Are procedural stories the future of RPGs?

[Yosharian]

Apples and oranges, Dwarf Fortress isn't an RPG.

 

[NatureOfMan]

If there's procedural generation like in Dwarf Fortress but in a game with modern graphics then it's going to be quite the shitshow to run.

 

[Makabb]

Apples and oranges, Dwarf Fortress isn't an RPG.

adventure mode

 

[user]

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least. Human brains will eventually be depreciated. :'(
https://www.theguardian.com/technol...musk-backed-ai-writes-convincing-news-fiction

 

[V_K]

The entire argument in this thread is "wow, simple algorithms are bad therefore all procgen will forever be bad", you sound like Luddites.

To be fair, it's an equation that works for most RPGs. Unexplored is the only one to my knowledge to do something more advanced with its level generation - and it's been played by, what, 6 people here?

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least. Human brains will eventually be depreciated. :'(

The limitation of computer creativity, at least at the moment, is that it's unavoidably formulaic. You can train an AI to imitate certain existing styles, and even produce variations in those styles, but it cannot properly innovate as it lacks self-awareness. Like, you can have an AI design tabletop wargames in all shapes and sizes, but it can never make the step from wargames to DnD.

 

[rusty_shackleford]

The entire argument in this thread is "wow, simple algorithms are bad therefore all procgen will forever be bad", you sound like Luddites.

To be fair, it's an equation that works for most RPGs. Unexplored is the only one to my knowledge to do something more advanced with its level generation - and it's been played by, what, 6 people here?

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least. Human brains will eventually be depreciated. :'(

The limitation of computer creativity, at least at the moment, is that it's unavoidably formulaic. You can train an AI to imitate certain existing styles, and even produce variations in those styles, but it cannot properly innovate as it lacks self-awareness. Like, you can have an AI design tabletop wargames in all shapes and sizes, but it can never make the step from wargames to DnD.

How many RPGs have innovative stories?
Being able to simply rehash thousands of good stories into a new, unique story would be a step above most RPG stories.

 

[user]

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least. Human brains will eventually be depreciated. :'(

The limitation of computer creativity, at least at the moment, is that it's unavoidably formulaic. You can train an AI to imitate certain existing styles, and even produce variations in those styles, but it cannot properly innovate as it lacks self-awareness. Like, you can have an AI design tabletop wargames in all shapes and sizes, but it can never make the step from wargames to DnD.

Don't think it's that big of a leap. In machine learning it's just a matter of training the "AI" enough, with gazillions of stories, books, literature etc that already exist like rusty says. With the right algorithms it will be able to even dish out new concepts, if it's fed the things that make a concept tick.

 

[laclongquan]

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least.

Hell, even now AI generated stories can be looked at as something manmade.

Something badly written, I mean.

And it's not as if something rare. 99% of written materials for amateur novels is badly written on Net, now that they no longer have a quality check gate called "publisher's draft reading sessions". Everybody and his dog can publish whatever the fuck they can write in a drunken/high moment, and call it good.

But I say this: present with two badly written stories, one is by AI procedural algorhim, and another is by an amateur you never heard the name before... More people will choose the amateur's version than the code generated.

It is gamers' psychology. I cant presume to speak for readers (as they are too bloody diversed) but I can safely speak about gamers.

 

[V_K]

How many RPGs have innovative stories?
Being able to simply rehash thousands of good stories into a new, unique story would be a step above most RPG stories.

I'm not arguing against that, for most kinds of entertainment formulaic is more than enough. I was addressing a more general point.

 

[Atrachasis]

I highly doubt that procedural generation will play much of a role in designing RPG narratives for the foreseeable future - but not because I would doubt its theoretical potential. But where would the pressure come from to develop and implement something like that? "Ideas guys" who can come up with a narrative at the same level of quality that a procedural algorithm can are probably a dime a dozen, there aren't a lot of costs to be cut there. And after having procedurally generated your plot, you'd still be lacking writing and art assets.

I just don't see the commercial incentive or the ROI for any major developer to make such an investment. Would the appeal of being able to play auto-generated shovelware at no additional cost really translate into that many additional sales to outweigh the investment into the algorithm? Compared to, say, hiring a random guy from an internet RPG forum to come up with a shovelware plot (and selling it as DLC)?


Dwarf Fortress is a different animal, of course - the appeal comes from its game mechanics, not the story, and it has no problems reusing the the same assets for each instance.

 

[Atrachasis]

Don't think it's that big of a leap. In machine learning it's just a matter of training the "AI" enough, with gazillions of stories, books, literature etc that already exist like rusty says. With the right algorithms it will be able to even dish out new concepts, if it's fed the things that make a concept tick.

