AI discussion (split from Canva thread)

[Bartek]

8 minutes ago, PaulEC said:

It makes a difference to the person who owns, and chooses to sell, the images!

Is there any obligation for graphic designers to provide the earnings of photographers? It should all be a matter of my choice whether I spend the time/electricity and use AI to generate the wolf image or spend the time/skill and go into the forest and photograph the wolf or spend the money and buy the wolf image from stock services (if I happen to need a wolf image).

[PaulEC]

Obviously it's up to you where you choose to get the images that you use. Personally, I would never employ a graphic designer, I would just use AI. (or, maybe use Canva!)  😁 
 

[Bartek]

1 minute ago, PaulEC said:

Obviously it's up to you where you choose to get the images that you use. Personally, I would never employ a graphic designer, I would just use AI. (or, maybe use Canva!)  😁 
 

Think about all those unemployed chimney sweepers who lost their jobs because you chose electric heating instead of using a fireplace.  

Once upon a time, there were attempts to destroy mechanical looms because they took away the work of weavers.

Signs of the times. The question is whether you want to fight it.

[William Overington]

Many and varied are the reasons why various users of Affinity software use Affinity software.

Some people use Affinity products to seek to earn their living by producing work for clients, some people are amateur artists, and so on.

It seems clear to me that it is not a one size fits all situation.

How Affinity moves forward in relation to AI needs to ensure that each end user can achieve good results according to that end user's chosen needs. This could be anything from no AI whatsoever and provably so to customers, to delighting in AI generated images that closely follow the text prompt input by the end user. Bing Chat AI currently outputs AI generated images only as 1024 pixel by 1024 pixel images. If Affinity incorporates AI in some Affinity products, will Affinity buy-in an AI system or will Affinity develop its own system?

When I have Bing Chat AI output an image, I do not claim to be the artist. I quote the prompt that I used and I write that I authored the prompt.

William

 

[walt.farrell]

1 hour ago, Bartek said:

Is there any obligation for graphic designers to provide the earnings of photographers? It should all be a matter of my choice whether I spend the time/electricity and use AI to generate the wolf image or spend the time/skill and go into the forest and photograph the wolf or spend the money and buy the wolf image from stock services (if I happen to need a wolf image).

In the case of AI, the question is whether the developers of the AI had the permission of the artists (photographers) whose work they used to train the AI, and whether the artists were properly compensated for the use of their images for that training. 

You are of course free to take your own photographs in the forest. And you're free to use a Stock agency, who will have appropriately compensated the artists whose work they carry.

The source of the AI training material, and the compensation for using it, are less clear, and that is what is causing the ethical issues from what I've read.

[William Overington]

When I was young, steam locomotives were mainstream and metal type printing was mainstream too, now both are mostly gone, scrapped, but survive in a much limited extent in heritage railways and traditional letterpress printing activity.

Look how Unicode has changed since emoji became allowed to be encoded. Yet the introduction and popularity of emoji, encoded in plane 1, and thus their support led to better support for rarer scripts in plane 1 as a good side effect.

So where will whatever engagement that Affinity has with AI lead?

William

 

[R C-R]

1 hour ago, walt.farrell said:

In the case of AI, the question is whether the developers of the AI had the permission of the artists (photographers) whose work they used to train the AI, and whether the artists were properly compensated for the use of their images for that training. 

Hypothetically, if it were humans who wanted to train themselves to create better work & they did so by studying the work of other artists, should the studied artists get compensated for that, & if so, how?

I think this is just one of the several issues that complicate if or how AI generated content can be regulated.

[R C-R]

5 hours ago, RichardMH said:

If Photo goes the generative AI route, then we need them to join the content authenticity initiative and editing adjustments be in the meta data. I've been led to believe Photoshop has it already.

How would that help for printed material that is fully or partially created with the help of generative AI that someone then sells?

[walt.farrell]

13 minutes ago, R C-R said:

Hypothetically, if it were humans who wanted to train themselves to create better work & they did so by studying the work of other artists, should the studied artists get compensated for that, & if so, how?

It is well accepted, I think, that humans can study the work of other humans and learn from it and improve their own work. Even there, though, if you study the work of a living artist and produce something too similar to one of the existing works, you may be sued for copyright infringement.

The difference with AI, I think, is more a matter of scale, and of not knowing exactly how the source material is being copied or modified.

[gguillotte]

22 hours ago, William Overington said:

Do you know why that is please? Is it anything to do with concern over intellectual property rights in case what is provided by AI is an unauthorized copy of someone else's work?

