pouët.net

AI art in compos

category: general [glöplog]
Now for the interesting question; can AI art be copyrighted? Copyright law applies only to humans right? Prompts should be copyrightable.
added on the 2023-01-03 12:51:29 by El Topo El Topo
Australia and South Africa have clearly seen the light.
added on the 2023-01-03 19:16:17 by El Topo El Topo
@farfar: Do I understand this article correct that the problem is that AI art cannot be copyright as an artwork created BY the AI because the AI isn't human?

Because this won't exclude copyrighting AI art to a human who created it using an AI.
added on the 2023-01-03 21:51:10 by gaspode gaspode
@farfar: Oh, I misunderstood the article. The problem is that Thaler tried to copyright the work he received by his AI as an commissioned work.

But I'm surprised that this is not possible because an AI is a piece of code. And when I code a software that generates images on a computer this result CAN be copyrighted AFAIK.
added on the 2023-01-03 22:00:56 by gaspode gaspode
Well Like I already wrote - first comes a new tech - and then the regulations. So - I think it is just a matter of time. ai-art-protest-disney-characters-mickey-mouse
added on the 2023-01-04 00:47:40 by titus^rab titus^rab
And if that not enough - not later than that moment where music can be generated as excessive as the images already - the music industry will turn up.
added on the 2023-01-04 01:04:24 by titus^rab titus^rab
Quote:
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.


who needs promptbase when you can just ask chatgpt to write it for you? :D
added on the 2023-01-04 12:47:21 by smash smash
.. trained on scraped prompts from promptbase!
added on the 2023-01-04 13:02:07 by farfar farfar
they pretend to not know each other!

BB Image
AI playing stupid is tediously obvious.
added on the 2023-01-04 16:56:25 by El Topo El Topo
Quote:
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".

15ish years ago, I predicted something like this to happen, as the next iteration of 64k intros after the "coder creates tool, artist uses tool" era. So far I've been disappointed...
added on the 2023-01-04 22:07:12 by Blueberry Blueberry
That era has ended? Sad face .. :/
added on the 2023-01-05 10:17:47 by farfar farfar
Quote:
That era has ended? Sad face .. :/

Well, evidently not. :)

That approach still delivers increasingly amazing 64k intros, and that is all well and good.

My prediction was (and still is) that eventually, we will start collaborating with AIs to produce content that no artist could think up (let alone create) on their own.

The reason that these AI advances come to images and text first is that these are well-understood, mainstream media that a lot of effort has been put into generating. There is nothing in principle stopping an AI from producing demos as well. It's just a more complex medium and therefore more difficult, and there isn't (yet?) a huge research effort behind it. We may need the help of some AIs to get there. ;)
added on the 2023-01-05 13:57:06 by Blueberry Blueberry
I suppose that'd be a logical step after online-only parties - if you don't need to meet people, you don't need to involve them in demomaking either.
added on the 2023-01-05 14:34:09 by Gargaj Gargaj
Quote:

who needs promptbase when you can just ask chatgpt to write it for you? :


no joke. a fellow demoscener who is into ai stuff and code some stuff around this topic, did an set up an ai learned programm for prompts on prompts that are already out there and uses this as random input for stable diffusion. so prompts are just one thing: lame !
added on the 2023-01-05 16:24:43 by titus^rab titus^rab
xzibit would be proud
dear sceners,

After 14 years working with my old computer
I have hands on a modern one.

The last 2 days I’ve been experimenting with
sd and it is crazy what is possible.

on its own, however, it is not satisfactory for me

Ok, it is cool to see how my own work would look
Like when another artist would have done it, but
I will go back now and explore other things with it:

These two things came up to my mind:

Since I’ve been learning to make music for a while now
I´ll test the following:

• Make a beat - turn this into a graphical version of it
(I think its called spectrogram)
Then put this through stable diffusion - make it sound / look
Like Dali :) and then mix it here and there with photoshop
With content aware fill - with the original one - and then back to audio - lets see

• Making a 3D Model in Zbrush - create a low poly version
And a displacement map from the hires version - and then
I´ll try what it will look like when this displacement map will
Be sent through sd on the resulting geometry later. unfortunately it is not 16bit atm

I think sd is maybe a door for people that have no idea about art
And play with it. And maybe some people will be wanting more
And pulling out a painting program, a sculpting tool etc. and
do something on their own. We will see what this tech will bring up
- and what will be destroyed / or how it will change.

Ps: the approach of the zenta party: reverse engineering a given
Image is a cool idea.

Lets see,

Cu, love

Cosmic
erm - I mean: atm 32bit exr for a proper displacement map.
and what I mean: maybe some people learn with these tools
how good images are composed - in terms of flow, composition, rhytm etc.


In this example the AI-art ist not the problem but the dumb Mod, because in the near future AI can copy every type of art and then he has to ban every member of his subreddit.
added on the 2023-01-07 18:08:40 by gaspode gaspode
Quote:


In this example the AI-art ist not the problem but the dumb Mod, because in the near future AI can copy every type of art and then he has to ban every member of his subreddit.


And now take that further and do you still enjoy the AI...
added on the 2023-01-07 20:14:05 by Serpent Serpent
Can someone reccommend a book that starts off with the fundamentals of Machine Learning and then proceeds on to touch on the subjects that enables f.ex Stable Diffusion, DLSS and Super Resolution?

There are probably lots of online resources that could be strung together, but it's a big topic so perhaps starting off with a good old 500 page book would be a good idea :)
added on the 2023-01-08 06:06:10 by rloaderro rloaderro
So many "jobs" have been "taken" over the history by new things appearing. There's nothing new here.

Why should AI not be permitted in compos, but high level languages or even whole 3D engines be OK? Why should music with prerecorded samples be allowed? Or demos streaming the music from a compressed audio file? The reasons to not allow AI art in prods clash with the reasons to allow many other things in other areas.

I'm sorry, but "but I had to spend so many time learning to do this and now everyone can do something similar with no effort" is not a valid argument for anything. Just ask coders about that.
added on the 2023-01-09 14:00:20 by Debvgger Debvgger
https://www.deeplearningbook.org/ is a great textbook, IMHO.
added on the 2023-01-09 15:43:53 by Adok Adok

login