pancakes

MicrostockGroup Sponsors


Author Topic: Original Work at Risk - AI image to image exploitation  (Read 435 times)

0 Members and 1 Guest are viewing this topic.

« on: Yesterday at 00:00 »
+1
Recently, I came across many contributors which have used my work in generating similar AI-generated images.

Creating and maintaining original works takes time and with such AI tools creating replicas is super easy.
This is serious copyright infringement and the companies needs to look into this.

Companies should keep the original work and AI generate work as a separate two categories.

I feel sad to see this.  :(


« Reply #1 on: Yesterday at 01:40 »
0
This is a real problem.

I even found one case where somebody copied my image with a German word, but the word is spelled wrong in his image and he has good search positions with it.

What I am trying to do is make my descriptions much shorter and more generic. Then at least they have to load a screenshot of my file into an ai for description, one additional step.

I hope they prefer to go to a port where every description is perfectly usable directly for ai.

But it is a problem.

I do believe especially with people, many customers will choose to exclude ai from their search, which is very easy to do.

« Reply #2 on: Yesterday at 04:57 »
0
This is a real problem.

I even found one case where somebody copied my image with a German word, but the word is spelled wrong in his image and he has good search positions with it.

What I am trying to do is make my descriptions much shorter and more generic. Then at least they have to load a screenshot of my file into an ai for description, one additional step.

I hope they prefer to go to a port where every description is perfectly usable directly for ai.

But it is a problem.

I do believe especially with people, many customers will choose to exclude ai from their search, which is very easy to do.

Making the description shorter will not stop someone trying to describe your image. You can send the image to an AI API, let the AI describe the image and use that as a prompt. But yes, it is an extra step and ill-fated newbies who just watched on how to become a millionaire with adobestock youtube video won't be able to do that, but people who understand ai and use it programatically will.

« Reply #3 on: Yesterday at 05:37 »
0
Recently, I came across many contributors which have used my work in generating similar AI-generated images.

Creating and maintaining original works takes time and with such AI tools creating replicas is super easy.
This is serious copyright infringement and the companies needs to look into this.

Companies should keep the original work and AI generate work as a separate two categories.

I feel sad to see this.  :(

You can be sure that these contributors are reading you on this forum... They like AI power sooo much...

« Reply #4 on: Yesterday at 10:05 »
+2
This is a real problem.

I even found one case where somebody copied my image with a German word, but the word is spelled wrong in his image and he has good search positions with it.

What I am trying to do is make my descriptions much shorter and more generic. Then at least they have to load a screenshot of my file into an ai for description, one additional step.

I hope they prefer to go to a port where every description is perfectly usable directly for ai.

But it is a problem.

I do believe especially with people, many customers will choose to exclude ai from their search, which is very easy to do.

Yes, as I've said many times... "AI" is not an actual "thinking" machine, but rather (for the most part all jewish-run as well) theft based machines. So basically they steal/scrape data, then arrange it into "buckets" - then "blend" the output (this is also true of things like chatGPT, etc).

So if you have a "detailed" description, i.e., "Sunset in a blue mountain with three seagulls flying above at 3:05 p.m." - when the so-called "AI" steals the images - there are very few images (if any aside from that one) that it can arrange into buckets called "Sunset in a blue mountain with three seagulls flying above at 3:05 p.m.". (There of course is a little more to it than just that, but for simplicity this is how I am explaining it).

So if someone uses that exact same prompt - because the "reference" material the AI stole from only has that SINGLE image - it will most likely "create" (aka "blend") something very similar - because it doesn't have other reference images to blend together with... So that is why the image(s) will look virtually identical.

HOWEVER, if you have something called "Sunset". Well, there are probably 100's of thousands of images just called "Sunset" - so it now can steal ("blend") from 100,000 images - and so it is unlikely that it will look 'just' like your image...

So Cobalt - you are correct - by using generic descriptions - it makes it much more difficult for someone to simply steal your work.

Because the "ai" tools are ALL THEFT based. They are NOT "creation" tools - they are "THEFT-BASED" tools. That is why it makes it "easy" for people to "steal" things using the same prompts.
« Last Edit: Yesterday at 10:11 by SuperPhoto »

« Reply #5 on: Yesterday at 12:37 »
0
Maybe leave out the racism, how about it?

And IT stuff usually comes or is co developped in Asia, so there will be lots of hindu, muslim developers as well in addition to atheist, christians and maybe wiccan believers...

The biggest disadvantage of having ultrageneric titles and descriptions is that it supposedly is bad for seo. Your files will be more difficult to pick up by the crawlers and probably achieve less specific rankings.

It is also bad for the agencies for the same reason.

But I think a lot of people will start doing it in the hope that the copy cats move on to the ports with all the prompts still included.

There are plenty of legit ai creation tools, i.e. the ones where we get paid for data set trainings.

Today I got 71 dollars from pond5, in November it was 180. For 1100 simple clips that is a nice add on.

The people with 30k ports are probably again swimming in money.

So if we get paid, ai training is a different thing, it becomes additional income.

