Anti-AI tools for user created images

[Bobby Henderson]

While AI has the potential to eliminate a lot of creative jobs, I think computer users in general should be really alarmed at the idea of software companies using AI to write code. If developers lean more and more onto AI as a crutch for writing code their own understanding of how the code works will atrophy. Then when problems, such as security vulnerabilities are discovered, the developers won't be able to fix the problems as quickly. Or they'll be dependent on AI to somehow fix the problem.

[Old Bruce]

2 hours ago, Bobby Henderson said:

While AI has the potential to eliminate a lot of creative jobs, I think computer users in general should be really alarmed at the idea of software companies using AI to write code. If developers lean more and more onto AI as a crutch for writing code their own understanding of how the code works will atrophy. Then when problems, such as security vulnerabilities are discovered, the developers won't be able to fix the problems as quickly. Or they'll be dependent on AI to somehow fix the problem.

I am more concerned about bad actors using machine learning to write code with well hidden security vulnerabilities which can be exploited later.

[Mr. Doodlezz]

7 hours ago, VectorVonDoom said:

simply bad

Still, what does that even mean? I get the feeling you're trying to troll me. It's kind of working, though.😄

There is no indication how to determine what is bad art and what is not, for humans or AI, so it remains highly subjective.a majority thinks a work of art is bad doesn't make it bad per se. It just proves that it's controversial, not in tune with popular taste or understanding.

Many artists faced rejection because their style was new, unconventional, misunderstood or, to put it very simple, bad and out of step with the period in which they lived. These artists, such as Van Gogh, Monet and others, are now studied and highly regarded in universities and art schools.

Are there and will there be critics who claim that the work of these artists and future artists is bad? Of course!
Do these or any other critics ultimately decide or define (and therefore think) for you or me what is good and what is bad? Definitely not!
But do they influence large groups of people aka the mainstream? Yes.
Is the mainstream the measurement of things? Unfortunately, for unconventional artists, and for most of the time, yes.
Does all the above make and artist's work actually bad? Nope.

The only arguments where I personally could see »bad« as the fitting label are questionable intentions/messages being conveyed or the subject/matter/issue (such as abuse or violence, although this is also a matter of context, e.g. if it is about awareness of the subject). Oh, and what some people might sell as AI-generated art, of course, ha.

Back to the question and AM's idea to flood the internet with bad art: How to determine what is bad art and what is not, for humans and AI?
AI doesn't care if we think something is bad, it doesn't criticise its training material the way we might look at and judge art, unless it's also trained to ignore certain aspects.
I think they have also grazed and stored enough material to be able to generate a wide range of styles that are compelling enough for the industry to use as a tool as it is, looking at some ads and commercials I came across.

That's why I had to chuckle at the idea of flooding the internet with bad art. Because if we take everyone's or every group's opinion into account, it's already full of rubbish. For some reason/s one group thinks pixel art is bad, another thinks comics are bad, another thinks 3D rendering is bad art, another thinks someone's child's drawing is bad art, another thinks digital/traditional art is bad and so on ‒ but of course no one's ultimately right, no one's ultimately wrong.

Regardless of this discussion: There's always use for it for AI and information to be extracted, even if it's just the stroke of a doodle from school or »my nephews kindergarten crayon drawing«. Someday someone will need a crayon drawing for their project and enter the prompt »photo of a crayon drawing of a 4 yo depicting a spaceship landing on mars, 16:9 crumbled paper, shallow DOF« or something like that, because they need it for a poster, ad or movie or what ever. And it will create just what they need, based on my nephews »bad« art. Actually, it's already happening.

[debraspicher]

Bad art is art with no thought put into it. Similar to bad design. Not everything hangs on aesthetics.

[v_kyr]

Well AI generators do and will evolve over time too.

