As we’ve seen the last few years in society, generative artificial intelligence (AI) is capable of quite a lot, even if it clearly has its limits. Then it hit me the other day, it would be pretty cool to play a video game that was made completely by AI, with the only human involved being the one who types in the prompts and hits “Enter”. I admit, I might be relapsing a bit but the idea was too tantalizing to pass up. So, today I’ll go over what I envision this game looking like, and how to realistically pull it off. The goal of this is to create a game that’s at least passable (let’s be honest, I’m not expecting Hideo Kojima here). I will assume no DLC’s or patches, and no human intervention from a quality control or editorial standpoint.
While I’m not here to give any ideas (shout out to The Man Himself) to executives of gaming studios, I wonder how those who work in the video game industry will view this thought experiment. I’m banking on a set number of sharpened pitchforks; that’s a given. Doubly so since the Hollywood writers just went through a major strike regarding the use of AI in screenwriting, though employees of game developers have no such labor union to protect them. I also expect a small though vocal percentage of the industry to be arrogant in how they’d obviously make something better than AI could, though I’m not convinced of that anymore. Mostly though, I expect a quiet air of anxiety as the nerds look on to what will take away their 100-hour-per-week jobs. Let’s begin…
First, we’ll have to set some parameters. Ideally, the goal of this hypothetical game is to be at least somewhat playable, and to have reasonable entertainment value measured by at least a mid-level score by a respected gaming news outlet. Thus, the goal is not to break the AI, but to give it tasks that it should be able to accomplish (should being the operative word here). That’s right, I’m not trying to be Trunks right now. However, balance is necessary and while I’m not suggesting that the AI’s be pushed to their breaking points, I’m not interested in making this too easy either; generating an Undertale clone (in terms of graphics and gameplay) constitutes a failing grade.
First and foremost is the story of the game. This one should actually be the easiest part of the game to produce. Essentially, all that’s truly needed is an AI that can create a passable fictional story. In fact, there has been a wave of students in English Lit and Creative Writing classes across the nation that have been using ChatGPT for this very purpose. Personally, I’d prefer if the game to be made-and hence, the story to be written-was an RPG of some sorts. This is because RPG’s tend to be the most story-intensive games, thus we’d get to see what the AI writing is truly capable of. I’d also be interested in seeing the world-building aspect of the story; thus, I’d have a bias towards seeing a fantasy setting. This is because allowing the AI to base a game off of real-world nations would be too easy. I’d also be interested to see how the AI would create dialogue.
Of course, no reputable RPG is complete without side quests and the occasional optional super-boss. The same AI that writes the story will also be tasked with creating a plethora of side-quests, with the prime directive that the side quests must tie back into the main plot of the game. The other major stipulation is that the side quests need to be different from each other; nobody wants a repeat of Final Fantasy XV. Additionally, the rewards for these side-quests need to have a prize that is enticing enough for players to partake in; nobody wants a repeat of Final Fantasy VII…
The graphics and the artwork should be a cinch (though I say this knowing that I very easily could be proven wrong). Much of the art work and graphics used in gaming today is created by similar technologies, thus an AI should be able to create something similar to it, whether something photorealistic or something with a more cartoony vibe. Personally, I’d like to see something with a cartoony vibe; photorealistic would be too easy as it requires less creativity. Thus, something in the art style of Genshin Impact or Zelda: Breath of the Wild would be a tough-but-fair challenge for the AI to create.
The control scheme would be created by another AI. This could vary wildly based on both the training data and what platform the game is being made for. For the best chance at a smoother experience, I’d honestly rather limit it to the PS5 or Xbox One. Color me biased, but the fact of the matter is that the controllers of the consoles have far less inputs than a keyboard. This is crucial because it’ll limit the possible mistakes made by the AI and uncomfortable combinations (i.e. not needing to hold down the A-key and F12 at the same time).
The game mechanics would also be created entirely by AI. However, my only question would be if this would be the same AI that lays out the controls scheme? Reasonable minds can disagree on this, and I can be swayed either way on this topic. It would be interesting to see how an AI would create menus that a human player would (or wouldn’t?) be able to intuitively navigate. It would also be interesting to see if the AI could create mechanics that tie into the plot of the game; such as Death Stranding’s blood-as-ammo mechanic.
Next, the story/dialogue, mechanics, controls, and artwork creations need to be combined with each other. This, of course, would be done by a separate AI. This one wouldn’t create anything, it would be more of a conductor AI, tasked with intricately weaving the creations of the AI’s together seamlessly and without a hitch. Of course, if that last sentence caused you caustic fucks to snicker sarcastically, then I’ve done my job!
Now that the sauce, dough, and cheese of this digital pizza have been layered and baked, it’s time to taste-test it. However, I don’t care if you’re salivating because it’s once again time for the AI’s to shine. I’d like to see how well an AI would do with quality de-bugging a game. Truthfully, this could go either way. Humans miss things in plain sight all the time, hence a rules-based AI should (there’s that s-word again!) excel at this task. However, the AI could also go awry and focus on some of the most inconsequential minutiae ever, while leaving most of the game a total mess. Truthfully, I’d expect the final product to be filled with bugs, but isn’t that half the fun of this thought experiment?
Finally, we would arrive at the marketing of the game. The marketing, again, would be handled by AI. AI should be able to handle this task well, Exhibit A. The ad wouldn’t even need a lot of dialogue either, as many game advertisements would really just need some game footage and kickass music. Some clever prompts from the human wielding the AI could be used to great effect here. Obviously, the algorithms of all the social media platforms and video/streaming platforms would come into play as well, making sure that this game would keep coming up in potential customer’s feeds.
I look forward to playing the end-product, even if it’ll have more bugs than a beekeeper…
Home Page 