Today I received an email informing me that Linkedin is integrating generative AI, except for āusers in the European Economic Area, Switzerland, and the United Kingdom. For users outside of these regions,ā they are making available āa setting that allows the user to opt out of sharing their information for this purpose.ā
Since I donāt know the extent of this change yet, Iām choosing not to share it for nowā¦
Iām wondering why they canāt make that opt-out setting available for everyone? Thatās not fair. 
People in these countries arenāt being scrapped for AI, hence the lack of option to opt out.
The rest of us plebs have to manually opt out and be thankful theyāre being gracious to let us for now. 
The other day I saw a Brazilian artist commenting that her commissions had dropped a lot because of AI. She had to start selling small products with her art for fans in order to support herself.
Another thing to think about: it is common for artists to send a layout for the client to approve before creating the final artwork, right? This attitude should be reconsidered, since there is already AI that takes a layout and converts it into a finished artwork.
So, there is a risk that the client asks you for a layout, you create it, they take your layout, use AIā¦ and never get back to you. You waste time and money.
Yeahā¦ Whatās worse, I observed a much higher quality of AI images recently (especially for anime stuff). Itās very difficult to spot that itās AI, I would say even impossible for a person not familiar with art or the subject matter.
Perhaps these images are now spot-fixed, or maybe some actual real artists incorporated it in their workflow to churn out more stuff quickly (and then correct the AI mistakes).
On the other hand, my experience has been with opposite scenario
I have had client send me rough layout that they did with AI and asked me to work based on those as sketches.
@YRH - In fact, we learn a lot by observing the AIās mistakes. One of its weakest points is the lack of criteria in what it selects. I noticed that many generated images have imbalances in brightness and contrastā¦ exactly because many images posted on the internet have this problem!
Another weak point is the inconsistency in the characters and scenarios. For example, there was a pirate: in one image he was wearing a frilly ornament on his chest, in another the AI āāreplaced it with a belt. The ornaments on his hat were one moment composed of a skull, another moment of three, and then of strange spiral shapesā¦
@raghukamath - In this case, there is still the risk I mentioned, because at some point you will need to send a sample of the artwork for the clientās approval. Of course, this does not necessarily mean that this will happen, because there are always honest clientsā¦ but it is worth taking precautions.
Iāve heard about people sending AI images as references but thankfully didnāt get sent any yet.
About the risks of not being paid after sending a preview, as a precaution Iāve changed my policy for clients I donāt know well to full upfront payment. I used to do split payments, but now only accommodate higher amounts and commissions from trusted clients.
Gen AI works fairly well for random images, but it canāt replace positions like concept artists for example. Anything requiring precision, patterns, or consistency canāt be done without serious work by actual artists. And as long they keep throwing more data and power at ML, itāll never be able to do these things because the issue lies with how it generates images and thatās not changing.
Itās still disruptive. Itās certainly eating market share in small private commissions, prints, and jobs in companies that think they can replace artists with ML. By the time the fad dies and they come round to re-hire some artists I hope people have raised their rates. 
Iāve also noticed the horrible contrast of gen AI. Some of it may be due excessive prompting, which causes images to kinda burn, but AI just sucks in general at very bright and dark images because not only it doesnāt know composition but it also has some limitations due the technology.
Have you noticed the eyes? People pay attention to hands but for me the eyes are downright creepy. Itās always excessive, with outlined irises when theyāre not simply misshapen.
I mostly do art for advertising and publishing. While publishing side has not sent me Ai generated references. the advertising side has done it. I think in this year alone I have got mood board filled with AI generated images. Sometimes the client just takes the images from pinterest and doesnāt know that it is AI generated. Sometimes client has generated the image by themselves.
Now there are two types of client one which asks for retouching the generated images fortunately I did not get to work by those types yet, and I rejected few queries which had this type of work. The second type knows that the images are AI and expects you to redraw and do the work from scratch with your own take but keeping the idea similar to the references they generated.
I am still trying to figure this out and how to deal with these.
True, but Iāve heard about some new Flux ML that solves the consistency problem, and can generate actual character sheets. Didnāt dig too deep, but it appears to be a paid program (correct me if Iām wrong).
Contrast and ācreepy eyesā problems can now be solved with the use of LoRA. These act like filters and can fix lighting, hands, eyes, etc. So it becomes harder and harder to figure out which images are AI/ML generated, and which arenāt. What Iāve found while experimenting locally with custom SDXL models is it canāt generate subjects it has no data for. For instance, if it has no patterns attached to the word ābatā, it will give you some random animal instead. The less data ML has, the less things it can do.
The best solution is to block access to new data for ML models. As long as it can accumulate more and more patterns, weāre in troubleā¦
I actually had second-hand experience of a company using tons and tons of AI art for 90% of their stuff. Yes, it looked bland, inconsistent, and terrible, but it looked good enough to investors so they genuinely didnāt care since they had unpaid interns slaving away to make things look passable in the actual āproductā (hard to call it a game when there was no ounce of creativity in itā¦)
It was a small company, sure, and they would hide the fact that they used AI to be able to copyright it all. It was wild. So, yeah, weird things are happening around the world and some people genuinely donāt care as long as they can pocket the money.
My faith lays not with execs but small communities of artists, forums, niche web corners. Theyāre my cures for all that ails me, and they definitely soothe the heartache I feel when I see just another soulless husk of what could have been.
The world is slowly but surely sinking into madness:
Eventually, even the code will be written by AI bots and they will criticise each other at very high speed until they disappear up each otherās output ports.
Well, apart from the fact that this did not happen out of sheer laziness on the part of the dev responsible, but because the dev responsible for this lacked the mental capacity to generate the required code themselves, they obviously did not see through the errors that were implemented in the software with the code generated by AI. A fact that I would excuse, however, since someone has tried to solve a problem that exceeds their own abilities with tools that are advertised as supposedly suitable for this purpose. Up to this point, this is nothing reprehensible in my eyes.
However, in order to then, immediately afterward, let the tool, which has already proven that it was not able to generate correct code, describe an error that it cannot recognize itself, otherwise the faulty code would not have been generated initially, the dev responsible for this must be mightily mentally underexposed.
Michelist
Add: The report shared by @AhabGreybeard is also bizarre. At least it sounds to me as if would-be error finders are using AI, i.e. other peopleās feathers, to decorate themselves and stand out. Like these freaks using generative AI to state they now are artists ā¦
We also had one or two cases here on the forum where someone wrote Krita allegedly has a security vulnerability and that they got told so by ChatGPT (if I remember correctly). Also several āplug-in developersā wondering why their AI generated plug-in wont work (Spoilers: because the AI pulled the API out of their virtual behind).
I saw an interesting videoā¦
There the creator talks about an article where OpenAI made the following statement:
āWe are always working on how we can make our systems more robust against this type of abuse.ā Ā¹
Which I can understandā¦ it is a lot of work to scrape images from the internetā¦ without permissionā¦ and when someone than uses glaze or nightshadeā¦ 
I remember someone suddenly unalive when he wanted to give testimonies about openai using copyrighted materials. Which is potentially bankrupt the whole thing.
Oh, boo hoo. OpenAI is crying because DeepSeek may have used their model for training without permission.
concerns grow over a potential breach of intellectual property.
OpenAI says it has evidence Chinaās DeepSeek used its model to train competitor
https://www.ft.com/content/a0dfedd1-5255-4fa9-8ccc-1fe01de87ea6
OpenAI ā¦ Well, itās āOpenā, isnāt it? 
It seems that theyāve been āhoisted by their own petard.ā (blown up by their own bomb 
I learned that from a Perry Mason episode)