In theory it sounds amazing, but in practice I see a dark future ahead where it quickly becomes impossible to find good games. There's already too much garbage flooding every marketplace if you're looking for one to play. Even the Switch store is full of absolute bottom of the barrel drek like Hentai Puzzle Elves and whatever.
If you're on the other side? Today it's purely luck if your game out of the 50 released in any given day will get any attention. Now imagine competing against 2000 AI games churned out every day by creatively bankrupt grifters without a passionate cell in their body.
People with the motivation to create things are already doing it, so I don't see this as a boon. Shovelware will blot out the sun.

>>105335
that's why we have AI driven recomendation systems!

The writing would be worse than a human so it would not be great for story telling and character creating. Plus this >>105335

However, yes it could create open ended games that could not be made otherwise so it could create that niche. But even that probably would not be as great as it sounds.
That Shopkeeper would have inconsistent and poorly written dialogue and he would react in bizarre and game breaking ways to things. It's the kind of thing that would work in a city builder or something like that where you don't really care about these things and you want a bizarre and living and breathing world.

Not really related to the OP topic, but it would be amazing if the rougelike genre was expanded to also include proceduraly generated quests, story and even game mechanics.

>>105335
Are you sure you're not conflating games with AI generating contents, and games created with AI generated contents?
There will be no demands for the latter type once a former type that does everything they can plus much more is released and outcompete every single one of them.

How do I use AI to kill Jashin-chan?

When it comes to the CYOA potential of AI I have some sincere concerns with its current capacity for generation along the lines of a consistent story and how much input you need to keep things on track. From what I've seen the "token limit" that exists right now extremely limits the background lore you can add for your AI to adhere to, like I was trying to add some MGE lore to make a CYOA for myself in novelAI but with the 4k limit I was able to only fit like 6 entries within it while really cutting down on context and detail. Not good, but it seems like the potential is rising as better models are developed. However another issue with those remains as the GPU power you need for them is immensely high.

Then from experience despite using tokens and trying to make a contained story, the AI still has some hangups. Where it heavily relies on user input to create and follow a structured story. I see this as an issue where the AI lacks some kind of focus and the marvel in it right now is still in the fact that it can do this to any extent at all.

And even if you are able to get something from one of the mainline sources that fits both criteria and generates within a reasonable time, you're most likely using a creatively neutered model that will never give something truly satisfying with all the corporate censorship in place. Until /g/ or some other place collectively gather together their autism to upgrade that facebook leak model to perfection there's probably not much of an avenue for AI CYOA games.

Honestly, AI has been going at such a breakneck speed that I can't even tell how it'll look a month from now.
I genuinely can't tell how far it can go right now, and I'm pretty tech savvy and can usually almost immediately tell what the true potential is behind most technologies.
In the case of games, I THINK it won't be able to generate good FPS games, since the movement of the player is the most important but also most vague thing in shooty games. At least, that's my presumption. Before SD came around I was sure that AI wouldn't be capable of producing art, and that ended up being the FIRST thing that AI did convincingly well.
That being said, RPG games? It's already possible, you just need to jerryrig stable diffusion and some text generator together with a healthy amount of tuning. The writing won't be mindblowing in the first renditions (...probably), but that gap in quality can be bridged simply because you can generate the game to match your taste, including whatever heroines you want in them while still being able to leave it up to the AI to generate them with enough randomness so that it doesn't feel like your own creation.

Personally I'm looking forward to an AI visual novel generator. People have already been using text AI for smut, going that one extra mile and getting AI to generate the CG and maybe sprites if necessary, that'll be crazy

>>105347
AI smut is very far away with none of the big tech AIs wanting to be associated with it

>>105347
>>105352
There are already Ai written VNs on F95 zone. I don't know what they are like, they are not the type of thing I would read anyway.

I think AI team mates would be interesting, you could play multiplayer type games but with bots you can tell what to do. I think it would be really good for games like Space Engineers, Escape from Tarkov and Squad.

>>105354
I wonder what the sort of people who swear out their teammates for screwing up will think of that. The AI will make stupid mistakes like an NPC in an escort mission. But you can't chastise an AI, it doesn't have feelings.

>>105360
I couldn't explain why but I feel I'd feel more guilt cussing out an AI than I would a human. There's some weird innocence to AI, though I guess it's mostly because AI chatbots actively avoid being hostile while fleshbots are a lot more prone to getting angry?

>>105332
What I'm more interested in is humans using AI tools to express themselves. Video games are probably the strongest contender for this, because there's already precedent for it with procedural generation.

I'd love to see something like Sburb form Homestuck (sans the reality altering part, obviously), where the broad strokes rules of how the game works are set in place manually by a real person, but the specifics are all unique to you the player, with the way you interact with the world having a real impact.

>>105347
I suspect the current wave of AI is going to do what every other AI wave since time immemorial did, which is to say, dry up overnight. You can only go so far with "the same thing but more powerful", especially since "more powerful" has diminishing returns

We'll see plenty of games use AI for character dialogue and limited character actions long before we ever see games made mostly by AI.
I doubt AI will ever be able to generate an entire game by itself, they would have to solve AI coding completely before that's possible and imo AI coding is currently shite and its future overhyped.

>>105370
Remember the DOTA 2 AI? After some huge initial success on a limited game version, they ended up retiring it still not being able to play on live. And now 4 years later there's been essentially no progress in MOBA AIs since then. So yeah, "dry up overnight" is pretty much in line of expectation.