Start a new topic

Ban on Ai Art

The current trend of Lensa, Midjourney, etc is rent seeking ghoulish behavior that seeks to destabilize  artistic labor even more than it already is. 

This is NFT shit all over again and theres needs to be a zero tolerance of it. This is automated theft and exploitation to be stopped.

6 people like this idea

Angry name-calling and asking for extremely broad bans is not a great way to make this point when you will see that many people who are exploring this field on Cohost are engaging directly with conversations of ethics and societal impact, and that many engage with those new systems for exploration rather than for monetary gain.

8 people like this

I'm not interested in a website that bans an entire medium of artistic expression in its infancy. I trust that Cohost staff can ban bad actors and remove instances of copyright infringement and other TOS violations without a blanket ban on generative art. So far, I have been impressed by Cohost's thoughtful and principled approach to their TOS and community guidelines, and I hope they will continue approaching hot-button issues by continuing to improve users' abilities to control their own experiences, rather than amending site policy in reaction to each new controversy. 

8 people like this
I mean, I'm uncomfortable with the many ethical issues with AI art too, but I *really* don't think a ban is the right call here. If anything, all it'll do is cause people to continue posting it without tagging it as AI-generated, and spark witch-hunts against artists for having work that's suspected of being AI-generated. Which you can argue already happens, but bans just encourage that sort of thing even more.

6 people like this

An expedition on muting/blocking tags would help this, along with the ability to block an entire account’s pages as currently people can circumvent blocks with alt-accounts.

Banning the entire topic is a bit heavy, articles/discussions and conversation surrounding the topic, absolutely, should be allowed.

It’s a very contentious issue for artists and they don’t feel safe with their work hosted alongside generated images which they know are likely scraped from sites without consent.

Asking for civility over an issue where the first civil step would be to ask artists if they wanted to contribute is kind of a big ask, especially as everyday after I see fellow artists getting their works spitefully taken and abused when they dare to ask, after the theft already occured, for some understanding and empathy on the issue.

It’s easy to feel it’s disrespectful seeing AI images posted knowing artists are in a living nightmare right now.

Ban AI images? Artists will sing to the high heavens and happily post without dreadding as much about cohost brushing shoulders with prompters who might feel bold enough to just scrape the art tag. (Do note, banning ai prompters won't really prevent this from happening, it's more a principled stance for the staff to take to mildly comfort artists, who will have to deal with this everywhere else until legislation gets involved).

Don't ban it? Artists will feel like the owners are taking the side of the failed-nft-grifters which is an unfair reach, but for artists who deal with being undervaluved enough to get their freely posted work stolen, and only valued when someone else profits off their skills and labour, it's tiring and artists will pack their stuff and hike it. (Again this wouldn't save them from getting their work stolen but some artists will just up and disappear and never share their work again, or draw again.)

Personally, give me the ability to block an account and their subsequent pages, after logged-in posts gets enabled, I'd feel more comfortable posting.

If staff did ban such posts, I’d feel a lot more relief even if again it’s merely a principled stance, you can’t stop vindictive or ignorant people from taking something that isn’t theirs because they’re having fun at the expense of others, it would show staff are not stepping over the overwhelming response of artists decrying this to have the discussion with tech/legal folk if this is morally/legally ok to host and (without intending to) encourage this behaviour.

6 people like this
It absolutely sucks to be afraid to post your work. Nobody should be afraid to post on Cohost, and nobody should act in the way of those bad actors act on Cohost. But who has acted this way on Cohost?

Sure, terrible jerks are involved in AI image generation on Reddit, Twitter, and other sites. But where are they on Cohost? If they are all "failed NFT grifters", among the 20 people or so who posted in the stable-diffusion tag, how many have an Opensea profile, versus how many are on the record being very negative about cryptocurrencies?