It is one thing to train an AI to imitate sentence patterns and word sequences. An unsophisticated one might simply use Markov Chains to string words together, more sophisticated ones might have internal variables keeping track of what the text is about, and prioritize the inputs accordingly.

But keep in mind: You can only train an AI on STRUCTURED data. To auto-generate a narrative, someone first has to go and convert these "gazillions of stories, books, literature etc." into structured templates, defining predictors and target variables in the process, that you can then feed into your ML algorithm.

 

[Atrachasis]

Not in our generation. I mean AI and procedural stuff is a young field. AI can compose a beautiful symphony now. In the future, AI could probably write a new dostoevsky novel without Dostoevsky actuAlly being involved and it probably can create another planescape torment. But for now this is all sci fi shit. A possibility that we dont know how to reach yet. For the near future, most great RPG will still be created mostly by fleshies, but i am willing to support people who dare to try to build the foundation imo. ( the guys from archmage rises tried, but failed in the pipeline due to unforeseen circumstances)

i still genuinely think old brain in a person's noggin still the best way to create entertainment, but i dont belittle people who try to revolutionize AI in making entertainment.

After all we achieved much with music composing, even technically it is far less complex than videogame or movies.

https://qz.com/864199/you-probably-...en-bach-and-music-written-by-ai-in-his-style/

You are expressing a sentiment that I can get behind - but the example you are quoting is actually much less impressive than it sounds at first. Baroque music should be a comparatively low-hanging fruit, because it was bound by very strict rules, but that AI doesn't even "create" the principal melody, it just adds harmonies to a given melody. And even in doing so, it commits tons of grievous errors, such as hidden parallels, demonstrating that it hasn't even "learned" the most basic rules of contrapuntal composition that were widely known back then. It's cargo cult composition, sounds like the real thing at first, but ain't.

 

[Lacrymas]

If they were hiring professional writers for each game then a story generator would be really valuable. But most of the time I think it is an amateur writer at best, usually not even a writer, some programmer or artist that thinks they can write will give it a shot. And that's why you end up with so many weak stories in games.

They ARE hiring "professional" writers, but when the programmers gave it a shot in the olden days, the stories came out better.

You are expressing a sentiment that I can get behind - but the example you are quoting is actually much less impressive than it sounds at first. Baroque music should be a comparatively low-hanging fruit, because it was bound by very strict rules, but that AI doesn't even "create" the principal melody, it just adds harmonies to a given melody. And even in doing so, it commits tons of grievous errors, such as hidden parallels, demonstrating that it hasn't even "learned" the most basic rules of contrapuntal composition that were widely known back then. It's cargo cult composition, sounds like the real thing at first, but ain't.

It sounds like what a composer that isn't on Bach's level would produce, there are some very awkward sounding passages, but I can imagine them being written by the standard church composer or apprentice back in the day. It sounds stilted, there is no clever use of sustained dissonances, or almost at all, the rhythm is basic, there's a lot of parallelisms and not only hidden ones, it's like a harmonization exercise. And yes, this is probably what we can expect from procedural stories, functional but stilted.

 

[Optimist]

I wish I were procedural.

 

[Nutria]

If you released Covert Action today but it wasn't limited by disk space and could take up 14mb instead of just 1.44mb it would blow the fuck out of anything being released now. So no, don't tell me that procedural generation is just a buzzword and doesn't have any practical application. Microprose was going in this direction for a while also with Sword of the Samurai and Darklands. It died out because when CD-ROMs were introduced it was easier for some Japanese guy to make stupid narrative games that drooling retard weebs would worship. That doesn't mean it was an evolutionary dead end.

 

[Wunderbar]

I don't think so, no. Same reasons why we don't have machine-written books or machine-directed movies.

i think disney are already doing machine-directed movies, just look at all of these marvel super heroes.

 

[Ysaye]

If they were hiring professional writers for each game then a story generator would be really valuable. But most of the time I think it is an amateur writer at best, usually not even a writer, some programmer or artist that thinks they can write will give it a shot. And that's why you end up with so many weak stories in games.

They ARE hiring "professional" writers, but when the programmers gave it a shot in the olden days, the stories came out better.

You are expressing a sentiment that I can get behind - but the example you are quoting is actually much less impressive than it sounds at first. Baroque music should be a comparatively low-hanging fruit, because it was bound by very strict rules, but that AI doesn't even "create" the principal melody, it just adds harmonies to a given melody. And even in doing so, it commits tons of grievous errors, such as hidden parallels, demonstrating that it hasn't even "learned" the most basic rules of contrapuntal composition that were widely known back then. It's cargo cult composition, sounds like the real thing at first, but ain't.

