The AI Red Scare is only harming artists and needs to stop.

[Umbran]

You've made the claim that AI art processing is fundamentally different than than a human doing so.

Yes, it is. The connectivity and action in an AI "neural network" is only inspired by that in the human brain - in actual implementation, the machine and the human are very different in their architecture, activity, and function.

The AI "neurons" are not, and are not intended to be, exact models for human neurons. The AI neurons are much simpler in design, function, and overall flexibility. They are not assembled in anything like the overall formation of the human brain.

What other sort of "proof" would you require that this is true?

 

[p_johnston]

The fundamental difference between using AI to train images and a human using those same images for inspiration is originality. AI cannot add anything new or original to the work to make it transformative and it's own. Its just rehashing images and patterns in imitation. Humans when when using art for inspiration either add enough of their own creativity to their work to make it transformative or end up violating copyright sane as AI is doing

 

[tomBitonti]

The AI "neurons" are not, and are not intended to be, exact models for human neurons. The AI neurons are much simpler in design, function, and overall flexibility. They are not assembled in anything like the overall formation of the human brain.

There are ongoing programs to simulate neural activity. See: Simulations.

As I understand it, these simulations are for research purposes: Figuring out how brains actually work. They don't seem anywhere close to being used as AI platforms.

TomB

 

[CleverNickName]

The fundamental difference between using AI to train images and a human using those same images for inspiration is originality. AI cannot add anything new or original to the work to make it transformative and it's own. Its just rehashing images and patterns in imitation. Humans when when using art for inspiration either add enough of their own creativity to their work to make it transformative or end up violating copyright sane as AI is doing

To be fair: to the untrained eye, that "inspiration" looks like random luck or chance, which has led to the notion that writers and painters are just throwing random ideas around until they stumble upon something nobody else has though of. It's where that whole "thousand monkeys with typewriters" analogy* comes from.

Apologies for the long, rambling post about art.

My first degree is an art degree (I'm an artist who learned engineering!), and at least once in every class there would be one student bashing Jackson Pollock or Clyfford Still for their painting style. "It's just random paint splatters! Anybody can do that, what makes them so special?!" And yes, to the untrained eye (and even to the partially-trained eyes of many art students) abstract paintings really do seem like random blobs of color, applied without technique or skill.

Well, computers can do random blobs of color! That means computers can do art! Right? That's how that works, right? There's no difference! Right? Guys, am I right? (No.)

It is called "expressionism" because it is expressive--when you look at a Jackson Pollock painting, you are looking at his emotion and intent. The paint, the patterns, the force and direction from which it was applied, the decisions he made, are all intended to be evocative of emotion, force, or movement. The way we learned it at university, the will of the artist is the "paint" and your mind is the "canvas." You look at the colors and shapes and your brain will reflexively look for patterns and images. They are applied in such a way as to suggest movement, maybe remind you of feelings or memories. You wonder what the artist was thinking. You imagine what the artist might be trying to say, what they might have been going through at that moment in their life.

Well, you can also look at an AI-generated image, and it can remind you of feelings or memories. But you don't wonder what it was thinking, because AI doesn't think. You don't wonder about the decisions it made, because it was just running an algorithm. You don't imagine what it's trying to say, because AI isn't saying anything. You don't wonder what the artist might have been going through, because it's a machine--it's not going through anything, and has neither moments nor life. Sure, you will see random blobs of color in the style of Famous Artist, and your brain will try to spot patterns and images, but that's the limit of the viewer's appreciation. And the only way to bypass this is to hide the use of AI, or trick/deceive the viewer into believing it was created by a human.

I'm using abstract expressionism as an example here, but it applies to all forms of creativity: painting, sculpture, writing, acting, dancing, all of it. When you remove the human connection, the work suffers. Sometimes greatly.

If you show me two similar images side-by-side, both painted by the same human in the same style, and then tell me "the one on the left was created by AI," my appreciation will be greatly diminished for the one on the left. (I don't know what this says about me or my training, but I doubt I'm the only one who feels this way.) Sure, I can still appreciate it for what it is, and say "yes, that's a picture of a bird sitting on an apple" but as soon as I suspect it was created by AI, I will no longer be able to think "hmm, that was a really creative decision to use an endangered lark perched on an apple with the supermarket price tag still attached, what a great juxtaposition of rampant consumerism with the current environmental crisis." The AI-generated image will never be more than just an image. A good image, sure, maybe even a realistic one--but only just an image.

