pouët.net

AI is more like remix.

It literally isn't. AI is something that a computer spits out once you've given it something to work with. At best it's a rehash of whatever the computer is given as source. There is literally no artistry in the outcome. Name me one computer capable of creating art without been fed thousands of works while not asking it to do any one thing, and I may consider changing my mind.

Just by looking at the video that started this, looking at the mangled faces at the end of the video, I feel disgusted. As an artist, I feel violated that this sort of garbage even exists.

Can a human creating art without any learning material?

Ok. That's one way to put it, but what's exactly the harm caused here?


That's your opinion. The author made the conscious decision to feed the AI specific parts of the source material he/she deemed important. Then, I assume, the author reassembled the output into coherent video compilation. Not only that, the author further utilized his/her artistic vision and added interesting tweaked music and sound effects to the video to adjust the mood accordingly.

You are free to think the outcome as a whole is trash, worthless and disgusting, but that does not mean there's no artistic value in it. Maybe there is not for you, but there certainly is for me. The fact that watching this video gave me mixed feelings and weird emotions like subtle anxiety is good enough reason for me to say that I am happy that it exists and I watched it, few times actually.


I agree, it was very emotional, like interesting art works tend to be.

FCAI or whoever you may actually be, you should have credited Byproduct for the remix that you slowed down.

Ok, maybe AI is more like a mashup. Take different sources and make something new.

Somewhere out there in the multiverse, this is the real pri.ary hand coded version of Second Reality.

And yet, if someone makes something like that (sans AI) in a democompo, you'd pass on it.

Perhaps AI was the missing piece of the "are demos art?" puzzle all along!

It's a video, not a demo. And it's some kind of art, at least by some definition of art. Reactions were induced!

Many parts of it look like shit, but not as bad as some coder demos. Is it funny? Interesting? I don't know. With these diffusion model thingies, it seems that you just need the tiniest bit of an idea, maybe not even that, and it might produce something interesting or funny. But I don't think anybody considers themselves an "artist" just because they were able to get something to come out. Very low amounts of effort are required to have fun. It's fun, not serious art.

There's a word in Finnish that this discussion sometimes reminds me of. https://fi.wikipedia.org/wiki/Kukkahattut%C3%A4ti I've definitely been guilty of that stuff many times myself! ;)


prompt: middle-aged woman in flower dress and flower hat

Let's make it a Finnish woman



Some more Second Reality fan fiction

Lol all this AI shit is zo bad

Like garbage tier bad.

So far in 2023, 5 demoparties had AI graphics compos.

Gargaj, i, for one, would be happy to see more like this at a democompo if it's realtime.

That's kinda the point though - it isn't, and noone is making the effort for it to be.

Went immediately to the opening YouTube link and disliked it. I'm fed up of AI being mentioned here so frequently, it's almost as if the sceners WANT their creative abilities replaced by AI rubbish!

Gargaj, noone is going to make any effort in this field if you guys immediately ban *AI* and call it disgusting and ridicule it. That's the whole point.

I hardly watch any democompos let alone attend any parties so you are probably right.

But if I did, over time my engagement would depend on how the use of AI evolves in context of demos. If it all starts to repeat itself and becomes boring in terms of the creation process and the outcome, I think it would fade away on its own. If however people would keep coming up with creative ways to extend and squeeze new ideas out of it, it could evolve into something interesting.

Mostly what we've seen so far have been low-effort stuff comparable to copying a code for rotating cube and making demo out of it, that's how you start. I don't think this is all the AI is capable of giving us if people are willing to experiment and give it a chance. Since the common argument against AI seems to be the worry of exploiting others work, imagine something like this:

Transitions in demos are often neglected. What if you'd create your demo as usual and feed video or images from different scenes of your demo to an AI in order to help you generate content for the transitions between scenes? Give it a screenshot from previous and the upcoming scene and make it generate a picture from that. Now crossfade or morph from previous scene to the AI picture to the next scene.

That could be interesting on its own, but that's only the most obvious and boring way to make use of this idea. Perhaps you could sprinkle influence from other scenes to the transition or even from your other demos. Even if you end up doing the transition completely yourself, you can milk potential material for the transition from the AI. The transition can be a dinosaur walking across the screen where the dinosaur is env mapped with a texture that was generated by the AI from the screenshots of your demo's previous and next scenes.

If all your experiments and ideas end up shit in the end, at least you did not waste that much time because it's fast to work with, I assume. Even if you don't use the AI directly, you can still use it to give you ideas and visions quickly while having some control of the broad direction.

Ok? Doesn’t make it less shit? Parties can do whatever they like. I did like the idea of generating a whole demo with AI that Syntax had. But you don’t expect that shit to be good, its just funny.

I don't think that this video was disliked because it was created by an AI. The main problem of the demoscene right now is that it lacks originality (concepts and ideas that have never been done before). I keep hearing this in one way or another, by many people on the scene. So yeah this video is an example of that. It has been done before but kinda better. Nowadays, AI is mostly used to aid in upscaling stuff that then are used in productions, so it would be great to see in, say, 20 or 30 years what it will possibly do by simply feeding it second reality and asking for a modern version of it. :)

I saw it and of course I gave it a thumbs up.

Proper link.

Stop the AI hate. This is just a funny video, dammit.

Here's my thick brain take...

I find the topic of GANs fascinating and I think we'll see more and more models being either built into the OS or augmented with cloud services from OS providers.

One such application is the Descript Audio Codec:
https://descript.notion.site/Descript-Audio-Codec-11389fce0ce2419891d6591a68f814d5

A 300MB model enables state of the art ultra low bitrate audio compression (44.1khz stereo at 8kbps). (Yes it sounds bad, but outstanding compared to the competition at the same bitrate).

It seems comparable to using gm.dls in demos, only in that it's another tool to achieve an end goal.

Points raised about it bypassing the hard work are absolutely valid though. I don't know the answers, but I'm not fundamentally opposed to the technology - I remember similar arguments from back in the day (demos have to be in assembly, shouldn't be in Windows, shouldn't be accellerated etc).

Disclosing AI usage seems like the responsible thing to do, and I anticipate it will become fairly commonplace over time.

Creepy stuff.
That's how a production-hell-nightmare of the coders might've been while being knee deep in making the actual demo ;D