aigan said:
A systemic story game will adapt to the player's choices. They should be able to go anywhere they want and find clever solutions to enter all those places. The story will adapt. The author can set up scenarios, but the end result could be very different depending on what the player does. If someone really tried to be good and clever they should be able to kill the big bad evil guy in the opening chapter, even if the author just intended it for the player to see what it is they are going to spend their next 200 hours to achieve. That doesn’t mean that it’s the end of the story. It just means that it will become a different story.

I thought that was clear by now. How can I explain this better?

Refining the two points which are not entirely clear to me:

1. Assuming the smart player kills the end boss after 5 minutes, how should a computer program be able to come up with a further story out of nothing? If human authoring is reduced to defining independent scenarios, how can a program connect them to form a deeper sense and drama, plot twists, and finally a new climax? Because without those things, there is no story in a classical sense. There may be communications, descriptions, even connections between past and potential future events. But that's no story. Meaningful words alone do not define a story.
2. It's clear that player want no restrictions, which means static story isn't possible at this point. And they also want a story, which isn't possible without an author. So that's a contradiction. We either aim for an optimal compromise of both, or we choose one or the other extreme. It's not entirely clear where you are here, since you mention setting up scenarios, or story templates to describe potential state and outcomes of a love triangle, for example.

To sum it up, it's unclear which work and control is left to human authoring and design,
and how the computer can make up the rest (which might be everything, including story).

And the technical question is: Which kind of data structures do you use to form an abstraction of story? What is the data? Which algorithms can turn it into natural language?

Your goals are clear eventually, but how can you make it actually work?
Without at least some vague overview of implementation, people will just think it's dreaming about features that would be nice to have. A theoretical discussion about design goals. But not something actually doable.

If people would think that's possible, then they would do it.
So you need to convince them that it is possible, by telling how it works.

That said assuming LLMs would not yet exist. Because now it is possible to generate stories from machines, as we all know. But you say you don't need that, while you don't say what to use instead.
Personally i just assume you don't want to expose secrets, which is fine.
But then you would not ask how to be more clear, since you would remain vague by intent.

Imo you don't have to be more clear, but then you have to accept the doubts until you are ready to show off.

aigan said:
Somehow FF 16 got praise for including a version of that. (“active time lore”)

Outcast 2 also got praise for displaying extra info during dialogue explaining used (alien) terminology, which is basically lore too.
I assume a lot of improvements we discuss here are underway.
If i don't stop my current retro phase, i might miss those improvements…
But i'm at the point where trying new games just to keep up to date feels mostly like a waste of time and bandwidth.

aigan said:
Actually, it’s the same thing as my principle of the quantum observer. In his version of it; the level of enemies in an area is determined by the player level the first time they enter it. When they later come back to the same area, they will experience how much stronger they have become. Thus, the game will adjust for the player power progression, and still keep the world consistent.

That's pretty good. : ) Solves at least a part of the problem i guess.
But i'm more worried about the player constantly switching play style (offensive, stealth, manipulating environment, …). If we would tune settings manually for goo balance, we would certainly find that different play style requires different settings. But i don't think classifying the play style and adjusting settings at runtime would work. It would be confusing, at lest subconsciously, i guess. But well, that's just speculation. Maybe there is no problem at all.

aigan said:
But for me… Levels are still a thing from the table-top board games. Computers can do better when it comes to simulation.

I think the same about government regulations and thresholds, actually. ; )

aigan said:
Love that movie. Have studied a lot of esoterica.

‘Esoteric’… maybe that's a good term describing what i love about those 70s scifi movies. 2001, Stalker, Solaris, also Clockwerk Orange, Rollerball and that stuff. For games, Planet of the Apes or Zardoz is a pretty practical inspiration. It's thrash, but does not assume a mindless audience, like Terminator or later Alien movies sadly do. And Star Wars ofc, which i never really liked.

aigan said:
I have been waiting for 30 years for story games to evolve. Was holding out hope that it would get better but the industry seems to stand still in that regard.

Well, wait and see. ChatGPT will surely ruin what's left. :D

aigan said:
That's why I started to write more about this topic. Hoping some people will steal the idea and start creating the games I want to play. If that doesn’t work I will continue to develop a proof of concept. I decided to call it “Adventure Alpha”, since it will sort of be in alpha forever. The best case scenario is that the five levels of implementation will correspond with Adventure Alpha 1-5.

Well ideas are nothing worth in this industry, they say.
So yes, you have to do it yourself.
Only then you well see it does not work, hihi :P

aigan said:
Adventure alpha will hopefully inspire others to do similar things. But it might also evolve to some sort of library to be used in other games.

That's a good plan imo.

aigan said:
For example, if you for each interactable have dedicated a place where the character should stand and then play an animation that doesn’t care about the physics. It should handle if some object is partly in the way such that the character must do the action in an altered position.