« Reply #6 on: Yesterday at 16:52 »
0
Maybe leave out the racism, how about it?

And IT stuff usually comes or is co developped in Asia, so there will be lots of hindu, muslim developers as well in addition to atheist, christians and maybe wiccan believers...

The biggest disadvantage of having ultrageneric titles and descriptions is that it supposedly is bad for seo. Your files will be more difficult to pick up by the crawlers and probably achieve less specific rankings.

It is also bad for the agencies for the same reason.

But I think a lot of people will start doing it in the hope that the copy cats move on to the ports with all the prompts still included.

There are plenty of legit ai creation tools, i.e. the ones where we get paid for data set trainings.

Today I got 71 dollars from pond5, in November it was 180. For 1100 simple clips that is a nice add on.

The people with 30k ports are probably again swimming in money.

So if we get paid, ai training is a different thing, it becomes additional income.

Lol - are you referring to me? It's not "racist" stating the companies are jewish-owned - because they are - check for yourself - and it is significant to point that out when you understand (in 'general') their idea of what "doing business" is. It's most likely not the same as yours. Also - it's important to understand how certain cultures conduct business - and - the fact a lot of "anti-hate" legislature is specifically designed to make people such as it seems yourself afraid of stating the obvious. Look into the ownership of the companies - hang around a few to figure out the "business ethics" (or lack thereof) and some things will become quite apparent.

As for employees - you are correct. A lot of IT stuff is indeed developed in Asia, as well as East india. However, many IT companies have had to re-hire europeans, because after they went to hire "east indians" -  the quality was so utterly crappy - they had to get it redone. (I've personally hired east indians too and experienced that first hand... MULTIPLE times. It's just how they do business - they see what they can get away with, with doing the least amount of work possible - and if can be stolen and passed off as their own - even better). For asians - (western) companies have realized that is not necessarily a good idea either - because (a) many times they don't necessarily know what they are doing, and (b) data "mysteriously" goes missing, and magically competiting companies pop up. Lol - it is "SO" bad in some cases - that even ASIAN companies are EXTREMELY secretive within ASIA so their own employees don't steal their trade secrets.

Not "racist" when its true. I'm stating the obvious.

(PPS, being German, I thought you would be aware of the "programming" designed to make Germans feel REALLY bad about themselves for something that really didn't quite 'happen' the way the (jewish-owned) media portrays it... But that is an entirely different topic).

Getting back to the titling...

Yes. Ultra generic titles (potentially) can be 'bad' for SEO, if you are hoping to get SEO based sales (primarily from google). So it is a tradeoff - have some east indian rip it off to "get rich quick" when you use a complete title - but - have a chance of getting a sale from SEO... or make it harder for them to rip it off, but potentially affect SEO sales... Up to you...

Re: the dataset earnings - is that for the last year? Or just this month? If that is something you wanted/opted into, and are happy with - good job!
« Last Edit: Yesterday at 17:14 by SuperPhoto »

« Reply #7 on: Today at 01:29 »
0
I opted into dataset trainings to get paid.

ALL my work and ALL my images anywhere on the internet have already been stolen without my consent to train ai software.

I refused to pay for Midjourney for a year because I hat their CEO. Still do.

But now I see getting an income from ai images and from dataset training as a way to get my royalties.

I would prefer to create ai only with software like firefly, but the quality is not as good.

So Midjourney and others stole all my files without asking me.

This is my way of getting paid.

I have 1100 videos on pond5 and few sales. Last month I got 180 dollars for dataset, this month 71. That is more than I got in sales, even though I had sales in both months.

Most of my clips are simple handheld shots done with my iphone.

We will se how it goes, but ai is a useful tool, it should just have professionally paid traning material.

If I am getting paid for datasets it means soemwhere a company is doing the right thing - paying for training material.


« Reply #8 on: Today at 05:46 »
+1
Hi Artist,

I have a same problem as you. Some of my unique images were copied numerous of times with same long specific description and keywords. There are 200 pages of same image in search on Adobe, with almost no difference between them because they all  use only one image (mine) for a source and there is no similar other in concept and style for learning.

I guess if I complain to Adobe, they will say that there is no proof that they are using my image as source, it is AI... (?!)

My income from adobe is decreased and these images are no longer best sellers.
I also notice many portfolios with AI images on Shutterstock and some of them are best sellers on first page!

Making original images is time consuming, but I also put a time and effort to describe and keyword them and they just copy paste it from me. I'm furious and I'm so demotivated to create new content for thieves.


 

Related Topics

  Subject / Started by Replies Last post
3 Replies
3324 Views
Last post March 06, 2008, 13:33
by Waldo4
5 Replies
4045 Views
Last post December 22, 2010, 16:58
by disorderly
0 Replies
2119 Views
Last post April 02, 2011, 22:56
by lola
0 Replies
2056 Views
Last post December 18, 2012, 17:41
by Cricket
13 Replies
6221 Views
Last post July 31, 2016, 15:00
by ShadySue

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors