@sooz: When I came across this passage you cited, only my ears got in the way of an even wider grin!
I hope that this class action will show the scrapers their limits and rights. Furthermore, it would be gratifying if it were possible to register as a joint plaintiff, so that the artist community could participate in the lawsuit and, by having as many joint plaintiffs as possible, lend the class action lawsuit much more weight. This could give the judges a sign, a measure, for the mass of those affected, and show the defendants how alone they stand.
You know, Iâm kinda of shock how long it took for lawsuits to start happening.
But the this issue degrade to the point that it is now an all out hate war/bullying.
Ai bros calling artist Luddites and anti tech, ai will replace artist, etc⌠is pure toxic.
The idea that âlet replace the field of creativityâ is downright insult to humanity. Cause most people would not be able to compete realistically, so the phase âwork harder and produce more.â Isnât helpful advice, rather it counterproductive.
â
since I decide to check twitter today, this came on my feed
At this point Iâm not surprised, and what more do I need to say.
â
There a lot more I want to say; but due to having hard time writing my thoughts as it is. I want to state this last thought about AI idealistic dream about building a Utopia.
Itâs an interesting study case on how AI works, answer is simple - it doesnât understand concept of anatomy, it analyzes arbitrary shapes and relations between them when analyzing images. It will see two crossed hands as one object, crossed fingers as possible pattern variation.
It âthinksâ hands can appear in many different forms when analyzing multiple hand photos. Human will immediately recognize what are fingers and how they work, AI wonât.
Weâve updated our Terms of Service to make clear that scraping and reselling or redistributing content is not permitted, and to clarify the prohibition against use of NoAI Content with Generative AI Programs.
We have also committed not to use, or license any third party to use, any ArtStation content for the purpose of training Generative AI Programs.
âŚmore and more weâre seeing this âletâs just call it artâ sentiment echoed everywhere â which is its own argument, but people are inevitably taking that a step further, calling AI art âtheir paintingsâ. Iâm seeing it increasingly online. People are blatantly lying to people, telling them that they âpaintedâ an image when they just hit ârefreshâ in their browser. They are building an entire online presence around this lie, and itâs no wonder weâre seeing such a vigilant, virulent reaction against it.
Well, as I explained earlier, in the first mega-topic about AI-generated pixel accumulations, we as artists must not let the term âAI artâ stand unchallenged. We have to reply to everyone who calls this art that it cannot be art at all, according to the very definition of art.
And we should also make it clear every time that anyone who types a few words into a prompt and creates an image at the push of a button has done nothing that would correspond or even come close to the creative act of creating a work of art, regardless of its quality.
Which is not to say that such an accumulation of pixels may be beautiful to look at. There is no contradiction in it, there are so many beautiful things on earth which have absolutely nothing in common with the creative act to create art. In addition, art can also appear very ugly in part, the decisive factor is the process of creation.
What is the essence of the above? We must oppose the attempt to lend a value to these AI pixel accumulations energetically and with justifiable arguments, negate this therefore, in order to represent the value of our creation and above all, in order to preserve the value of our work!
And with it, we must not stop, even if it may be exhausting and tiring, it is about our very own interests!
Looks interesting, though there are two concern from the top of my head:
Would the âprotected artâ being accused as AI-generated by some people when they look into those invisible watermark?
It canât protect art being used for image-to-image output, which have already been a concern in some twitter anime artists. (Though this is another level of lazy-stealing, not really related to the auto-grabbing-machine)
EDIT: After one day I have some other thought to reply myself:
AI tool like Stable Diffusion can already generate image without invisible watermark by removing some parameter, which make the âwatermark detectiveâ confused.
Throwing otherâs art into image-to-image cannot be prevented by simply add a watermarkâŚunless those AI tool remove this function.
The invisible watermark main goal is trying to protect artist from the mass auto-grabbing machine, so these concern I wrote previously seems kinda out of scope
Iâm abit concerned about invisible watermarks. I donât see how that would help.
With all the social media mess lately, I have pulled all my artwork from deviant art, twitter and facebook. ( and an obscure old blog I did 15 years back Iâd nearly forgot about ) I currently am only posting links to some of my print on demand sales, which is a bit unfortunate as my old customers will probably only click the link to see what I have to offer and a new one, most likely wont bother.
Oh well, it currently is the mess of a world we live in at the moment.
I wish there was something secure, I know folks are migrating from twitter and facebook to other places, but mastadon seems a bit of a learning curve I dont want to mess with right now. Not sure of any others that are otherwise promising.
Discrete wavelet/cosine transforms only work on jpeg images. You can no longer upload pngs of your artwork if you use this.
StableDiffusion and Midjourney use different watermarking methods (DWT vs metadata). Itâd be easy to include âfake signedâ images by just checking for the existence of both.
AI training datasets can and do ignore AI watermarks if the site they scrape from does not allow AI artwork. This either means the generated image was truly indiscernible from a real artist, or the watermark is fake. Either of which theyâd like in their dataset to train from.
Which brings up the final point. If a generated image is shared online, it can be used to train another AI. Itâs not inbreeding, itâs natural selection. If you fed all outputs back into the dataset, itâd get worse. If you had humans cherrypick the best ones and fed those back in, then you improve the AI.
There is a lot of fearmongering surrounding AI, stay informed to temper your emotions and battle disinformation.
tempering emotions is a bad suggestion.
This deals with the very livelyhood of some of us. information is stolen, used, abused without permission to create something that the very artist could have done" but now doesnât get a chance to, and does not get paid for.
AI is criminal, and needs to be regulated.
It âcould " have been used as a tool if it would have remained offline and sold as a product to an artist, fed with that artists own artwork as a source, but the developers went in a"creepyâ way of stealing under the pretence of using a word like âtrainingâ. It is theft not training.
It violates copyright, that these deveolpers of dalle, midJourney etc call a grey area. Copyright needs to be upheld. The music indutry does not allow this, you go try to stream more than 30 secods of a song that sorta sounds like another song and you get hit with demonitization flagged etc⌠So why should we not be ANGRY, FURIOUS!!! and want proper justice.
Do not tell me ti is fearmongering when I have been battling severe depression and sorrow, and cannot even pick up a pen in 5 months to draw⌠Do not tell me to temper my emotions.
No. I will be angry, sad, sorrowfilled because this is the same stages of grief over death⌠What I wont do is become compliant, I will shout, I will voice it all as lud as I can, frankly because I do nto want this injustice to win.