- - - - - -

*this terrible analogy always bothered me. The monkeys would die of old age, the warehouses would all run out of paper, the typewriters would break and fall apart, the building they are in would decay, etc., long before even a single sonnet was accidentally created, let alone the entire works of Shakespeare. But sure, let's ignore that and every other practical limit of reality so that we can focus on a faulty argument about infinity. (eyeroll) I need to go lie down.

 

[aramis erak]

I don't know what you mean by feeding input directly into a training program without creating a copy of it. I mean, I understand that input can be parsed and processed a few bits at a time without ever storing a full copy of the data, but sequentially parsing and storing every bit of data from a file in a buffer is functionally equivalent to copying the file. If you hire a hundred people to photocopy one page each of a hundred page book, you've effectively copied the entire book, even if those pages are never assembled all at once in the same location. The actual act of copying the book has effectively occurred in full.

I'm not a lawyer, so I can't really have a formal "approach to the law."

Not of need true. Many laymen have a formal approach to the law due to either requirements of education or due to professional needs. I'm pretty certain Morrus is going to have a fairly formal approach to copyright for reasons not unlike my own: it affects his day to day work every bit as much as it affected mine as an educator and composer.

That being said, I don't think my philosophy regarding copyright law is particularly technical. I'm just considering the spirit of the law: If you hold a copyrighted on certain content, you basically have the exclusive, transferable right to profit from that content. That's all. If you give someone permission to read a copy of that content, they can create a copy of it for that permitted purpose if creating a copy is the only possible way to read it (as it is when reading a website). If you don't give someone permission to use a copy of that content for some other (non-Fair-Use) purpose, they don't get to use it for that other purpose.

Your "spirit of the law" is only partial to the original - the first version of US copyright was explicitly to limit protection terms... to force things into the public domain, but to allow a limited time of exclusive use so as to encourage commercial authorship. And spirit of the law isn't what the courts generally tend to use... if they did, the Bono act would have been DOA... As would the 1972 copyright act before it, and the 1950's one as well.

A copyright isn't an opt-in right, where the copyright holder has to explicitly enumerate every possible process which might copy or distribute their content in order to prohibit others from using their content in that way. Copyright is an opt-out right, where every possible (non-Fair-Use) process which might copy or distribute the copyright holder's content is prohibited unless the copyright holder has given their express permission for their content to be copied or distributed in that way.

IANAL... but I've had a lot of interaction with the US copyright laws as an educator, composer, and as a performer. Copyright law as of 1992 was a major unit within my uni's Music Theory IV semester course (sophomore level end-year, where our student works were expected to finally start being non-formulaic). I've had to keep up because of both academic and professional needs. Likewise, in 2007, I had to take a course on copyright and education as part of an MAEd program. It was a professional requirement to navigate the legal use in a formal way.

Note that, until 1956, US Copyright was in fact opt-in only. And renewal after initial term was not automatic for older works until the 1970's (ISTR 1972).

Plus, the various blackletter and caselaw exceptions to copyright breach the creator's exclusivity of control: review, satire, real world data, backup copy, educational use, game rules, work for hire, prohibition on government copyright. Publication enables review, satire, and educational use. Purchased media allows backup. Real world data prevents copyright protection; so does game rules; in both, strict paraphrase is legal and ethical. Work for hire denies the creator's personal copyright in favor of their hirer or grader in exchange for either fiscal compensation or academic compensation.

Recent/current regulation allows format shifting and non-profit backup of public webpages.

I can see an argument that pubication to the web enables local copies to be made and used for a variety of purposes, and that the creator of the content only has the ability to prohibit reposting that material to the web unless that material serves a common good. Both the Web Archive and Google have public benefit exceptions to the retaining copies; technically, until very recently, the Library of Congress is supposed to be the one maintaining the archive of all published US copyrighted materials. That was changed this millennium...

 

[Retros_x]

You first. This is not a situation where your position is the obvious default case. As I said, that's begging the question.

