Remaking Super Mario World in Godot
youtube.com
youtube.com
I've sat through a few game development tutorials - mostly Unity and Godot ones, because one day I'd love to create the game which puts the Monkey Island series to shame.
Upon witnessing how much work it is to actually get something to happen in a single scene, to get things to behave in a non-broken way using one of these engines, I'm not sure I'll ever have the patience to get there. I don't imagine having much of a problem coding the game logic (character says "Yes" door unlocks and becomes interactable etc) but having everything from the art style, scenes, sprites, and applying it all in a non-sucky, tasteful way? That's high art.
And yet game devs are inexplicably being fired like there's no tomorrow. No justice, I say. None whatsoever.
Great artwork feels like something impossible to deliver, but artists think exactly the same about code. I say this after witnessing how quickly one of my friends creates great looking art while being impressed by some trivially easy programming feats.
I would use an engine for 3D, but not 2D. For 2D work an engine like Godot is like trying to use a no-code platform to build apps, you end up with a bespoke and unmaintainable hindrance when you could have just built yourself a good foundation from scratch with slightly less effort. Leverage Copilot to get you up to speed with the boilerplate.
And the premise that a project within an engine is a "bespoke and unmaintainable hindrance" as opposed to a pile of raw AI generated code is simply laughable. If you're going to use Copilot anyway, why not also use an engine? It's obvious at that point you care less about code and code quality than getting a product finished as quickly as possible.
When everything is code, it "makes sense". You have a main loop, you can reason about what is happening when. You can set a breakpoint and walk through everything step by step.
I have no idea what's going on in Unity or Godot. Everything is completely unfamiliar, there are a bunch of abstractions I'm unfamiliar with and I can't rely on any of my past experience. I'm basically starting from scratch.
If I was tasked with making a game quickly, I could probably do it faster in SDL or Libgdx (if that's still around) than I could learn unity or Godot.
I'm not disparaging low-level coding here, I write projects in C, LuaJIT and SDL myself. But I know from personal experience how liberating it can be to not have to deal with 90% of what amounts to uninteresting boilerplate and focus on the actual creative part of a project.
I worked alone and used SDL. This was also one of my first attempts at making a game.
I managed to make a couple of command line inputs and outputs in the terminal, and then use SDL to render a single sprite on screen that did not move or anything. Needless to say, my entry did not win any prices :p
https://www.smwcentral.net/?p=section&a=details&id=28856
And created enough resources for that game that they basically know the engine inside and out. They definitely have the knowledge to finish a remake like this, and it'll be interesting to see how it goes.
My one worry however is that remaking any sort of finished game as a fan project tends not to end too well. If there's one type of fan game that companies tend to shut down, its remakes, with ones that clone the original game 1:1 being the primary target (see the Link's Awakening DX PC port recently). So I'm also not sure how long this project will last.
If you have any other SMW hack recommendations, please share!
Because there are some hacks I'd recommend if you're pretty experienced with tough games, but which would probably be too much if played casually (JUMP/JUMP 1/2/YUMP/Send Your Regards to the Czar)
For something a tad less punishing (but still tricky towards the end), The Second Reality Project Reloaded and The Second Reality Project 2 Reloaded are excellent choices:
https://www.smwcentral.net/?p=section&a=details&id=16696 https://www.smwcentral.net/?p=section&a=details&id=9543
And SMW2+3: The Essence Star is pretty good if you want a more relaxed (albeit story driven) experience:
Why bother with laborious programming and traditional modeling and rendering techniques, when AI can generate playable photo realistic worlds in seconds based on prompts by the player itself?
Tech like Google's Genie is still primitive in comparison, but we're a few short generations away from surpassing what can be done with traditional game programming.
Personalized media and interactive experiences are just around the corner, which is both scary and exciting.
Second, you assume players know what they want to play so they can prompt it. You also assume this in a broader sense on all "personalized content". I think that the value of totally AI created media is completely unproven - if you only get what you can describe, it's going to a boring time. We like content envisioned by other people.
For an AI to actually create a meaningful, full length game, movie or series - we might be decades away if this is even possible without AGI.
Tl;dr: learn programming.
ChatGPT came out a bit more than 1 year. Now we have 60 second highly realistic and creative videos generated. Movies aren't that far out. Games aren't either.
Sora is very cool and definitely shows progress but it is still far from what's needed for a story where people interact with each other or the environment.
It was incredibly hard to think an AI would beat someone in 1v1. Yet, they did it. There's no reason to believe an AI won't beat them in 5v5, even though the problem seemes orthogonal to the problem of 1v1.
60 seconds is mind blowing during just 2 years. Progress will accelerate, 2 or 4 years might be enough to fill the remaining gap.
But that won't be a reason for the majority of people who just want a product quickly and with low effort.
Because art - of which game development is a subset - is a form of creative expression, intended for humans to communicate to other humans. Note that this is a separate dimension to the purely technical aspect of content creation, which AI is increasingly capable of.