It sounds like what a composer that isn't on Bach's level would produce, there are some very awkward sounding passages, but I can imagine them being written by the standard church composer or apprentice back in the day. It sounds stilted, there is no clever use of sustained dissonances, or almost at all, the rhythm is basic, there's a lot of parallelisms and not only hidden ones, it's like a harmonization exercise. And yes, this is probably what we can expect from procedural stories, functional but stilted.

My theory teacher would have failed that "harmonization" effort, and even that would have been generous for her, but she was always very strict on the rules - I have to admit being surprised it couldn't do better given it was given the subject to work with...

 

[Haba]

You can have "AI" writing average Forgotten Realms novels in a few years, if anyone is so inclined. Few years more and you will have compooters coming up with unique ideas that no-one had come up with before.

And then we will have the machine overlords producing custom-generated waifu-porn catered specifically for boosting your own fragile ego. And after a few years mankind will perish.

 

[user]

Don't think it's that big of a leap. In machine learning it's just a matter of training the "AI" enough, with gazillions of stories, books, literature etc that already exist like rusty says. With the right algorithms it will be able to even dish out new concepts, if it's fed the things that make a concept tick.

It is one thing to train an AI to imitate sentence patterns and word sequences. An unsophisticated one might simply use Markov Chains to string words together, more sophisticated ones might have internal variables keeping track of what the text is about, and prioritize the inputs accordingly.

But keep in mind: You can only train an AI on STRUCTURED data. To auto-generate a narrative, someone first has to go and convert these "gazillions of stories, books, literature etc." into structured templates, defining predictors and target variables in the process, that you can then feed into your ML algorithm.

Afaik not necessarily. In "unsupervised learning" or more sophisticated deep learning the machine would be able to structure the data by itself.

In the not so distant future, AI will probably be able to create stories that will look as if they were man made. It's possible now, it's just that none has attempted it for games at least.

Hell, even now AI generated stories can be looked at as something manmade.

Something badly written, I mean.

And it's not as if something rare. 99% of written materials for amateur novels is badly written on Net, now that they no longer have a quality check gate called "publisher's draft reading sessions". Everybody and his dog can publish whatever the fuck they can write in a drunken/high moment, and call it good.

But I say this: present with two badly written stories, one is by AI procedural algorhim, and another is by an amateur you never heard the name before... More people will choose the amateur's version than the code generated.

It is gamers' psychology. I cant presume to speak for readers (as they are too bloody diversed) but I can safely speak about gamers.

Yea probably true. But I bet that will eventually change as well. Machine racism will be a thing and we will be like "when I was your age blahblah was made by real people".

 

[Grauken]

You can have "AI" writing average Forgotten Realms novels in a few years, if anyone is so inclined. Few years more and you will have compooters coming up with unique ideas that no-one had come up with before.

And then we will have the machine overlords producing custom-generated waifu-porn catered specifically for boosting your own fragile ego. And after a few years mankind will perish.

 

[Lacrymas]

The matter of fact is, it can work, but it would be like a not-good author who doesn't understand characterization or causal relationships beyond the most basic.

 

[user]

The matter of fact is, it can work, but it would be like a not-good author who doesn't understand characterization or causal relationships beyond the most basic.

Why? Even the deepest most intricate shiet, feelings or behaviors and whatever, are just data in the end.

 

[Falksi]

Lazy cunts. Just put some effort in to making good games you slacking wankers.

 

[Lacrymas]

Why? Even the deepest most intricate shiet, feelings or behaviors and whatever, are just data in the end.

Are they? Or are they a unique feature of the human experience? People don't act consistently when confronted with unusual circumstances. The mechanics of emotions and thoughts are known biologically, but how we react to them is informed by our whole life and experience up to that point. It's not as simple as "my wife was killed, beep boop, I must have revenge, beeeeep". I don't think intricacies and details that make sense can be procedurally generated.

 

[Haba]

Why? Even the deepest most intricate shiet, feelings or behaviors and whatever, are just data in the end.

Are they? Or are they a unique feature of the human experience? People don't act consistently when confronted with unusual circumstances. The mechanics of emotions and thoughts are known biologically, but how we react to them is informed by our whole life and experience up to that point. It's not as simple as "my wife was killed, beep boop, I must have revenge, beeeeep". I don't think intricacies and details that make sense can be procedurally generated.

Why not? You're underestimating the power of computing. You can have compooters analysing all of the English written prose mankind has produced. Certainly we are ways off still, but there are no technical limitations even today.

And conversely, once the patterns have been identified, compooters can come up with things that human writers might only chance upon.

At the moment there simply isn't enough financial incentive to pursue those avenues, it is easier to go for the low hanging (and extremely well paying) fruit first.