Dwarf fortress is going to complete obsolete as a story generating game in about ~5 years time by AI Models.

[Murderer Drug Dealer]

Mark my words. There'll be a game that comes in the near future that'll absolutely blow DF out of the water by using large AI language models. At the moment technology like ChatGPT is already insanely impressive, but it'll get exponentially better at a really fast rate. Meanwhile DF is moving at a snail's pace and using tech that's bordering on 20 years old. When someone constraints a language model AI by some rules of a game like RimWorld or DwarfFortress, it's just going to be on another level to the point where DF will seem like an ant compared to an intelligent human being.

[CommandoKal]

Or... DF will simply import/buy the licensing/develop the AI for this very reason.

[Mercury]

Originally posted by Griggith:

Originally posted by Halcyon Days:

Mark my words. There'll be a game that comes in the near future that'll absolutely blow DF out of the water by using large AI language models. At the moment technology like ChatGPT is already insanely impressive, but it'll get exponentially better at a really fast rate. Meanwhile DF is moving at a snail's pace and using tech that's bordering on 20 years old. When someone constraints a language model AI by some rules of a game like RimWorld or DwarfFortress, it's just going to be on another level to the point where DF will seem like an ant compared to an intelligent human being.

I mean Rimworld already has the storyteller AI that makes a playthrough exponentially more interesting than DF. I'm actually shocked playing this game how peaceful it is. My fortress has gone completely smooth so far with pretty much no issues.

They made the default settings easier for the steam release, play sparse minerals and hard economy and invasions for the real experience

[Mercury]

crypto bros on their new fad that will be useless or die inevitably

[Kyrah Abattoir]

Originally posted by diji:

Originally posted by Kyrah Abattoir:

I have nothing against automation, please, do automate the stuff no one wants to do.
But... why are we automatic artistic work? It's one of the few joys left in the world.

It's not so much about stopping it but about re-affirming that we want humans to keep creating art. In order to do so, society has to provide some kind of viability to artistic careers.

And the only way I can think of that keeps artists into the loop is to consider that AI models trained on their work are a form of derivative work.

Because I can't draw and can't afford to pay $50-200 for a commission every time I want art for stuff like this.
I also just think it's kind of funny that history is repeating itself by calling AI generation evil, the same way machine automation was treated by farmers, bakers, etc.
I'm not sure why they consider hold their profession sacred above others that have been/are being/will be impacted by automation. A bloated sense of self-importance, perhaps?

AI tech isn't evil, far from that.

The people who are building the tech have however been extremely reluctant to cooperate with artists whose work they rely upon.

The irony is that it all boils down to the fact that most artists simply aren't wealthy enough to sue in order to defend their rights.

[ccwscott]

Originally posted by Hambone:

The two primary issues is that for one it is legally not possible to copyright the output of an AI, making it unattractive for commercial use.

What? You can copyright the output of an AI.

The other one is the issue of sourcing material to train it with. Sample sets aren't a new thing, science has been using them for a long time and there's a process to actually being able to use images in them. As the field stabilizes that will likely become more and more of an issue, as much of the samples used to train AI currently are used without artist consent. This is particularly an issue if you say, want to monetize access to your AI, since it uses things you do not own to actually create the output product.

Doesn't really matter. As long as it's transformative enough using someone's art isn't a problem.

Frankly, AI will likely not replace artists. There's a variety of legal reasons, but also just the hows of the way AI operate that make it a non-viable alternative. However for disposable and quick art that you would never have paid to commission anyway? Sure.

It's nearly there already. Basic logos, pfps, character portraits, "draw me but as an elf fairy", fursonas - a lot of stuff in the $15-$200 price range is nearly able to be done by A.I.s or is able to be done.

[MintiM]

Try 70 years instead of 20. People rn don't even want to buy VR or upgrade their PCs, recession is huge and is gonna be even worse. The technology is partially already there, it'd just cut into profits of companies like Intel to release it to the public so early.

Imagine if your statement was formulated around the time Duke Nukem 3D released. "In 20 years there's gonna be better games that will make this one obsolete!!!". Yeah no ♥♥♥♥ sherlock. In the meantime I'm gonna play Duke.

