pouët.net

Most creations in art are remixes of others before, because »Everything is a remix« and I don't see a problem in this. It's a rare exception when something is completely new. And even those things are inspired by art done before.

In demoscene a lot of demo-content is a remix of the same effects we saw a gazillion times before: Cubes, plasma, tunnels, greetings, scrolltexts, fractals ...

Yeah, lets make this world so generic that nothing will matter when the doom of extinction will be inevitable! \s

And to all apologists of AI generated crap, if you don't have quality standards and real values in your life, it doesn't mean that everyone should accept your generic and worthless mindset so you won't feel so miserable and alone. Go build your dystopian future where talent will be worthless on some other planet, maybe ask E. Musk for help or something.

Just have a separate AI compo if we absolutely need one and let's keep the regular Art compos free of AI stuff and ask people to provide steps to make sure frauds won't ruin the party.

"Everything is a remix" is nice, but completely lacks the explanation of origins and is, in my humble opinion, therefore just a pile of shit or gets stuck at the "hen-egg-problem" at some point.

And there still is a difference in painting/pixeling something from scratch and just letting A.I. compile a collage out of different ripped sources, even if they are blended into each other well enough.

So if we strip all these sources from the AI image generation I would guess there wouldn't be much of an output, but if you lock up a real artist in a room without any sources he still would be able to create something meaningful, which is what I'm interested in graphics compos.

I tried several of the tools and found it quite amusing for some time, but in the end I just realized there isn't even talent involved in formulate the instructions. Just simple communication skills. At least for me this was the outcome of testing it.

If people want to have A.I. compos on demoparties, I'm fine with it. I'll welcome and use the compos as toilet or coffee breaks. \o/ :D

Addition:
I know there will be someone coming up with something like "the human mind holds all the copies", but this more or less applies to basic stuff like "the face normally has eyes, a nose and a mouth" and not to what A.I. is doing with it's copy-jobs.

This idea guy approach may work better in the domain of the conceptual art but that's a rather distant from the demoscene. If we see the art as a form of human communication – which is probably a highly relevant point of view – then AI cannot take it over since the ideas have to come from humans ...and do also technical details of the image form an essential part of that communication? This would be more scene relevant.

Of course it may be also possible that one day we start perceive AI as a relevant communicator along with humans.

To say »everything is a remix« was not to the justify the way AIs »create« art. It was related to tomkh, who said that AIs cannot create original new art. Because humans cannot create completely new art, too. Every piece of art, even by humans, is inspired and influenced by other art that came before.

gaspode: it's a remix, but there is occasionally a bit of new content in every remix. Otherwise you would still remix cave paintings. The same is not true for current-gen AI art. It's just a remix.

I also see many of the problematic aspects of using AI in compos, that were already mentioned here.

But open up some thoughts, please consider this naively formulated: "Just imagine, there are new concepts out there to make computer generated graphics - why are of all things the sceners the ones who rejects this new concepts at all?"

All this new AI stuff to make gfx or even demos:

1. Nobody knows where all this will end up
2. This set of tools will get more or less standard everywhere in the next years - like i.e. typical post processing tools for gfx nowadays. Is it realistic completely ban this stuff from the demoscene at all?

I do not want to vote for a "allow AI in gfx-compos" right here, right now. But rejecting all of this sound on the other side very naive to me. Just like times where people said: "This can't be art, because it was made with a computer" or "this is not music, because people need to play real instruments, otherwise its not." And many arguments in this thread give the the memory to this kind of sentences.

And to be honest: I would be curious, if something special would happen, if sceners would totally go down this way (although at the moment my imagination only allows, to see an AI compo just similar to a low-end photo-compo).

Maybe nothing special will happen and all critics here were proven right. I could live with that. But maybe there will also also the opportunity for a demoscenish approach. Because again: new ways to make graphics with computers...

Organic materials for LHS can be packed in few kilobytes. It's the best example of gravitationnal intrinsincs known as time perception AI. You can found base for understanding of this kind of visualisations on my web site.

nodepond: some good points there, I especially agree that it will be unrealistic to ban if it's part of everyday tooling.
Also trying to compress some useful generative models to say 64k might be fun and in size-coding spirit.

I cannot speak for others, but I think the biggest concern is lack of regulations when it comes to obtaining training data. Some people just feel ripped off here and I totally agree with them.

I'm also quite concerned about putting any code now to public domain (not just art) - I bet even if you put restrictive license on it will and it probably is already used to train AI coding assistants. They are already part of everyday tooling (GitHub copilot) and you can use it to code your next demo without anyone knowing about it. I can imagine similar stuff can be created for writing shaders - and probably someone is creating a model for it as we speak.

I guess most arguments for and against AI have come up now that we are on page 10 in this thread. No one is going to ban AI, how is that even going to work? As long as it can be used to reduce the cost of labor intensive graphic design like the faux surrealist car ads I see everywhere it will be done. That doesn't mean that I have to like it or embrace it, especially since it's purpose is the take humans out of the creative process while generating sweet $$$ for venture capitalists. It's not in my interest to make their wallets fatter by using their black box technology, it's not why I make pictures in the first place. The argument seems to be that ideas can be art too. Sure they can, but for me it's also the process of creating them myself, not by using a tool that thinks for me.

I think I'm with Keops - have a separate AI compo if there's demand for it, and see what happens.

And on the tech side, while these AI models are gigantic multigigabyte in size, maybe if we have AI in 64k intros like tomkh suggests, and if they evolve into a rich 64k AI scene that can produce amazing things from tiny models, that would be an amazing way to show them (academia + industry), yet again, "how it's done".

It is unavoidable. Look at this stunning AI video:

https://www.youtube.com/watch?v=kGLo8tl5sxs

From this Youtube channel: Hueman Instrumentality

Made with Deforum: https://www.youtube.com/watch?v=lztn6qLc9UE

I wonder who will be the first group to make a demo about this. =)

Did not notice Stable Diffusion was mentioned earlier in the thread. Anyway, this is just the start, and I can't see a way back. It will improve and eventually become realtime. Boom, AI demos. =)

Since art is my hobby and not my livelihood, words can not describe how hard I avoid making AI art, AI music, AI demos and AI thingamajig :) Time will tell how long I will succeed with this.

Well, I have a similar sentiment regarding Unity, Unreal etc usage in compos, but that seems settled already and I won't try to derail the thread...

btw lol, if you ask stable diffusion to generate vector art, it even includes iStock watermarks in the output :DDD

@fra: Perhaps ChatGPT is trolling you? Just noticed how you did use the expression "actual painting skills" and she rephrased it as "traditional techniques" and "traditional media". :D

yeah, it should've added "traditional" like landline phones and printed newspapers, boomer

@ham: yup, I didn't mention specifically the "traditionnal" arts, not sure if she (is this a "she" ?) had an internal representation of digital vs trad. arts (let's find out...).

I'm a big fan of AI-generated works, I even have paid Midjourney-account. Currently I'm using it to draw a up moodboard for my next demo. Also have been using it for covers of music on Spotify etc. I'm a bit suprised by all the negativity here about what I see as a great tool. Personally I only consider the dataset of stolen images as a negative aspect, besides that I think is a great contribution if used in a proper, respectful way. I'll see if I can find the time to write up something more indepth later on.


I would welcome the day a demo can be generated by a single pronpt and even than I won't stop making demos, but just like decent AI music we're not there in a decade or so. See also the reply from Smash about building a demo from a prompt.

I think this is rather a amazing actually. It also makes a lot of sense right? There are many people in search of good prompts and a lot of non-technical, non-creative people who would like to get certain results from AI, so having a marketplace to sell & buy prompts makes a lot of sense.

Imho AI art is fine in compos as long as it is clearly stated which tool was used and the prompt is given as well. Having a separate compo for AI art with mandatory prompt and tool (chain) reveal would be fine as well.

I think the organisers of each party should simply set the rules on this according to what they personally prefer (or think their party thrives from).

If this were a 64k intro It'd be pretty impressive. In its current form it would be an OK entry for the typical wild/animation compo. ;)