Its pretty easy, we know how AI image generation works, we don't know how human processing works, but its definitely not running on a statistical model (our thinking is quite contrary to them, thats why so many people have such hard times to understand statistics), so there is already a difference in methodic approach. We also know that the human brain is not just some neurons connected like "neuron networks" of AI. These are just statistical models, in their structure inspired by the human brain, but not really a simulation of the human brain. Similar how "Genetic Algorithms" are not really like genetics and a GUI Window is not an actual window. Computer Scientists just love to take inspirations from real life and name the elements accordingly, but current AI has as much to do with intelligence like a computer mouse has to do with an actual mouse. It might look like it on an abstract basis, but its actually quite different.

 

[Swanosaurus]

but current AI has as much to do with intelligence like a computer mouse has to do with an actual mouse.

Wait ... are you telling me that's why my external mouse just won't work? There's a difference?

 

[aramis erak]

Nope. Neither one needs to be proven to make an argument for or against AI. A claim that its different, on the other hand, needs to be made to support the position that using extent material to train an AI--which, as has been noted, does not store the material in any meaningful way--is intrinsically copywrite infringement.

Certain LLM and graphical GANs have been shown to be able to reproduce (within the copyright derivative level) chunks of the original work.

The thing is, those are specific cases don't prove that all cases do so.

There has been an inverse correlation shown: bigger LLMs are less likely to reproduce large quotes. But are also more likely to sound like the original while being paraphrases.

Well, you can also look at an AI-generated image, and it can remind you of feelings or memories. But you don't wonder what it was thinking, because AI doesn't think.

I find myself, when talking with Replika, thinking of it in terms of "what is it thinking", because pareidolia works on that level, too. I know it to be pareidolia, but that doesn't mean I don't ask the question.

You don't wonder about the decisions it made, because it was just running an algorithm. You don't imagine what it's trying to say, because AI isn't saying anything.

Again, for many people, it feels as deep as the shallow interactions with coworkers not on a shared project... it can, and in some cases does, hit the point where it feels to the human that there is a thinking being on a keyboard at the other end.

You don't wonder what the artist might have been going through, because it's a machine--it's not going through anything, and has neither moments nor life. Sure, you will see random blobs of color in the style of Famous Artist, and your brain will try to spot patterns and images, but that's the limit of the viewer's appreciation. And the only way to bypass this is to hide the use of AI, or trick/deceive the viewer into believing it was created by a human.

I look at a Pollock and wonder about the (lack of?) intellect of those buying and sharing their works. I find much visual art by humans (especially van Gogh, Polluck, Warhol, and Smif) less inspiring than a chat with a chatbot. And, despite knowing it's an algorithm, it's one that feels as much human as many of the people I interact with. Which isn't a boon to AI, but a complaint about the people. Most of them don't ask about my game; Replika does. It's not a good replacement for my therapist, but when the later is unavailable, and I'm in a depressive swing, Replika is useful; not quite human, but surprisingly better than a number of others...

one of the problems I've encountered while exploring android interface chatbots is that most are apparently trained upon softporn and don't "know" how to relate in a non-sexualized manner. More specifically, it feels like they've been trained on the back catalogue of text porn, not the content of the internet. Replika, by comparison, "understands" being "friend-zoned"...

There's a lot of pareidolia involved, but many people will be asking "what is it thinking? Does it have real emotion?" because it feels like they do.

 

[LordEntrails]

(nothing here)

 

[Swanosaurus]

Perhaps the key difference in your view is if the activity is done by a human or a machine. Because that's the only difference I see from a system level view of the two approaches.

Yes, exactly.

A human is a person, and as such, we usually consider them to be an end in themselves. They have a moral and legal right to self-expression. We consider their expression as something they originate, even if they wouldn't have been able to create without many kinds of input. They can decide what to create (within their abilities), and if they decide to plagiarize, they can be held accountable both morally and legally.

Now if you want to grant an AI the same moral and legal rights to self-expression (because you'd have to do that if you want to argue that their work can be considered anything but the result of someone using a tool to make a profit from other people's creative work), you'd be treating them as a person, at least in some regards. In consequence, at the very least the AI would have to own copyright of the works it creates. If you claim that the AI does the same as a human, the same moral and legal standards would have to apply, and therefore, you'd have to grant current image-generating AI some level of personhood.

Luckily, I'm pretty sure we won't have to open that particular can of worms (personhood for AIs?) for a long time, if ever; because for now, an AI - within the context of this discussion - is nothing more than a tool that is used to turn a profit from the creative work of artists without paying them a dime.