in Lionhead Studio’s strategy game, Black and White, which features a creature that develops based on the player’s interactions. For example, the creature can be taught that it’s not okay to eat innocent villagers by punishing it after it does so.
Generative AI is a type of machine learning where computers can generate original new content in response to prompts from the user, most commonly text (as in ChatGPT) and imagery.
The possibilities this AI unlocks for game developers are almost limitless. For example, Asobo Studio used generative AI to help build its enormous 197m square mile recreation of Earth for its Microsoft Flight Simulator.
At Inworld, we worked with the creator of The Matrix Awakens to launch Origins, a playable short where players must investigate an explosion in the fictional city of Metropolis by questioning completely unscripted NPCs powered by Inworld AI.
How is AI used in gaming now
Pathfinding algorithms were one of the first applications of AI in gaming and, as the name suggests, were used to determine the trajectory of in-game characters. Today, AI has several additional applications which game developers can use to create better, more expansive gaming experiences. Here are some of the most popular AI applications in gaming:
- Decision trees - These use a form of machine learning to create a branching storyline where players’ decisions influence the game's future. They’re common in most modern video games but are especially prominent in those with a heavy narrative focus, such as Life is Strange or The Dark Pictures Anthology.
- Enemy AI - Simple advancements in enemy AI can transform a video game. Released in 2006, FEAR is still highly regarded for its sophisticated Goal Oriented Action Planning (GOAP), which allows enemies to react dynamically during a shootout. For example, when a soldier realizes it's in danger, it will retreat but if it can’t identify a safe path to do so, it will hunker down and blind fire on the player.
- Procedural generation - Some games use AI for procedural generation to create open-world environments, levels, and other assets. For example, No Man’s Sky uses this technology to generate planets and build an entire galaxy for players to explore.
How AI will be used in gaming in the future
With generative AI becoming increasingly sophisticated, we expect game developers will find all-new ways to utilize AI when designing video games. Here are some predictions about how AI for game development might evolve in the future and how game developers will use AI tools to circumvent some of the current disadvantages of AI in gaming.
Games such as No Man’s Sky already use AI for procedural generation, creating environments based on the rules a level designer inputs to guide the AI. Generative AI can be utilized similarly but can deliver far more impressive results, including worlds built in a specific style or time period.
For example, one generative AI tool, Promethean AI, was used to recreate the world in Stranger Things in just 15 minutes As one of the biggest advantages of AI in gaming is its ability to streamline tasks that typically take a lot of time and resources, generative AI can help indie game developers build more diverse and expansive game environments providing more gameplay value to gamers and helping them compete against bigger titles.
It might also help AAA developers create a first draft of game environments allowing all games to create larger worlds giving players more gameplay options.
NPCs (Non-playable characters)
One of the most exciting aspects of generative AI is its ability to help create more believable and interactive characters. At Inworld, we built a character engine to craft characters that have distinct personalities and demonstrate genuine contextual awareness, meaning they can dynamically react to the world around them while staying in the game world. Better yet, they are seamlessly integrated into real-time applications with built-in performance optimizations.
As we demonstrated with our playable demo Origins, having NPCs that can respond and react to in-game actions can be transformative. Imagine this technology's impact on some of the biggest triple-AAA open-world games, such as a future Red Dead Redemption? A study we conducted in collaboration with Bryter market research found that 99% of gamers think AI NPCs like Inworld’s will enhance a core aspect of gameplay.
Image generation is one of the most common uses for generative AI. Tools such as Midjourney, Stable Diffusions and Dall-E 2 can be used to create high-quality 2D image from text, and these techniques have already made their way into some of the biggest video game studios. However, the images often still need editing to fix common AI mistakes like too many fingers and unnatural body positions.
As tools continue to develop, many suspect that generative AI won’t be used for just 2D assets and concept art, but also to help create and animate 3D models. Or to create a first draft of a model that a designer can refine, saving design time.
However, given Valve’s reported stance on AI-generated art and issues with copyrighting generated art, it’s likely that this practice will face some challenges to becoming a key part of video game development.
Some of the most successful video games of all time are not just recognized for their outstanding gameplay, but also for their iconic soundtracks, something that’s particularly notable in series such as The Legend of Zelda and Final Fantasy.
Creating a soundtrack that captures the game's look and feel without becoming repetitive (and let’s face it, annoying) during its lengthy runtime is challenging, especially for smaller and indie developers. The dynamic nature of music also needs programming through audio tools such as Wwise, which can take up a lot of time and resources.
In time, the advancements being made to AI will benefit in-game mechanics. For example, most games today have several difficulty options, usually easy, normal, or hard. Using AI, game developers could build an adaptive difficulty that alters the game's rules based on the player's performance, creating a more personalized and satisfying experience.
While experiences such as these do already exist in games such as the original Resident Evil 4, they’re few and far between due to the difficulties of programming them. Advancements in AI are already streamlining this process significantly.
