It's not about "analysis" but about for-profit use. Public domain still falls under Fair Use. I think you're being too optimistic about support for UBI, but I absolutely agree on that point. There are countries that believe UBI will be necessary in a decades time due to more and more of the population becoming permanently unemployed by jobs being replaced. I say myself that I don't think anybody would really care if their livelihoods weren't at stake (except for dealing with the people who look down on artists and say that writing prompts makes them just as good as if not better than artists). As it stands, artists are already forming their own walled off communities to isolate their work from being publicly available and creating software to poison LLMs. So either art becomes largely inaccessible to the public, or some form of horrible copyright action is taken because those are the only options available to artists.

Ultimately, I'd like a licensing system put in place, like for open source software where people can license their works and companies have to cite their sources for their training data. Academics have to cite their sources for research, and holding for-profit companies to the same standards seems like it would be a step in the right direction. Simply require your data scraper to keep track of where it got its data from in a publicly available list. That way, if they've used stuff that they legally shouldn't, it can be proven.

[–] ClamDrinker@lemmy.world 1 points 1 month ago (1 children)

If you think I'm being optimistic about UBI, I can only question how optimistic you are about your own position receiving wide spread support. So far not even most artists stand behind anti AI standpoints, just a very vocal minority and their supporters who even threaten and bully other artists that don't support their views.

It’s not about “analysis” but about for-profit use. Public domain still falls under Fair Use.

I really don't know what you're trying to say here. Public domain is free of any copyright, so you don't need a fair use exemption to use it at all. And for-profit use is not a factor for whether analysis is allowed or not. And if it was, again, it would stagnate the ability for society to invent and advance, since most frequent use is for profit. But even if it wasn't, one company can produce the dataset or the model as a non-profit, and the other company could use that for profit. It doesn't hold up.

As it stands, artists are already forming their own walled off communities to isolate their work from being publicly available

If you want to avoid being trained on by AI, that's a pretty good way to do it yes. It can also be combined with payment. So if that helps artists, I'm all for it. But I have yet to hear any of that from the artists I know, nor seen a single practical example of it that wasn't already explicitly private (eg. commissions or a patreon). Most artists make their work to be seen, and that has always meant accepting that someone might take your work and be inspired by it. My ideas have been stolen blatantly, and I cannot do a thing about it. That is the compromise we make between creative freedom and ownership, since the alternative would be disastrous. Even if people pay for access, once they've done so they can still analyze and learn from it. But yes, if you don't want your ideas to be copied, never sharing it is a sure way to do that, but that is antithetical to why most people make art to begin with.

creating software to poison LLMs.

These tools are horribly ineffective though. They waste artists time and/or degrade the artwork to the point humans don't enjoy it either. It's an artists right to use it though, but it's essentially snake oil that plays on these artists fears of AI. But that's a whole other discussion.

So either art becomes largely inaccessible to the public, or some form of horrible copyright action is taken because those are the only options available to artists.

I really think you are being unrealistic and hyperbolic here. Neither of these have happened nor have much of chance of happening. There are billions of people producing works that could be considered art and with making art comes the desire to share it. Sure there might only be millions that make great art, but if they would mobilize together that would be world news, if a workers strike in Hollywood can do that for a significantly smaller amount of artists.

Ultimately, I’d like a licensing system put in place Academics have to cite their sources for research That way, if they’ve used stuff that they legally shouldn’t, it can be proven.

The reason we have sources in research is not for licensing purposes. It is to support legitimacy, to build upon the work of the other. I wouldn't be against sourcing, but it is a moot point because companies that make AI models don't typically throw their dataset out there. So these datasets might very well be sourced. One well known public dataset LAION 5b, does source URLs. But again, because analysis can be performed freely, this is not a requirement.

Creating a requirement to license data for analysis is what you are arguing here for. I can already hear every large corporation salivating in the back at the idea of that. Every creator in existence would have to pay license to some big company because they interacted with their works at some point in their life and something they made looked somewhat similar. And copyright is already far more of a tool for big corporations, not small creators. This is a dystopian future to desire.

[–] EldritchFeminity@lemmy.blahaj.zone 1 points 1 month ago (1 children)

We're already living in a dystopia. Companies are selling your work to be used in training sets already. Every social media company that I'm aware of has already tried it at least once, and most are actively doing it. Though that's not why we live in a dystopia, it's just one more piece on the pile.

When I say licensing, I'm not talking about licensing fees like social media companies are already taking in, I'm talking about open source software style licensing - groups of predefined rules that artists can apply to their work that AI companies must abide by if they want to use their work. Under these licensing rules are everything from "do whatever you want with my code" to "my code can only be used to make not-for-profit software," and all derivative works have the same license applied to them. Obviously, the closed source alternative doesn't apply here - the d'jinn's already out of the bottle and as you said, once your work is out there, there's always the risk somebody is going to steal it.