You want things tagged? Sure, I've taken to using the #AI Generated tag due to it being neutral and unambiguous, does it work for you? It's also disclosed in the description, sometimes with a discussion of the process or of an interesting quirk of the algorithm. For pictures, it's also embedded in the metadata (Cohost strips it for privacy, but I don't strip it myself). There is also an invisible watermark added to every picture to ensure they can be detected easily. I think blocking tags is on Cohost's todo-list, for now you can mute or block entire pages.

But you want such tagging enforced? Are you ready for witch hunts? Are you ready to relitigate every fight about tracing 3D references and using photobashing? Do you want people taking the zoom and eyedropper tool to your art to send reports if your shading technique is a bit sussy? Do you want to end up forced to clear up your name of malicious accusations because you use a tool such as Kirta, which has a SD inpainting plugin, by having to provide a screencast of your painting process? Do you want to waste the time of a small staff by forcing them to police whether art is contaminated?

And why are you casting your net only over images? Code, text, audio, 3D models are all concerned by this field. Will you ask Staff members to take down their own posts using ChatGPT?

Additionally, you are asking Staff to adopt the "collage engine" interpretation of what current AI image generation techniques do. This interpretation is extremely contentious, as it is easily argued against on technical grounds. Moreover, there are literal collage artists on Cohost, and hip-hop producers who are likely to use sampling. Do you want to relitigate the legitimacy of their techniques as a side-effect?

If you wish to argue guilt by association, you have to demonstrate that association exists. People can't just say AI bros are NFT ghouls then fail to point out bad behavior. It's just going after targets of opportunity because they are in your immediate line of sight. And on Cohost, targets of opportunity are mostly LGBT weirdos into weird outsider art engaging with the technology critically, and interrogating it from an artistic background.

I am primarily a musician, machine learning being a natural extension of exploring stochastic music, making my own tools to explore chaotic deterministic systems and aleatoric systems, generative themes such as fractals and markov chains, euclidian rhythms, machine learning based lookup tables, how to use those techniques for live modular performance, and how those aleatoric techniques can be hybridized with a more compositional approach. It is very unlikely that Staff has the skills to figure out what any of this means, and to decide whether or not some of my live performance "use AI". They would have to make difficult pronouncements such as "does an untrained neural network constitute AI".

Most AI users you find on Cohost do not fit the mold of the NFT bro who harasses artists on Twitter.
You may be able to chase away from Cohost a few well-behaved people who have fun playing with Stable Diffusion, but you'll never take this tech away from the thousands of 4chan guys who use it to make mediocre porn, from the brands that use it to make even more mediocre clipart, and the millions who use apps that make it a one-click photo filter.

Additionally, asking for the ability to block every single page by the same account would necessarily leak their identity, no matter how you go about the implementation. Shouldn't Cohost be a place where questioning people are safe to create an alt page with different pronouns to see how they feel about it, without fear of being outed?

6 people like this

Wanting compensation for your labor does not makes you a landlord.  Landlords to not contribute anything and deny people something that is required to live.  Art is a luxury good and anyone can learn to be an artists.

Not only that but, these AI models are imperfect and are improving by having people, like you, use them and test them without compensation to you. You are being exploited into free labor by the people who own the AI models. By continuing to use them and allow their use you are aiding these companies who, at any time, can make it so these programs are no longer free.

You are helping the landlords build the house.

5 people like this
The idea that you think of art as only a physical resource and not a representation of the time, skill, and labor that went into making a piece is the issue.

5 people like this

The conversation in this thread has ceased to be useful or constructive. I'm locking this for now. We do not have any pending community guidelines changes re: AI Art to announce at this time.

4 people like this
I also support a ban of anything generated on any model trained on unlicensed data acquired without consent. If a generative program isn't built off data scraped from thousands of non-consenting creators then it would theoretically not be a problem, but as far as I know that doesn't exist.

4 people like this
It's pretty easy actually to ultimately be against copyright law and still think that tech startups bursting into a room, vacuuming up a bunch of stuff they had nothing to do with so they can put it into a blender and resell it to people (and eventually other corporations) in the hopes of replacing the people who made the original stuff sucks a big one.

4 people like this

Putting words in people's mouth by using imaginary quotes conceding your absurd premises makes it clear that the threat you are upset with exists only in your imagination, Kaden. 

The best part of Cohost are its extensive controls to curate your experience, and a userbase eager to accommodate people who require material that upsets them tagged. You already have all the tools you need to curate your dashboard and limit your exposure to this material.

3 people like this
I told myself I wouldn't jump into this argument, but eh.

I personally believe that there's little value that could be added in repeatedly attempting to make this entire debate about one single argument based purely on shock value. It makes the arguments against AI art look simpler than they actually are, and makes the actual hard-hitting arguments take a back seat. Not to mention the article itself that this info came from (which I definitely think hasn't helped here with its clickbait title) also goes into detail on the ways they're already trying to combat stuff like this - how they're trying to completely filter out extreme materials like that.

Finally, I personally believe it's not even relevant unless you're actively trying to produce materials like that, in which case, you're a sicko (derogatory). If you tell it to generate stuff like that, I feel the fault lies more with you than whoever trained the model - which also included things like safety filters that you likely had to disable to make it output such things. Depending on what horrible things you chose to create, you might even be violating the license you had to agree to before downloading the model. Regardless, I don't exactly think any good model is referring to revenge porn or ISIS beheadings when you tell it to create a picture of a cute kitten.

I think the biggest argument against AI art has to do with its capacity to displace artists, and the impact it will have on artists looking to make a living if its existence is normalized. This is a problem that can't be solved by a model from someone who actually completely filters out the bad material - it has to be solved outside of that.

I like to think that moves that places such as the US Copyright Office are making helps (in the US, AI-generated art can't be copyrighted as it lacks "human authorship", and multiple AI-generated works have already been rejected for copyright due to that reason), but not all governments are the same in that manner, unfortunately.

My opinions on this topic are a bit mixed, admittedly. I think, in some regards, there has been a Pandora's Box that has been opened - AI art won't cease to exist, even if it were completely and totally outlawed. I think the ideal solution would be making it completely unviable to use in any situation that involves the exchange of money - but that involves some massive systemic change, and I'm sure companies that want to profit off of these models won't like that.

As far as if it should be banned from Cohost, though? I think that'd cause some problems - mainly, how can you tell what art is AI-produced and what isn't? Sure, some models invisibly watermark their output, but then there's folks who will remove those functions, or will destroy the watermark in some way through manipulating the image. Even worse, some folks know the watermarking is so that the generated works won't be included in newer datasets, so there might be a few artists that deliberately watermark their own works so they won't get trained on.

Sure, some images might be more obviously AI-generated by just how badly they fudge fingers or other artifacts, but not every AI-generated image is like that, and what if the poster made a very minor edit to fix such details? Even worse - what if someone's art style involves some level of simplification of features that looks similar *enough* to how AI fudges those features?

I think a blanket ban on AI art will just lead to people obscuring the fact they generated their works through a neural network, and will lead to people trying to witch hunt folks that they *think* are making AI generated art - even if they aren't using neural networks in their art. I think it should be a situation where people are encouraged to tag AI-generated works as such, so that tag muffling can do its job.

If there was some way to identify, with perfect accuracy, which images are and aren't AI generated, I think a ban would be a better idea. As is right now, however, I feel like it'd just lead to more toxicity, with AI art wordsmiths trying to obscure the fact their works are generated by neural networks, and some folks choosing to try and hunt those people down and perhaps not always coming up with the right answer.

3 people like this
AI generated images need to be banned completely. Ai has been trained on revenge porn, executions, and stolen ip. The use of AI is completely unethical in its current form.

3 people like this
Yeah this feels like working backwards to justify why it's actually fine the shiny new toy you want to play with was built with stolen labor.

3 people like this
Yeah, not just stolen ip but also revenge porn, stolen medical photos, and isis executions. If you want to build an ai off of completely public domain images and works that the authors themselves submit. Share fine whatever. But there is no current model that does that. This person is just clearly a troll hiding behind pseudo Marxists language and has never talked to an actual artist ever.

3 people like this