AI NPCs run by character engines like Inworld’s can also transform game mechanics. For example, Inworld has recently introduced an updated goals and actions feature, a player profile feature, and a relationship feature.
Goals and actions give NPCs defined goals that they then work to execute dynamically therefore moving the narrative forward or changing game play. Meanwhile, player profiles allow NPCs to respond contextually to a player. For example, if a character is from one faction, an NPC from another faction will respond to them rudely but if they choose to play as another faction when they play next, the NPC would respond to them in a more friendly way.
Finally, our relationship feature allows players to progress in their relationships with each NPC based on every interaction with them. If they’re rude to an NPC, they could end up making themselves an enemy. If they’re nice to them? They could end up becoming their best friend and getting more support from that NPC. Dynamic features like these could have a significant impact on game mechanics.
Many people believe AI-assisted coding will make a huge impact on gaming. However, while it is possible to generate code using AI, this still requires extensive testing and verification from the developers to ensure everything works as intended, so there are only limited gains to productivity compared to other areas.
Until further advancements are made, adoption will likely be limited, which means coding is one of the current disadvantages of AI in gaming due to the additional time it needs to check code.
Examples of AI in gaming
We’ve touched on many of the most impressive applications of AI in gaming, but there are still many more fascinating examples that are worth checking out for yourself. These titles use a variety of AI applications to power gameplay mechanics and are must-plays for anyone looking to learn more about how AI is used in gaming.
- NetEase’s Cygnus Enterprises uses Inworld’s character engine to power the game’s sidekick PEA. Not only can you chat voice-to-voice with PEA, but she’s also contextually aware and can warn you of upcoming dangers in the game world.
- Alien: Isolation uses an advanced AI to power its digital take on Ridley Scott’s iconic sci-fi predator, the Xenomorph. The creature behaves differently every time the player encounters it, and it learns from past mistakes, ensuring each meeting is as terrifying as the last.
- Middle Earth: Shadow of Mordor (and its sequel, Shadow of War) is widely acclaimed thanks to its Nemesis System, which allows the Uruks players face-off against to adapt and grow alongside them. The Uruks can even remember past encounters with the player and taunt them.
- Left 4 Dead constantly feels new thanks to an in-game AI which manages each level, referred to as ‘The Director.’ This decides when to send a horde of zombies at the players, where to spawn challenging encounters, and when to give the players time to recover.
Advantages of AI in gaming
When implemented correctly, AI can have a transformative effect on game development. Below are some of the advantages of AI in gaming:
- AI can be used to create highly-intelligent NPCs helping to create the illusion of a genuinely real in-game world, boosting immersion.
- Certain games, such as Microsoft Flight Simulator and the upcoming Starfield, are only possible because they use generative AI to achieve their enormous environmental scale.
- Generative AI can release NPCs from their scripted shackles so they can have almost human-like interactions with players.
- Generative AI can speed up game development and reduce production costs by automating basic tasks.
Disadvantages of AI in gaming
Countless AI applications in gaming can benefit players, but there are some drawbacks. These are the main disadvantages of AI in gaming that developers need to be aware of:
- Using generative AI reduces production time, but it also means that developers have less control over certain aspects of their game. For example, the ‘palaces’ players explore in Persona 5 were the first in the series not built using procedural generation. As they were handcrafted, critics felt they had more personality than previous installments.
- Generative AI is still a very new technology, meaning there are only a select number of game developers with the expertise to use it. If you plan to use AI for game development, you may need to invest in additional training or expand your team.
- Generative AI can build entire virtual worlds, but that doesn’t mean they will be perfect. Developers should invest extra time into testing to ensure everything works as they intended, or they risk increasing the number of potential bugs.
The promise of generative AI for game development
Generative AI has seemingly endless potential when it comes to game development. It can help build more expansive worlds filled with seemingly living characters while drastically reducing production costs and speed up development by automating menial tasks, enabling teams to focus on building more imaginative game mechanics and other features.
There’s no need to imagine what generative AI in modern video games might look like. One modder went to the trouble of implementing our character AI directly into The Elder Scrolls: Skyrim, allowing them to converse with their companions and other characters about almost anything.
AI for game developers
Gaming is the most expansive entertainment industry worldwide and is estimated to be worth more than £321 billion by 2026. As the industry grows increasingly competitive, video games must find new ways of delivering innovative experiences to capture players' attention. By adopting generative AI for game development, studios will be able to build immersive worlds and exciting new mechanics to help them stand out on the market.
Our research found that almost everyone who plays video games feels that AI will enhance gameplay, with 81% even going so far as to state they would be willing to pay more for a game that uses these features to provide a more immersive experience through NPCs.
If you’d like to learn more about how generative AI can be used to create in-game characters with unique personalities, head to our studio.
If you’d like to see examples of studios implementing our character engine and other AI in gaming, you can check out our other blogs: