pouët.net

I have made lame low-effort prods without AI.

Here is 4.4 million items they can train their AIs with that doesn't rip living artists off: https://www.si.edu/openaccess

Let’s put everything else (and there is a lot) aside. This one aspect should be enough: when you use AI (a multi billion dollar capitalist venture ripping offeverything left and right) it feels (to me at least) that this goes against the basic feel of garage/diy ethos of the demoscene. And if things go this way I think it will mark the beginning of the end of any kind of future garage/diy ethos, as it will raise the entry bar so much in all things that it will just become unreachable for anyone not backed with millions of dollars in AI datasets and such.
When individuals (and the demoscene is notorious for this) don’t care about copyright - I salute that!!!
But, when multi million/billion dollar corporations rip off all humanity without consent - I say f*ck those m*otherf*ckers!
I wish the scene would just ignore/boycott this abomination.

I was delighted to see Revision make this provocative declaration:



I assume this means I cannot use an image that an AI spat out -- but what problem are they trying to solve? Is it because they are concerned artists will become redundant as AI owners seek profit? Or do they have a general disgust for the technology? Does AI threaten some property of the party they are actively trying to preserve?

For me - I would answer some amount of 'yes' to all of those points. However, I think this rule is impracticable and should be removed. I will be preparing popcorn in advance of the inevitable "that is an AI pic, or inspired by AI"/"no it is not" squabbles.

In the mean time, I recommend this wonderful piece of work from Rich Lloyd at Nova last year:
https://www.pouet.net/prod.php?which=92202

MusicLM examples - sounds good
https://google-research.github.io/seanet/musiclm/examples/

I think Revision has struck the right balance here, by disallowing AI art in the graphics compos but allowing it in the other compos.

If I need graphics for my demo, it shouldn't matter whether I get it from a human or from an AI, as long as I have the appropriate license to use the material and give credit where credit is due.

In the graphics compos, on the other hand, there's an expectation that the submitter is the creator of the picture, which would not be the case if it was made by an AI.

I'm still confused by what "or inspired" means, cos if it's meant literally, what the fuck.

Ah! Yes, Blueberry - thanks for gently clarifying that for me. I skipped over the headline and jumped straight to indignation:

Inspiration is hard to police :)

Obviously it is difficult for humans because our memory is very limited. I think we can do it with some kind of AI. :)

Noise2Music examples - also sounds good
https://noise2music.github.io/

that sounds godawful

If you're preparing a graphics entry for a compo, you might wanna look into this to fuck with AIs: https://glaze.cs.uchicago.edu/

AIs will soon learn about Commodore 1530 and how to adjust the tape head azimuth position and how to crack the glaze protection. And then they'll add crack intros to the cracked works. And then other AIs will be very impressed by the crack intros and then they'll start focusing on the intro stuff completely and an AI demo scene is born! Just wait.

@yzi: That might be ... and they have names like ... fAIrlight, AIsd, conspAIracy, trsAI etc.

@gAIspode: Names like gAIgART/conspAIracy? :]

@Gargaj
I'm not sure what to make of this. At the bottom of the page they have a 'proof of concept' schematic on how their method blocks the AI from generating a 'correct' result from copying an image. Hey, looks legit! Fine. One the other hand, the 'failed' generated image looks to be about 16384 times more interesting than the original it was based upon. Hm.

@rexbeng: I am not convinced. No paper --> No algorithm --> Smells like bullshit marketing?

The days without AI are gone. Welcome to the XXI century.

OK. I read it too fast. Actually, there is a paper here but I'm still not convinced because I'm sure it's possible to circumvent these "protections" modifying the AI training process.

Here's two takes on AI and how it may interact with creative activities.

This one is on the consumerist applications of arts/crafts (https://www.youtube.com/watch?v=xm7BwEsdVbQ while the other one is on the institutional, or 'pretentious' if you prefer, side (https://www.youtube.com/watch?v=G2XdZIC3AM8).

I post the following thoughts to this thread because I think they somehow fit in here.

I recently initiated a thread with the question whether there are demos coded in CIL. I got a reply that this is unlikely to be the case because while coding in CIL is difficult it does not really bring a benefit.

Now I would like to object that one of the things that have fascinated me about the demoscene has always been that it is quite difficult, and challenging, to code a demo. In fact I never admired the scene much because of its aesthetics but because sceners have managed to solve the challenging problems that are connected with demomaking. For this reason I wonder what you think - if democoding were not difficult, would you still be interested in demos? I think that this fits to this thread because AI is one means to make demomaking less challenging. Would you still be interested in the demoscene if it were allowed to submit AI-generated demos to compos?