Per statements by companies:

"Our customers expect a human touch to our releases, and so long as the ethical and legal circumstances surrounding these programs remains murky and undefined, we are unwilling to associate our brands with the technology in any way."

or:

"Our internal guidelines remain the same with regard to artificial intelligence tools: We require artists, writers, and creatives contributing to the Magic TCG to refrain from using AI generative tools to create final Magic products. We work with some of the most talented artists and creatives in the world, and we believe those people are what makes Magic great."

and, after being confronted by customers for using generative features in marketing images:

"Now we’re evaluating how we work with vendors on creative beyond our products – like these marketing images – to make sure that we are living up to those values.”

and in marketplace advisement, such as:

"AI art, AI writing, or algorithm-based creation is prohibited on the Infinite Platform. Any product posted to Infinite that contains AI-generated art or writing where very little human work was done is subject to removal."

In practice, this has included artworks that were primarily human-generated but used tools like generative fills, and texts where creators admitted that they used generative AI to supplement, edit, or restructure their own writing.

In chat rooms and on social media where creators in this field congregate, they raise products to each other that they suspect contain generative AI, investigate them together, and collectively report them to marketplaces until the products are removed.

[R C-R]

2 hours ago, walt.farrell said:

It is well accepted, I think, that humans can study the work of other humans and learn from it and improve their own work. Even there, though, if you study the work of a living artist and produce something too similar to one of the existing works, you may be sued for copyright infringement.

The difference with AI, I think, is more a matter of scale, and of not knowing exactly how the source material is being copied or modified.

But for generative AI, is there really any actual copying or modifying of existing work occurring? At worst, isn't is more nearly like 'in the style of' some existing work or artist?

These are not rhetorical questions. I really do not know enough about it to say how it could or should be considered, ethically, legally or otherwise.

Also, on copyright infringement, it is up to the copyright owner to actively protect & preserve his or her usage rights, suing the infringer if necessary to do so. When AI generated artwork is involved, it may be very hard to determine who the actual infringer is, who should be sued, & if there is much chance of getting a favorable settlement.

So regarding scale, unless somehow the courts decide the infringer is the maker of the AI-powered tool or compiler of the source material used to train it, it may be futile to try to sue what could be dozens or hundreds of individual people who used the tools for any purpose that infringed on their rights.

I do not think there will be any resolution of these or the other vexing issues surrounding the use of AI tools any time soon.

[R C-R]

I'm not sure how much to trust Adobe about it but they claim their Firefly generative AI technology was trained only on the several hundred million images they own & from other sources that are not encumbered by license restrictions that otherwise would prevent their use for that purpose.

I have to wonder if what if anything Canva will offer, either as something built into a future version of Affinity or as a subscription based add-on, will make the same claim & how much we could trust it....

[William Overington]

3 hours ago, R C-R said:

But for generative AI, is there really any actual copying or modifying of existing work occurring? At worst, isn't is more nearly like 'in the style of' some existing work or artist?

You might like the following post and the one following it.

The modalities of using Bing Chat AI to produce art (Page 2) — Art & Literature — Alfred's Serif Users' Forums (punster.me)

In the United Kingdom, copyright continues for seventy years after the end of the lifetime where the author is an individual human.

William

 

 

[R C-R]

2 hours ago, William Overington said:

You might like the following post and the one following it.

I'm not sure how relevant that is to what I said since it is about AI-generated work created in the style of artists, not actually copying or modifying their work.

[debraspicher]

AI takes other people's work, throws them into a wood chipper where it then applies it own signature to analyzed bits, then uses its unique algorithmic magic to interpret those works and thus their bits are used to "reform" new works based on previous works. It's a form of theft in the sense it takes other people's materials and uses it to feed the process... which it cannot do on its own without those references.

The copying is merely the act of feeding something into the algorithm. It has less to do with the output, more to do with the fact that many creators do not approve their works being used to train computers against their consent. I get the feeling for some it is akin to identity theft, which I really can't blame them.
 

That said, it's not illegal to trawl the public web and store people's information against their consent, so nothing will be done. People who hate to pay for things will welcome AI. Many people still pirate music. Just like many don't have any issue with search engines and social media sites mass collecting our information, they will be fine with it because it provides a free service in return.

[SrPx]

12 hours ago, R C-R said:

How would that help for printed material that is fully or partially created with the help of generative AI that someone then sells?