AI doesn't create art, it generates product. If your only view of programming is labor to generate product, then yes, maybe you should quit, because it's likely AI will replace you. But if you consider programming, asset creation, design etc as means of creative expression, then AI is unacceptable, because it can only express generalities. None of those games are going to be fun, engaging, thought provoking or expressive. You're never going to get the unique fingerprint of the author. That such games may be technically complete and minimally viable products is besides the point. It's just extruded entertainment product.
Human creative expression being suboptimal and maladaptive in a world where all culture is generated by machines only makes it more necessary.
I think you're vastly overestimating what entertainment means for most people, and underestimating the capabilities of AI today, and in the very near future.
People spend countless hours in idle clickers, chance games, and many other "simple" time wasters. Entertainment doesn't have to be thought provoking.
Though what makes you think AI couldn't craft such experiences as well? Can't it already come up with thought provoking stories today?
FWIW, I used to think like you a ~year ago. Art is art, AI can't replace artists, etc. And to some extent I still think that will be true. There will always be a demand for human-created art and products.
But for the vast majority of people, the ability to say what they want to experience, and be able to control all aspects of that experience, will be much more engaging and personal than anything another human could ever create.
On the contrary, I think AI can and will replace artists, and the ruthless, cynical exploitation of artists and commoditization of talent by companies using AI is what bothers me. I can even concede that AI generated art can be considered art (as generative art is already established as such) - I just don't think the wholesale replacement of human creative effort with AI results in an equivalent exchange of value. It's fine for the low-effort stuff that never meant anything to anyone beyond being a means to making a quick buck, but that isn't what all creative effort should be. Not everything is cookie clickers and time wasters.
I mean, look at the debacle around the Willy Wonka "Experience" in Glasgow. That was mostly generated by AI. That's the future of everything.
>But for the vast majority of people, the ability to say what they want to experience, and be able to control all aspects of that experience, will be much more engaging and personal than anything another human could ever create.
You can't control all aspects of that experience with AI, that isn't how it works. The only way you can do that is by actually doing that, and that requires effort, and study, and hiring people. The end result of AI can't be personal, because AI can only work with what already exists within its training set. And anyone with the same model and settings can create exactly the same thing, which renders any intent on the part of the human being meaningless with regards to the end result.
Huh? You can already do that with text and static images _today_. Doing that for video is almost here, and long-form entertainment will be next.
> The end result of AI can't be personal, because AI can only work with what already exists within its training set.
That's not true, even today. Techniques like RAG and very large context sizes can augment the training data with anything the user wants.
The end result can indeed be very personal. Imagine interacting as yourself (so not a role or character) with your deceased relatives, favorite celebrities, romantic interests, etc. They will refer to you by name, and know detailed information about you. No video game programmer and artist can craft such experiences for you, let alone in the time it takes you to think about it and communicate it to the AI.
Is an AI generated facsimile of my dead relatives supposed to mean something to me? It's not a person. It's a thing generated from what I can only assume is their publicly accessible social media data and probabilistic conjecture, taken without their consent.
Celebrities? Again, it isn't them. AI generated Einstein or James Cagney can't offer up any real insight, because they can't know anything not already in their data set that isn't randomly generated. It's all superficial.
Romantic interests? I mean, sure, you could strap on vr googles and a fleshlight and fuck the AI generated facsimile of your crush but it isn't the same. It's just closing your eyes and beating off with extra steps, and probably a subscription.
I think you're confused about the difference between "personalized" and "personal" here.
> No video game programmer and artist can craft such experiences for you, let alone in the time it takes you to think about it and communicate it to the AI.
Of course they could, people already do it, it's called identity theft. There is nothing an AI could do that human beings couldn't do given time, because AI works from human-generated data. That AI could do it faster is a bit of a red herring, since we're talking about quality of experience, not speed of turnaround.
I can tell you're engaged with the fantasy and it isn't likely that we're going to find common ground here. But I'm looking at what AI is already being used for and all of it - all of it is worse than the human-created content it replaces.
I wish I could set a /remind for this thread and come back in 5(?) years to see who was closer.
I will just say that 5 years ago the state of AI[1] was quite primitive compared to today, and 10 years ago modern advancements were practically unthinkable. There is no sign of progress slowing down, and we can assume that it will only accelerate. So I suspect that 5 years from now it will be even more entrenched in and transformative to our lives, in ways we haven't imagined yet.
Guy knows his audience.
Some of the most known:
* Mario Teaches Typing
* Super Mario Bros. Print World
* Mario is Missing! (also released on SNES)
* Mario's Time Machine (also released on SNES)
* Mario's Early Years! Fun with Letters
There are also (bad) Donkey Kong ports for home computers of the 80s such as Atari 800XL, IBM DOS, VIC-20, Commodore 64.
Full list and more details here: https://www.mariowiki.com/List_of_games#PC
...There were those games from Access Software which touted their Realsound tech, or whatever it was called. The Sound Blaster was, for me, the beginning of a newfound respect for PC hardware; I was no longer jealous of my friends who had their Amigas. I really miss the awe and wonder of those days. The days where every seemingly minor advancement actually served as a huge leap in terms of the experiential qualities of games. These days people whine about ray-tracing performance and having to play stuff at < 120FPS. I get it - I just really miss the "before times". Like the first time seeing an image rendered in full 256 colour VGA. </end oldtimer rant>