Good Uses for AI in Video Games (And Bad Uses)
top of page
  • Writer's pictureCJ

Good Uses for AI in Video Games (And Bad Uses)

The latest ‘future of’ or ‘enemy of’ video games has floated to the surface, like cream or scum. Artificial intelligence, AI. I think most gamers, that being people who actually play video games, think it’s probably a terrible idea. But this doesn’t stop the ‘brilliant’ minds in the board rooms of tech companies throwing millions of dollars at any start-up promising ‘The next great era of gaming’.

It has recently been revealed that Steam is blocking games which use AI-generated assets or dialogue. Which is a huge blow against tech bros leading the AI charge.


While I generally agree that AI shouldn’t replace game developers, and for the most part just can’t, I do believe there are a few uses for AI in video games that would work quite well.


(Please be warned, the following may contain instances of AI)


Creating Bespoke Dialogue for NPCs


With my Twitter timeline littered with ‘big tiddy anime girlfriend’ adverts that use AI to simulate conversation for the bored and lonely, it does seem that AI is quite capable of mimicking human interaction quite well.


Picture the scene, you’re walking through the lands of Tamriel and every NPC you talk to has been programmed with some basic knowledge of who they are, and all their dialogue is generated by ChatGPT to fit their character. This still requires programming by a person, someone to design their character and who they are. But once this personality is set, they’re given the freedom to talk and respond how they see fit.


With Microsoft owning Bethesda and the Elder Scrolls series and owning ChatGPT, this isn’t a particularly far-fetched idea.

Maybe in Elder Scrolls 6, you can finally ask for some more background information regarding how exactly he took an arrow to the knee. A different story for every playthrough.


Random Generated Side Quests


In a similar vein, randomly generated side quests could be done simply enough with AI. I specify side quests as frankly, I have my doubts of a current AI being able to generate something original and engaging for a main story. But I’m sure it can handle ‘Miss Miggins has had her apples stolen by bandits and needs you to hunt them down and bring them back for some XP and gold.’

Frankly, following a formula to generate semi-interesting versions of the usual fetch quests doesn’t sound super interesting, but it does sound exactly like half the side quests in open-world RPGs that currently get created by actual human writers. It’s making me question if this isn’t already in practice for coming up with ideas on RPGs.


While I don’t personally care much for grindy missions like this in games, they are present, and they aren’t going away. We could spice up playthroughs by making every experience different. Just for some less repetitive post-story content at least.



Improving Enemies for Souls-likes


Now this one might be the trick to make Souls-like diehards admit they’d like an easy mode.


AI is basically machine learning, it follows an outcome of events and learns it so that it can replicate the same outcome or differ from it next time. Picture this being used on your enemies in Elden Ring. You learn their patterns to defeat them but wait, they’re also learning yours. They come in with three swings which you have learnt to roll away from, and then you attack. The next time, having studied your pattern, they go for four swings.


Honestly, it sounds exhaustingly difficult. Which is most definitely not for me. But to the hardcore sadists who love nothing more than to throw themselves against a brick wall until it breaks, this could be the next platform of excellence.


It might be that this appeals to no one. After all, learning those patterns of the enemy so that you can take them on without taking a hit is the goal here for most fans. It’s a skill that’s rooted in muscle memory and pattern recognition. Which an AI-learning baddie would strip from you.


Well, git gud…



Procedural Generation


This next one I have put in as a fitting experiment. I asked ChatGPT the following question.

‘I'm writing an article on good uses of AI in video games, can you please give me an example of one with a description about it.’


The first response happened to be the idea I suggested above, more or less. Except better written but minus the charm (I like to think). Which really does speak to replaceability when it comes to AI.

I asked for a second example.

While you read this, I’m going to be looking up off-grid cabins I can move to in order to avoid the upcoming AI uprising…


From ChatGPT:

“Certainly! Another great example of a good use of AI in video games is the application of procedural content generation (PCG). PCG refers to the use of algorithms to generate game content dynamically, such as levels, environments, quests, and items. This technique not only saves time and resources for game developers but also provides players with unique and diverse experiences.


One standout game that leverages PCG is "No Man's Sky." This open-world space exploration game features a vast universe with billions of procedurally generated planets, each offering a unique combination of terrain, flora, fauna, and resources. The game utilizes AI algorithms to create these planets and their ecosystems on the fly, ensuring an almost limitless variety of worlds to discover.

The AI in "No Man's Sky" is responsible for generating the landscapes, weather patterns, animal behaviors, and even the architecture of alien civilizations. By employing complex algorithms and data-driven processes, the game creates an expansive and immersive universe that feels remarkably realistic and organic.


The beauty of PCG in "No Man's Sky" lies in its ability to provide players with a sense of wonder and discovery. Every planet they encounter is an uncharted territory, with unique topography, diverse ecosystems, and unknown secrets. This keeps the gameplay experience fresh and engaging, as players are constantly motivated to explore and unravel the mysteries of the universe.


Additionally, PCG allows "No Man's Sky" to offer a vast, living world without requiring an impractical amount of manual content creation. The AI algorithms generate content on the fly, resulting in a game that feels expansive and rich in variety, even though it is practically impossible for developers to handcraft such a massive universe.


Through the use of AI-driven procedural content generation, "No Man's Sky" demonstrates how technology can empower game developers to create massive, dynamic, and infinitely diverse game worlds. This approach provides players with an ever-changing and awe-inspiring experience, where each journey through the cosmos feels uniquely their own.”


Experiment over. I feel numb.


While I’ve listed some ideas on where AI can be used to improve video games and where it works there most certainly are areas which are dead ends yet seems to spout up all the time. And much like the blockchain / NFT fiasco we recently endured and appear to have bullied out of the mainstream, here is a 'bad' use for AI in video games.


Creating Games from Simple Input


‘Anything you can think of the AI can build. Just type what you are thinking and away we go!’


This is the promise that gets greedy tech CEOs wet. Unfortunately, for them at least, this just isn’t something that can be done. I’ve seen some attempts going around with my own eyes, and it’s safe to say the jobs of game developers are very much secure.


‘Dream up any game and make it happen’. Sure, as long as that game is a fortnight clone with minimal customisation of location and skins for the baddies.


Is that what you wanted, was that the big dream game you have hoped to make? A generic shooter with piss poor artistic direction and less than zero originality.


Devolver Digital’s annual industry mocking games show went right for the throat on this very aspect this year. With their Volv-E mascot robot churning out new games by just giving it a description. Which, as a premise is wonderful but very much unattainable with our current level of machine learning. And this isn’t likely to change any time soon. (I’d give it 200 years).



While it’s a nice idea for a layman of coding like me to be able just to reel off my amazing video game idea I had on a longer visit to the bathroom, it’s fantasy.


Having seen these failed attempts to reach the ambitions their marketing department promised, I’m left feeling that the whole thing is a bit of a scam. Create a company promising the undeliverable, get bought out by bigger tech companies, and walk away with the money.


Recently I’ve seen tech companies’ stocks go up when they announced investing in various AI projects. It’s impressive, it really is, that we’re so very eager to bring about Skynet. But luckily for video game devs, the promises of these AI games are truly hollow.


One day the world of Volv-E may be real, with enough pre-programmed assets and correct asset tagging something very generic may come along and threaten game production as we know it.

For now, I think we’re safe.


If you can think of any more interesting uses for AI let us know. If you’re a tech bro deeply upset that I insulted your AI companion, maybe don’t leave mean comments. Remember, a human wrote (most) this article.






Post: Blog2 Post
bottom of page