[moffkalast]

> I can procedurally generate levels all day long in a video game, but for me, they're never going to be as compelling or interesting as a lower-resolution and low quality textured game from the '90s or '00s where every single tree and rock is placed with intent.

The popularity of Minecraft and other procedural games would imply that there is still a large number of people who value exploring the unknown generated content, even if it means it's not curated.

Yes the quality won't be as good, but you do get quantity instead. And the quality will improve.

[hammyhavoc]

Yup, I ran a Minecraft server for about 8 years, I loved a bunch of procedurally generated games. I've worked on games in the past and present that use procedural generation. Furthermore, the most interesting part of Minecraft was what people built, not what world the seed generated, IME. Usually we'd WorldEdit vast swathes of it away to have flat areas to build and have fun, or we'd WE with brushes and materials to hand create terrain because the procedural generation was so-so.

My point still remains that there's a very different experience in something intentional and exact. One feels very human.

The quality isn't the issue, the broad strokes without intent are.

This is like preferring impressionism over a Dutch master. Audiences for both. I'm not invalidating any of them, I'm just saying that for me, I prefer something more human and different.

[syntheweave]

For a long time I wondered where popular misconceptions about facts(usually historical) came from. Then I read about the history of the Encyclopedia Britannica and its checkered editorial history and realized I had found the cause: we relied a lot on truth-making authorities in the past, and they wrote in their biases.

So when we speak of intentional creation, we are in some ways looking a mode of truth reminiscent of Britannica: the world as someone experienced and described it, biases and all. Neither a sensory truth like "it feels cold today", nor an emergent collective belief like our rules of language.

What makes the LLM chatbot interesting is that it resembles a lens into collective belief; it starts with one bias, but you can tell it, "now explain it the way this other type of person would," and it will oblige you with its best in-character approximation, changing facts and details as well as aesthetics. It is capable of deep and radical changes to its output with minimal changes in prompting.

That's something that you could access in a limited sense with traditional PCG, but almost entirely in terms of mathematics operating on premade assets - fractals and chaos functions and so forth. Here you have a "fractal encyclopedia", working over an enormous reference library. It can, in fact, attempt to place every brush stroke "in the style of a Dutch master" if recursively prompted to do so, although in any existing implementation, you will run into technical limits with its dataset and working context. But if we look over towards diffusion model image generation, we know that that general idea does work - it's always a little less exact than the original human, but it can get us far past the uncanny valley.

And that's valuable because all the human creative stuff is built on references, too: original ideas emerge in one context, then get transferred to another. So there's a sense of, yes, you can definitely build a "make game" button here, and it'll be an approximation of a hypothetical human character that built the thing, but you can easily turn that output into a reference for something else that injects more bias and humanity into it or adds more structural rules to rigorously shape it. Often that's the actual path of creativity: make a longer pipeline of refinement with more layers to it, and you end up with an increased sense of transformative elements and intentional structure.

[lelandfe]

Not to mention the middleground of handmade pieces and putting them together in a randomly generated way. That's how a ton of games work - from Noita, to Cataclysm DDA, to Dead Cells.

[hammyhavoc]

That's quite literally how Minecraft works. That's also how games I've worked on function.

[strken]

Pretty sure Minecraft uses Perlin noise for terrain, while Noita uses entire pre-made tiles. They both have setpiece structures dotted around the terrain, but I think lelandfe was making a distinction between Wang tiles vs a gradient noise generator.

[hammyhavoc]

You have ignored the original point made entirely that intentional level design is generally far more interesting than anything algorithmically generated.

Specifically, yes, but my point was more that there's pre-created elements thrown in there too. It's a very surface-level argument of generative art versus intentional art, the technology itself isn't what's being discussed, it's the output and how interesting it is or isn't.

I'm not even saying procedural generation et al is bad, it has its place.

People use stock/library content in games all the time, whether it's textures, rigs, whole environments, or even mechanics. Can interesting things be made? Yes. Are the usually unique or innovative? Well, generally no.

What made Minecraft so interesting for most people was what people made within it, the world seeds themselves were really whatever for most people. If they were so interesting then people wouldn't be importing schematics for models, wholly pre-built worlds, hubs, and even some creative mode servers. Sure, some people like exploring the procedurally generated stuff, but once you've done that for a few tens of hours, it's boring and you've largely seen what's worth seeing.

I spent more time using WorldEdit than screwing around with seeds because I knew exactly what I wanted my server's main area to be like rather than relying on a seed.

[strken]

I was responding to the "literally how Minecraft works" part rather than the discussion further back.

That being said, world generation is a bit like a toybox, or a backdrop to a play in a theatre. Minecraft's world gen isn't worthless, and so improving the backdrop is worth discussing. The differences in the output of implementations matter a lot (to me, at least) and I think Minecraft would be much less compelling if the world was generated using a 3d implementation of Wang tiles. "A few tens of hours" is a lot of content.

Would Minecraft be even more compelling with neural networks doing some of the content creation? I think it might. Would that totally replace the need for human input and invalidate everything players build? No, but that doesn't make it worthless.

[hammyhavoc]

Perhaps!

But a few tens of hours is very little versus the quite literally thousands of hours spent running a server for 8 years and using plugins on it. Truth be told, I think even if a Minecraft world was literally a blank base plate and it was digital collaborative LEGO, it still would have been a hit.

I guess the thing with Minecraft is that it was many different things to different people. Some people loved "Hunger Games" style worlds/modes on servers, others like single player mods, or even just vanilla Minecraft.