Adobe Firefly for example designs pictures and lettering from text inputs with artificial intelligence. Vector graphics, videos and other content should follow. Firefly was trained with curated and community-free content - for legally harmless use. Safe from cautions ... According to their own statements, Adobe trained his AI service with curated data from the image agency offer from Adobe Stock and with public domain content and those in which copyright law has already gone out. The group wants to minimize the risk of unwanted copyright violations. Firefly is also involved in the Content Authenticity Initiative (CAI). Adobe launched her to document the origin and processing history of image material. The Firefly service provides the content created with tags, which you identify as AI generated and on request with a do-not-train Tag, so that they are not in turn readed/used by a AI. In the current beta phase, all generated images still have the note "Image not for commercial use". Adobe has not yet announced when the beta phase ends. So far, Firefly only generates images and lettering (text effects etc.). In addition, the service can recolor vector graphics. As with DALL-E 2, Midjourney and Stable Diffusion, the basis for this is a text line in English, the text prompt.

Microsoft integrates the Dall-E image generator from OpenAI into the chat function of its Bing search engine and into the Edge browser. There, the so-called Bing Image Creator Preview can be called up via the sidebar. Its use is free of charge until further notice, but a Bing logo is emblazoned in one corner of each generated image.

The design web app Canva has also been extended with AI functions. An image generator creates photo-realistic or graphic images after a line of text is entered - even in Canva Free. Other AI-powered image editing tools adjust embedded photos and images: "Magic Edit" retouches content in images, "Magic Eraser" reconstructs the background. AI helpers are also available for designing entire documents: they create coherent layouts and assist with text.

Meanwhile, more and more sample images are appearing on social media, demonstrating how far version 5 of the Midjourney image generator, released in mid-March, surpasses its predecessor. Its major shortcoming was that it could not produce photorealistic images of people. They primarily resembled drawings and the skin looked as if it had been smoothed out. Version 5, on the other hand, produces much more realistic textures, for example of human facial features and hair.

Now also a "Segment Anything Model" research demo of Meta AI appered on the net, which offers to cut-out objects from images.

 

[prophet]

3 minutes ago, debraspicher said:

Bad art is art with no thought put into it.

How would you know how much "thought" was put into a piece of art?

[debraspicher]

7 minutes ago, prophet said:

How would you know how much "thought" was put into a piece of art?

What is plagiarism?

[Mr. Doodlezz]

18 minutes ago, debraspicher said:

Bad art is art with no thought put into it. Similar to bad design. Not everything hangs on aesthetics.

What if it's exactly what the artist wants? Did Pollock, for example, put thought into the marbles that rolled across the canvas and the drips that created the works that are now praised?

I very much doubt it, and to me personally it looks completely random. But what if (and maybe that's what he wanted, apart from the choice of harmonising colours and the general direction in which they moved) that's what his works are about? The complete randomness?

[StrixCZ]

On 4/9/2023 at 2:37 AM, nezumi said:

I hope Affinity team will not waste time on AI...

I would rather see good automatic tracing in Designer rather then some AI useless bloat

Obviously, you're misunderstanding what AI is or what it can do, as automatic tracing is exactly one of the tasks AI can be incredibly useful to help with. Creating images based on text prompts is only one of many possible uses of AI - it's definitely not the only (or even the main) one. And since Affinity is trying to compete with Adobe, they actually need to adopt AI tools if they don't want to be left in the dust. Have you seen what you can do with Photoshop's "neural filters" (based on AI models)? Considering they're only in beta stages yet, they're pretty impressive already. And if Affinity doesn't come up with something similar, it will eventually render their suite obsolete. I mean, not many people will choose to spend hours on tasks AI can take care of in a matter of seconds (and we're not far from this being the case)...

[StrixCZ]

On 4/4/2023 at 10:56 AM, PaulEC said:

Sorry, but I simply do not accept that anyone should be able to steal other people's work; any more than I accept that it's OK to break into someone's house and take whatever they want!

Sadly, the question is not whether they "should be able to". The question is how to deal with the fact the they "already can" (and I don't think any sort of regulation can stop it now). To put it in perspective with your "breaking into someone's house" metaphor, we're talking about the equivalent of trying to ban production/sales of crowbars or making it mandatory for anyone buying one to sign some sort of "ethical use" contract (and expecting them to never break it, LOL).