Yes, that's what i'll get. At least. But it's a lot already.
Animation is really the wall which stops individuals from making games.
You would need an animation clip for every action. You need to place the character at the right spot so it matches. You may need to rotate the character around the up axis, which looks silly. Everything looks robotic, unbalanced, and like shit.
I guess it's the primary reason why AAA cost so much to make.
And i really wonder why they never tried alternatives, because HW is powerful enough since we have multi core CPUs.
It's difficult ofc. For many years Boston Dynamics was the only company at all who had figured it out.
But we can do the same, and it is quite a bit easier for us. We can always cheat.
There was Natural Motion, but seems they no longer sell their character simulation middleware. No sign on their webpage. Seems a failure. : (

But well, now we have AI. And you will see, for any problem that is hard, just throw AI at it, fire some workers no longer needed, and all will be good. :D

aigan said:
The result is that the location of your gun has an effect on the world, since it affects how long it takes you to aim.

Yeah, the other result is increasing amount of players whining about latency.
Any new technology comes at some loss.
But this forces us to come up with new things, and that's really needed currently.

aigan said:
I would consider the animation systemic if it behaved physically correct.

No. It's just physically correct then. That's the proper term.
I'm glad you like it. But that does not make it ‘systemic’ yet. ; )

I see the advantage initially in more realism, like for the example of picking up an object, instead teleporting it to the inventory or hand. That's practically only a visual improvement!
However, there is no way around it. GFX is currently way more realistic than animation. Beside exceptions like TLoU, it's often even ridiculous. CP2077 for example is pretty robotic.
So we must fix this, and the fix will introduce lag affecting gameplay.
But because the realism is pleasant to watch, we can compensate the loss this way.

It will lead to slower games. So we need to move the interesting action closer to the player. More detailed interactions, which now becomes possible and practical because animation is no longer needed.

And that's the point where your systemic journey begins.
It becomes essential. It will become the norm, you will see.

aigan said:
I don’t get why games still do that in this day and age.

Because animation simply does not work, and they can not fix this from one more specific improvement.
It does not work, so it is better to accept this, get the job done, and focus on something more important.
Totally understandable to me.

Doing it with physics simulation has an advantage here, btw. If i do something that is not right, the character will loose balance and fall on it's nose.
You have no idea how hard it is to keep in balance. All our movements operate on the edge of what's physically possible, literally. It requires very high accuracy. Our support polygon is very small, but our center of mass is very high.
Because of that, i expect games will keep cheating a lot. They will use ML to make it flexible and looking natural, but they will apply external force to enforce balance.
Still puppets on strings, smoke and mirrors, but much less obvious than now.

aigan said:
(I would never not pause a game, but I have seen others that expect nothing to happen if they stop playing.)

But then you miss the most intense thrill some games have to offer? :D

aigan said:
This person I know doesn't want to play games with kids-like graphics. They prefer a realistic style, and an adventure in the style of Tomb Raider. No such game exists. Sadly. High quality story with female adventure protagonist and nice graphics. With no brutal tragedy.

The talk yesterday made me realize it's probably indeed necessary to adopt to many players, even when targeting a niche. Your example is a rare exception i guess, but you guys like to solve a whole point and click adventure just to make an arrow. While i'm indeed too lazy to select a powerup, but then press another button to finally activate it.
So the need for more options begins much earlier than i've thought.
However - since it's so difficult to come up with options, looking at it from this angle as well is more inspiration than problem, actually. : )

aigan said:
Game designers have in some parts already explored this space for graphics and music, with procedural generation and responding to the players choices. I’m just asking them to do the same for the stories. It’s still artistry, even when the stories are told through systems. It just hasn't been done for the main story in story-driven games yet.

It hasn't been done because generating a story represented by text or voice requires procedural generation of language. Many people tried this, but none of the chat bots before the LLM era passed the turing test. The programs scrambled words together, respecting rules of grammar, but it did not make a lot of sense. Neither did it understand human input.
Contrary, generating procedural images is pretty easy. Even if we just mix some noise functions, it might well imitate materials as seen in nature. It works without a deeper understanding of how the universe works, to some acceptable degree. But we are still far from the point to generate game gfx which look like real life. We are far, but colored boxes were already good enough to make fun games. That's because humans are very tolerant in this sense.
However, generating text to describe character relationships, including their conversations, will most likely fail for the same reason former chat bots have failed. There is very little tolerance here. If it's below the credibility of human written stories, people will notice it all the time and prefer human written stories even if they are static and limit their freedom. They don't expect freedom, but they do expect human level language.

Stories build on top of language. And that's why it hasn't been done. It simply was impossible until recently.