A, still not fair, yet "not so bad" solution would be to make it a requirement by law to state very prominently (and this certainly can be regulated, and it's not costly -at all- to do by the "prompters". I have indeed seen some doing it for moral reasons)  that the art was created using generative AI, instead of being the work of an artist. Still, it would remain the fact that such generation came from a non allowed use (whether to extract the style or blatantly copy parts of art pieces) of a lot of illustrated content for the creation of a tool (which is illegal in any other field), and then that print. This in the music industry is very well regulated. And at least for commercial usage, it is not that easy to bypass.  But the idea is not a digital protection, but enforcing the measure by law. As then, the average Jane and Joe will still be doing it not complying to the law (BTW, I have no complaints about the use in memes and non commercial use, in general), but like with piracy, unlawful use is not that practical for true commercial purposes. This is not my idea, it has been mentioned multiple times.

I think it is never comparable what is happening with AI to an art student learning. Besides that an AI is not a human, the AI is not learning what it takes to make a good art piece. It does not have a clue about what anatomy is, or at least does not really understand it, neither composition, color theory, art history, its own human context, etc. It has some parameters for making a sophisticated mashup, a prediction based system using enormous sets of artwork. And ultimately, it is a product developed using creative work (impossible without it) from artists, in a way not allowed legally by their authors.

I personally would have less of an issue if, by law, by regulation (otherwise people just don't comply) whether from EU legislation then copied by the US and the rest of the world, or the other way around, or if becoming a global thing in any other way, it was imposed to specify very clearly (in the credits and signature of the print, digital image, etc) that it is the result of generative AI. Then markets could be differentiated, many users would still prefer the cheap AI images, of course, but would be two categories. While currently, posters, prints, etc, are sold without any note about it (although the amount of errors is glaring, but not for everyone). The ethics of it, still not great, but at least it damages less the creators who made it, and contributes less to destroy an entire cultural field (which has built our history and it is more important to humanity than it is perceived to be), as it is doing already. The motivation for those not willing to take this simple measure would be quite telling, in my opinion.

But this is not in the hands of any developer or company (they could have the initiative to force a small watermark, tho, some are doing it. But the only definitive solution is regulation). I mention it as we are apparently discussing philosophic or ethical aspects of it.   

[RichardMH]

There's several ways AI is used in images. The focus here seems to be on generative AI. But there is also the image quality AI for noise and sharpness that Topaz and DxO use (and Adobe) and AI masking. Personally, I can't see any image editing software surviving long without AI masks. I suspect those of us that want the image quality AI are happy to use plug ins. Topaz are also developing AI lighting and white balance so there's a lot happening in the non-generative AI space.

[RichardMH]

12 hours ago, R C-R said:

How would that help for printed material that is fully or partially created with the help of generative AI that someone then sells?

I think the concern at the moment is more for documentary photography that claims to be a true representation of the world. e.g. in news media.  If a print is sold as a piece of art does it matter if some or all of it is AI generated? I believe there is already a lucrative market for AI art that is expected to grow.

[R C-R]

58 minutes ago, SrPx said:

I think it is never comparable what is happening with AI to an art student learning. Besides that an AI is not a human, the AI is not learning what it takes to make a good art piece. It does not have a clue about what anatomy is, or at least does not really understand it, neither composition, color theory, art history, its own human context, etc.

I think in this respect what it boils down to is how one defines intelligence, legally or otherwise, & I think there still unanswered questions about how to define non-human levels of intelligence, including but not limited to animals & ever more sophisticated & capable forms of machine intelligence. In fact, some highly respected people like John  von Neumann argue that it is inevitable that eventually machines will become smarter than humans, some form of the so-called singularity. 

[debraspicher]

1 hour ago, RichardMH said:

There's several ways AI is used in images. The focus here seems to be on generative AI. But there is also the image quality AI for noise and sharpness that Topaz and DxO use (and Adobe) and AI masking. Personally, I can't see any image editing software surviving long without AI masks. I suspect those of us that want the image quality AI are happy to use plug ins. Topaz are also developing AI lighting and white balance so there's a lot happening in the non-generative AI space.

Serif in a recent company statement shared by another user (see below) has stated they are putting resources into AI, so we may not need a plugin. I also agree with you, it's inevitable that this is where tech will go now as AI is the future. I don't have any particular issues with it being used to "enhance" work. My issue is when whole new pieces are created out of works that were added to a tool without consent.

If someone wants to take one of my color palettes and use it in their own design, that's a legitimate use as a creative tool, I think. That's akin to exploratory use-cases and is a suitable use for AI as a creative tool. It wouldn't be very different if I had input that color palette manually and had the algorithm work it... or used other filters, etc, to adjust. Using AI to "restyle" ones work, for example, provided it's not a total replacement. So the difference for me would be using an image we created as a "prompt" (causing AI to redraw it entirely), versus using AI to "filter" through the piece and add 3d effects, shadows, recoloring, adding objects, etc...

Spoiler

 

[SrPx]

5 hours ago, R C-R said:

I think in this respect what it boils down to is how one defines intelligence, legally or otherwise, & I think there still unanswered questions about how to define non-human levels of intelligence, including but not limited to animals & ever more sophisticated & capable forms of machine intelligence. In fact, some highly respected people like John  von Neumann argue that it is inevitable that eventually machines will become smarter than humans, some form of the so-called singularity. 

Well.. that runs way deeper than our practical case here (and our current problem as artists/designers/photographers), but I'd say I'm not sure about Neumann having enough data to predict that, so long ago, and more importantly: I have had that conversation often with quite a few individuals; it ends up always in having to agree to disagree. Mostly because I am convinced that we can't replicate intelligence to our brain's level in certain very key matters, even if we can make it quite more capable in several areas (as fascinating as the future arrival of "AGI" appears to be for some, if  that ends up happening). But mostly, because I value what is human (art is part of it) a lot more than any chunk of silicon or any piece of metal with some circuits on it, so, I wouldn't give the thing more rights than to a human, even if it got super intelligent (talking here about IP, copyrights). And overall, no matter how smart we can make it, it would not have the human's main characteristics. In any case, I'll retire way before AGI is even close. IMO what AI (which, btw, people in that field hate that it has been called "AI" by marketing folks, as it is not yet such, it's quite far) can do has been slightly over hyped, for VC's interests. 

Currently it does not boil down to that. At least not in its current state. It does not know what I know to make a proportionate human (hence the many errors, and it's only fixed (the 9 fingers, etc, etc, etc) with certain "tricks", not true understanding. It's just dealing with patterns), neither how all the 20 muscles of the forearm behave and change in every pose (neither do I, but yep till a sufficient extent). And neither is capable of many other skills achieved by an artist, those are not yet things that an "AI" is _really_ able to replicate, even less, consciously (but it can mashup stuff, use patterns, predict words in a context, etc), right now. Maybe in the future? Yes, probably. But that future is farther than people think. But the result, the output, is already something that can be sold/used to fully replace artists (the general public doesn't notice/even care, with said output). VC investors behind all those companies were not after giving artists tools to improve their work, give the artists more functionality to their workflows (even while a lot of devs will use AI for that, as there's a market for it, too, but much, much smaller). The main purpose was/is the replacement, otherwise these VCs would have not invested so much money in making full generative AI for images. Huge profit is at the core of all this, and only that. Replacing all labor of a field (not caring about the damage, not just to the artists, but to society, in the long run) means a lot of money gain, multiplying profit by a much larger factor than a more typical investment.    

[William Overington]

What please is a VC?

William

 

[Ron P.]

9 minutes ago, William Overington said:

What please is a VC?

William

 

Venture Capitalists

[William Overington]

Various posts about generative AI mention the legalities.

Please consider the following prompt to Bing Chat AI and one of the four images that was produced. Please note that I selected that image - two of the others each showed an okapi depicted wrongly with horns and one had the head to the left and the body separately at the right. Had they all been wrong I could have reset to a New Topic and tried again.

Please produce an original painting in the style of Claude Monet of a lady in a long green dress feeding an okapi in a garden where there is a pond with water lilies in flower.

Later I varied that prompt to the following.

Please produce an original painting in the style of Claude Monet of a lady in a long green dress feeding a stegosaurus in a garden where there is a pond with water lilies in flower.

As it happens I used that image to produce what is termed a photo greetings card and sent it to myself and I framed it to be part of my art collection.

As the template was notionally for a full-field square greetings card with the greeting inside customizable, to the great extent of being able to choose font style and size and have around fifteen or more lines of text in an ordinary text font, I chose Avenir at 15 point and added some text.

What do readers consider would be reasonable for that text to be, as if it were a public statement about the picture?

William

 

[William Overington]

Some readers might enjoy the following thread.

Compare and Contrast please (Page 1) — Art & Literature — Alfred's Serif Users' Forums (punster.me)

The first image is from some years ago.

The second image and the story were produced by Bing Chat AI in response to my prompts, based upon the first image.

William