[Miles Teg]

At some point Steam will be full of AI coded games with AI generated art. Just don't ask me to name the date.

[MASTER CHEEKS]

People have been saying this for years, but AI is so far off from writing actually good code that it's more of a headache than a reality. You're talking about literally a decade ahead, maybe longer, right now using AI to code anything complex is more of a liability because it constantly makes mistakes and requires high maintenance.

[McFuzz]

You can already use it to make furry porn.
Once people find that out, it will be demonized by the majority of people, and relegated to only being used by the furries.

Once again, furries have saved us all.

[Phagocytosis]

Originally posted by 😘MАSTЕR CHЕЕKS 😍:

People have been saying this for years, but AI is so far off from writing actually good code that it's more of a headache than a reality. You're talking about literally a decade ahead, maybe longer, right now using AI to code anything complex is more of a liability because it constantly makes mistakes and requires high maintenance.

I think a decade or more is a pretty reasonable estimate, yes, if you want to have a game of the calibre of Dwarf Fortress, and presumably still with a lot of input from humans at that point. The 5 years guess in OP might be a bit overly optimistic. But I would say a decade is still a relatively short time for something groundbreaking as that! I'm very excited for the future, assuming we all make it there.

[Remus]

Originally posted by Frank McFuzz:

What AI tech bros haven't realized is that companies are going to sieze your AI models, and monetize them to the high heavens. It will happen to every AI model you enjoy. It has ALREADY happened with AI dungeon, which now gives you like 15 posts unless you pay a subscription.

A new AI model will appear, someone like Adobe or some other Globocorp will sweep in and monetize it from out of the reach of the common man.

Welcome to capitalism.

Enjoy your stay.

preach brother. monthly subscription out the ass

[Doctor Zalgo]

Originally posted by Halcyon Days:

Mark my words. There'll be a game that comes in the near future that'll absolutely blow DF out of the water by using large AI language models. At the moment technology like ChatGPT is already insanely impressive, but it'll get exponentially better at a really fast rate. Meanwhile DF is moving at a snail's pace and using tech that's bordering on 20 years old. When someone constraints a language model AI by some rules of a game like RimWorld or DwarfFortress, it's just going to be on another level to the point where DF will seem like an ant compared to an intelligent human being.

Actually, DF just needs GPT3 bindings and its basically a bespoke RPG setting generator. A mate of mine took the description from a book generated in one of my games and used GPT3 to generate the first chapter of a holy book about mothmen who melted while they prayed to the sun.

[Doctor Zalgo]

Originally posted by Attack on Titan Fall:

Originally posted by Frank McFuzz:

What AI tech bros haven't realized is that companies are going to sieze your AI models, and monetize them to the high heavens. It will happen to every AI model you enjoy. It has ALREADY happened with AI dungeon, which now gives you like 15 posts unless you pay a subscription.

A new AI model will appear, someone like Adobe or some other Globocorp will sweep in and monetize it from out of the reach of the common man.

Welcome to capitalism.

Enjoy your stay.

preach brother. monthly subscription out the ass

I like the suggestion that companies aren't just run by people, but are instead just faceless entities that humanity is powerless to stop.

You could start up a company tomorrow that gave away AI services for free if you wanted to.

[Kamamura]

There is a lot of really stupid fanboi answers like "who cares", etc, but the OP is right.

The AI has crossed a threshold where it starts to really matter.

And who plays the game as story generators? Many people just use legends and read about the generated worlds and entities, I did it too at one time.

Dwarf Fortress is an interesting experiment, but the development approach is utterly impractical, especially in an environment like software that changes so fast.

[Doctor Zalgo]

Originally posted by Kamamura:

There is a lot of really stupid fanboi answers like "who cares", etc, but the OP is right.

The AI has crossed a threshold where it starts to really matter.

And who plays the game as story generators? Many people just use legends and read about the generated worlds and entities, I did it too at one time.

Dwarf Fortress is an interesting experiment, but the development approach is utterly impractical, especially in an environment like software that changes so fast.

If it helps, in the near future, someone is going to be able to feed the DF code into an AI and have it recode it into a modern multi-threaded programming language.