I guess you very much underestimate the problem due to desire and enthusiasm.
But even if you don't make it further than level X, it may be still worth the attempt.
You may arrive at a practical interface to join game engines and LLMs, for example.
I don't believe in entirely AI hallucinated games, and connecting traditional programs to AI surely isn't trivial.
If so, a system like yours is needed to model story and state of relationship on the traditional side. AI will care about language output, but likely AI should also affect the game model, similar to how the player does.

aigan said:
The language part I already talked about in length with the Wild at Heart article.

Not at all. You say what you want to achieve, but you do not even scratch the surface of implementation.
But that's the problem. We all want this stuff, but we can't implement it.

Now you've said you're no programmer. I saw you are a webdev on your blog, and so i've previously assumed you are a programmer.
But - if you do not know how computers and their programs work (which i don't believe), then i would just go ahead and tell you: ‘I'm sorry, but what you want is not possible with computer programs. We can not make computers talk or think. They can only do math and logic, and since we lack a definition of intelligence and thus also of the problem, we can not implement an algorithm to solve it.’

Your idea is based on your own human skill to improvise and associate. In works in your mind, but only because you have a mind. A computer does not have it, and a program can not emulate it, since we don't know what it is.

Machine Learning bypasses this issue. It's not a program. The people who have achieved ChatGPT do not know how it works. They made a system which can learn by detecting patters, and it can reproduce human behavior and skills from given examples. But we still don't know how humans come up with those examples in the first place. We don't know how intelligence works, what it is, and ML will not reveal this mystery, even if it can reproduce intelligent behavior.

Now i guess you knew all that, and you'll say ‘intelligence is not needed’.
But i think for dynamic language and story it is, and you will come to the same conclusion once you work on code, not design documents.

aigan said:
But I seem to fail to communicate the vision.

No. You fail at convincing us that the vision can be implemented.
The vision itself is fine. But believe me: We all have this vision. We just don't talk about it because we think it can't be done. Maybe sometime in the future, we thought.
And ironically, that time starts right now, due to arrival of LLMs.

You'll get what you want. However, personally i'm quite hostile against AI in general. They have ruined it for me by stealing artworks of the whole history of humanity, and then replacing the artists they have robbed. They do the same with coders. It seems AI is good for pretty much nothing, other than reducing a need for human work. It will hurt economy much more than it helps, which will distract us from working on the real problem we have: Climate change.

Personally i will not pay a single cent for an AI driven game. They steal the content, so i won't pay for it. I'll have no money anyway, since i'll have no job.
They'll have to give those games away for free, showing some adds about products no one could afford anymore.
Jensen says it's the new industrial revolution, but that's just a marketing lie as usual.
As far as i can tell, so far.

I think there will be a branch of gamedev rejecting AI, and i'm from that camp.
For me, systemic story at the cost of AI running on big tech cloud servers is no option.
It's not worth the price.
Your level 1 will be good enough for me. : )

aigan said:
So for that I am thinking of a tree of key/value tables

Hehe, obviously you are a webdev. ; ) But yes. Go all in, and see what it gives, what's the unexpected problems, refine and iterate…
I can image you can make it work so it could even generate new cases for the detective to solve, and while i think it becomes repetitive probably, that's still enough to make procedural story a mechanic. I image 2D top down JRPG visuals later eventually. Could become an interesting and inspiring webgame. The next NetHack. \:D/

aigan said:
So what happens when you kill the big bad evil guy in the first act? That’s easy. There is always something else behind the curtains pulling the threads.

Ok, but then we did not kill the bad guy in the first place at all, just some boss guarding the endboss.
It's like enforcing the werewolf has to die at the end of the quest, no matter what's the players choices, no? So the freedom is an illusion.
But i agree what that. That's how it works all the time. The goal is not to achieve freedom or X. The goal is to come closer to the goal, but never reaching it.

aigan said:
LLM as they exist today as transformers is not enough.

My tech site brought this up again today:

The german journalists were invited by NV and could play themselves. They were impressed, and speculated how this will change games, enabling new things.

Maybe. But to me it's just terrible. I can't watch the video to the end - i fall asleep long before. Most boring thing i ever saw. And i would never talk in a microphone to interact with NPCs, or Alexa, Siri, etc. Makes me feel like a slave of the tech industry. Like giving up on humanity for a Matrix in my basement.

But it surely will get better. E.g. if some proper gamedevs and artists work on that, instead those self appointed experts. Total amateurs, imo.

aigan said:
The dynamic adjustment should also be in-world. In MGS5, the enemy will install more flood lights if they realize that somebody has been sneaking around in the dark. They will requisition more helmets if they have lost several soldiers from sniper headshots. And so on for mines, cameras, and more. Other things in response to certain playstyles can be that they change passwords or keys. They check personnel against photographs. They start using guard dogs. They add more check points. They could also just call for reinforcement.

Played it quite a while, but did not notice any of this. Maybe due to lack of oversight? Or maybe the player should not notice? But actually i think Kojima just isn't as good as they claim he is. At least not for my kind of player.