[nezumi]

@StrixCZ And you obviously did not read how this thread started... Theres hardly anything complicated about pseudo AI (pseudo because its not artificial intelligence, but this acronym sells better then "machine learning"). You dont have to explain it thank you very much. 
I was referring to fact that instead of jumping on AI wagon as soon as possible we could first take care of things that are missing from the traditional tools. I dont mind if so called "AI" will be used internally for tracing. Or if Photo gets some tool for better content aware fill. I am not blindly against pseudo AI. Just against wasting resources for absolute GARBAGE GIMMICKS - like "now you can generate dress", "now you can add a smile and turn head 5 degrees on photo", "now you can transfer makeup" or use photo of a palm to make a fake shadow thats basically outline of the palm filled with black... Come on mate, all this are useless gimmicks. I dont know about you but I have ZERO needs for transferring make ups and generating dresses in day to day job... In all the stuff I saw from Adobe I was impressed by ONE. That "ai" tool that automatically was recognizing shapes in 3D and deforming texture accordingly. That was cool. Rest is bunch of toys for... I honestly dont know for who. And that includes their recent picture generator and generator for letters... All looks like toys made for kindergarten attendants not professional designers.

[debraspicher]

My thought is if someone needed AI that badly to be productive and it translated to much quicker turn around (ie more $$$), they could just use Adobe. The costs would pay for themselves instead of struggling to adapt to the shortcomings of other software that would slow them back down.

The other consideration is using a third party tool to supplement AI-esk tasks. This is the sensible option if Serif continues working out imperfections in their suite and makes it more productivity-based. A "master of all" approach, while guaranteeing more market share, is why Adobe has such a high price tag and a barrier for entry (subscription model). Price for value unfortunately is a major factor in these considerations. So while I can see them adding in some AI-like tools like @nezumi mentions, I can also see where they won't touch AI for quite a while and wait for those benefits and best use cases to mature a bit within the industry.

We forget also, because Serif is not a subscription-based, their customers pay for the software before hand. In some ways it would make more sense they wait until 3.0 to release AI and use that to justify people to upgrade (again). So I don't see the incentive to add in AI features as much as others when the potential backlash over the ethics might outweigh any benefits when the full impacts of these tools isn't exactly clear. Is it going to crater our livelihood by relying on these tools? Or is it another way to get more customers away from their Adobe robot masters, by becoming the robot master themselves? Just some thoughts.

[AffinityMakesMeWonder]

On 4/10/2023 at 8:13 PM, v_kyr said:

Well AI generators do and will evolve over time too.

Adobe Firefly for example designs pictures and lettering from text inputs with artificial intelligence. Vector graphics, videos and other content should follow. Firefly was trained with curated and community-free content - for legally harmless use. Safe from cautions ... According to their own statements, Adobe trained his AI service with curated data from the image agency offer from Adobe Stock and with public domain content and those in which copyright law has already gone out. The group wants to minimize the risk of unwanted copyright violations. Firefly is also involved in the Content Authenticity Initiative (CAI). Adobe launched her to document the origin and processing history of image material. The Firefly service provides the content created with tags, which you identify as AI generated and on request with a do-not-train Tag, so that they are not in turn readed/used by a AI. In the current beta phase, all generated images still have the note "Image not for commercial use". Adobe has not yet announced when the beta phase ends. So far, Firefly only generates images and lettering (text effects etc.). In addition, the service can recolor vector graphics. As with DALL-E 2, Midjourney and Stable Diffusion, the basis for this is a text line in English, the text prompt.

Microsoft integrates the Dall-E image generator from OpenAI into the chat function of its Bing search engine and into the Edge browser. There, the so-called Bing Image Creator Preview can be called up via the sidebar. Its use is free of charge until further notice, but a Bing logo is emblazoned in one corner of each generated image.

The design web app Canva has also been extended with AI functions. An image generator creates photo-realistic or graphic images after a line of text is entered - even in Canva Free. Other AI-powered image editing tools adjust embedded photos and images: "Magic Edit" retouches content in images, "Magic Eraser" reconstructs the background. AI helpers are also available for designing entire documents: they create coherent layouts and assist with text.