I'm not against AI, I'm simply against corporations being left unregulated to do whatever the hell they want. That's one of the reasons to make the distinction between people taking inspiration from a work and a LLM being trained off of analysing that work as part of its data set. Profit-motivation is largely antithetical to progress. Companies hate taking risks. Even at the height of corporate research spending, the so-called Blue Skies Research, the majority of research funding was done by the government. Today, medical research is done at colleges and universities on government dollars, with companies coming in afterward to patent a product out of the research when there is no longer any risk. This is how AI companies currently work. Letting people like you and me do all the work and then swooping in to take that and turn it into a multi-billion dollar profit. The work that made the COVID vaccines possible was done decades before, but no company could figure out how to make a profit off of it until COVID happened, so nothing was ever done with it.

As for walled off communities of artists, you should check out Cara, a new social media platform that's a mix of Artstation and Instagam and 100% anti-AI. I forget the details, but AI art is banned on the site, and I believe they have Nightshade or something built in. I believe that when it was first announced, they had something like 200,000 people create accounts in the first 3 months.

People aren't anti-AI. They're anti late-stage capitalism. And with what little power they have, they'd rather poison the well or watch it all burn than be trampled on any further.

[–] ClamDrinker@lemmy.world 1 points 1 month ago* (last edited 1 month ago)

You don't solve a dystopia by adding more dystopian elements. Yes, some companies are scum and they should be rightfully targeted and taken down. But the way you do that is by targeting those scummy companies specifically, and creatives aren't the only industry suffering from them. There are broad spectrum legislatures to do so, such as income based equality (proportional taxing and fining), or further regulations. But you don't do that by changing fundamental rights every artists so far has enjoyed to learn their craft, but also made society what it is today. Your idea would KILL any scientific progress because all of it depends on either for profit businesses (Not per se the scummy ones) and the freedom to analyze works without a license (Something you seem to want to get rid of), in which the vast majority is computer driven. You are arguing in favor of taking a shot to the foot if it means "owning the ~~libs~~ big companies" when there are clearly better solutions, and guess what, we already have pretty bad luck getting those things passed as is.

And you think most artists and creatives don't see this? Most of us are honest about the fact of how we got to where we are, because we've learned how to create and grow our skill set this same way. By consuming (and so, analyzing) a lot of media, and looking a whole lot at other people making things. There's a reason "good artists copy, great artists steal" is such a known line, and I'd argue against it because I feel it frames even something like taking inspiration as theft, but it's the same argument people are making in reverse for AI.

But this whole conversation shouldn't be about the big companies, but about the small ones. If you're not in the industry you might just not know that AI is everywhere in small companies too. And they're not using the big companies if they can help it. There's open source AI that's free to download and use, that holds true to open information that everyone can benefit from. By pretending they don't exist and proposing an unreasonable ban on the means, denies those without the capital and ability to build their own (licensed) datasets in the future, while those with the means have no problem and can even leverage their own licenses far more efficiently than any small company or individuals could. And if AI does get too good to ignore, there will be the artists that learned how to use AI, forced to work for corporations, and the ones that don't and can't compete. So far it's only been optional since using AI well is actually quite hard, and only dumb CEOs would put any trust in it replacing a human. But it will speed up your workflow, and make certain tasks faster, but it doesn't replace it in large pieces unless you're really just making the most generic stuff ever for a living, like marketing material.

Never heard of Cara. I don't doubt it exists somewhere, but I'm wholly uninterested in it or putting any work I make there. I will fight tooth and nail for what I made to be mine and allowing me to profit off it, but I'm not going to argue and promote for taking away the freedom that allowed me to become who I am from others, and the freedom of people to make art in any way they like. The freedom of expression is sacred to me. I will support other more broad appealing and far more likely to succeed alternatives that will put these companies in their place, and anything sensible that doesn't also cause casualties elsewhere. But I'm not going to be in favor of being the "freedom of expression police" against my colleagues, and friends, or anyone for that matter, on what tools they can or cannot not use to funnel their creativity into. This is a downright insidious mentality in my eyes, and so far most people I've had a good talk about AI with have shared that distaste, while agreeing to it being abused by big companies.

Again, they can use whatever they want, but Nightshade (And Glaze) are not proven to be effective, in case you didn't know. They rely on misunderstandings, and hypothetically only work under extremely favorable situations, and assume the people collecting the dataset are really, really dumb. That's why I call it snake oil. It's not just me saying exactly this.