Meanwhile, more and more sample images are appearing on social media, demonstrating how far version 5 of the Midjourney image generator, released in mid-March, surpasses its predecessor. Its major shortcoming was that it could not produce photorealistic images of people. They primarily resembled drawings and the skin looked as if it had been smoothed out. Version 5, on the other hand, produces much more realistic textures, for example of human facial features and hair.

Now also a "Segment Anything Model" research demo of Meta AI appered on the net, which offers to cut-out objects from images.

 

Late to the Adobe Firefly party, but, I have been playing around with Firefly for two days now, and, I am VERY impressed with the results!

Much joy and easy to bring hours after hours with Firefly… Recommend others to join the Beta-part of Firefly.

[Mr. Doodlezz]

6 minutes ago, AffinityMakesMeSmile said:

Late to the Adobe Firefly party, but, I have been playing around with Firefly for two days now, and, I am VERY impressed with the results!

Much joy and easy to bring hours after hours with Firefly… Recommend others to join the Beta-part of Firefly.

Um, I'm not sure you've commented on the right topic – I mean, the title clearly says »Anti-AI tools« while you're suggesting we join and try an AI tool.

Kind of ironic, isn't it? 🤔

[AffinityMakesMeWonder]

42 minutes ago, Mr. Doodlezz said:

Um, I'm not sure you've commented on the right topic – I mean, the title clearly says »Anti-AI tools« while you're suggesting we join and try an AI tool.

Kind of ironic, isn't it? 🤔

Ha, my sorry…

In the same time, people against AI should at least test some AI services so they have some sort of reference what they dislike…

Some services are really great - I use an AI online service that converts audiofiles to MIDI-files, and, the result is stunning!

[Mr. Doodlezz]

2 hours ago, AffinityMakesMeSmile said:

Ha, my sorry…

In the same time, people against AI should at least test some AI services so they have some sort of reference what they dislike…

Some services are really great - I use an AI online service that converts audiofiles to MIDI-files, and, the result is stunning!

Jag tror att de flesta av oss har testat AI någon gång. Det handlar inte om AI-tjänsterna i allmänhet, utan om hur leverantörerna får sitt utbildningsmaterial och hur snabbt det utvecklas till något som producerar anständigt innehåll för ett mycket lågt pris, vilket lämnar skaparna och konstnärerna med i princip ingenting, samtidigt som det är de vars arbete ligger till grund för AI-tjänsten i första hand. Detta gäller särskilt digitala konstnärer inom illustration och fotografi. 😔

För närvarande respekterar hela systemet inte alls förvaltningen av kreativa rättigheter. Det är därför vi är så upprörda (för att uttrycka det milt). 😑

Jag använder till exempel DeepL som ett verktyg för att förbättra mina översättningar. Eller för att översätta till ett språk som är helt främmande för mig. Men jag tycker inte att det kan jämföras med konst. Jag antar att författare skulle argumentera annorlunda om det också, men jag känner att ord är ord, och det är inte så att mina översättningar är skrivna i en särskild stil som från en berömd författare. 🤔😥

[dcr]

22 hours ago, Mr. Doodlezz said:

Um, I'm not sure you've commented on the right topic – I mean, the title clearly says »Anti-AI tools« while you're suggesting we join and try an AI tool.

On top of that, it was an Adobe tool and, I don't know about others here, but as far as I am concerned, Adobe can burn in the deepest, hottest, fieriest pits of Tartarus.

[v_kyr]

• "Like Photoshop on steroids": OpenAI boss admits risks of AI revolution (Google EN translation - of a Heise.de online article)
• "Wie Photoshop auf Steroiden": OpenAI-Chef gesteht Risiken von KI-Revolution ein (DE original Heise.de online article)

[Abdollin]

اشتريت هده الصورة لغلافي من أحد المصميمين و أريد تأكد من أنها عمله الخاص

 

Edited April 21 by Abdollin