AI generative fill

That is not entirely true. Probably the main reason is that artificially creating image content is not part of PhotoLab’s mission. Something like content aware fill would be.


If your photos are posted with explicit licensing that allows the companies to use it for AI training, then yes, no problem at all. But it has been shown that some companies used license-restricted images. Some of the generative AI models have actually been shown to include copyright logos from their training data in generated images.

What do I lose? I lose the right to control how my images are used.


AI isn’t just one thing. It can range from assisting us with tedious tasks all the way to the creation of self-aware entities.



They literally just scrapped Getty images database. No permission.

Deviant art as well. Also they scrapped images from websites and anywhere else they could.

I just tested it. I wrote Trump in a crowd. Basically stable diffusion training just took it off google image search it seems.

Japan Sets the Precedent for AI Copyright

Amid global governmental regulations, Japan has come to a conclusion already about AI copyright - it does not apply to AI training at all.

By Mohit Pandey

Imagine if you are Anime artists. Why bother even training for it, which was one of the big Japanese cultural exports. Why bother training to be anime artist when the goverment and corporations can legally steal your work and use it to make profits. Beyond ethical and legal issues, its also killing motivation for whole generation of people to invest in creative arts. This is like one child china policy. By the time the consequences are felt a generation or two later, it will be too late. Short minded panic and greedy solution for long term damage to whole culture.

China is well known for copyright infringement, theft of intellectual property, and lack of enforcement of any laws that do exist. This discourages R&D (research and development) because after company does all the hard work, another one simply copies the finished version. This is also problem for forign investors. And ultimately limits innovation in the country itself, because it rewards copycats, not inventors.

This problem is not unique to China and with the whole problem where user generated data becomes the product to sell, like in western tech world, it has other issues of privacy and protection of legal rights. Face recognition technology, data generated on social media platforms etc. All repackaged and sold to advertisers, investors, forign and domestic governments etc.

At some point the big companies and big goverment, simply monopolizes the market, via both often illegal and unethical means. But more to the point as well, it discourages innovation and creative work by many people, diminishing the culture. This has profound consequences when it happens on a grand scale and over few generations. And everyone is affected.

The doomsday talk of AI ending many jobs is a story that is not always true, but it can be. Some jobs are truly going to be replaces by AI, but a lot of it is pointless creation of jobs that really should not exist in the first place, and also there is a lot of hype so investors, gambles would bet on future AI earnings. Similar to bitcoin bubble before it AI speculation is another financial bubble, that is about to burst. When that happens, lot of companies using AI for marketing will be out of work.

People who talk about aI art, or AI created art, really should say, humans created everything, neural networks only statistically blend text prompt with image data to collage it. Machine does not understand, it has no creativity. Its just responding to statistical probability. Only humans have creativity.

1 Like

Well, we just look at this differently, I guess. I prefer to take a pragmatic view. Even if a company captures one of my copyrighted photos in a trawl of many 100s of thousands of online images,

  1. My image is not being individually displayed (its only data associated with assigned categories)
  2. I am not losing access to my image (so the stolen bike analogy doesn’t work)
  3. I am not suffering a loss of income
  4. I am not suffering a loss of credit/attribution as the image’s creator
    … and I may actually benefit from the product that is produced!

Moreover, if your logic is strictly applied, we may end up restricting all sorts of content such as independent movie and music reviews, reaction and response videos on YT, internet-based research (and likely many more I can’t think of and the moment).

In the case of AI training, your copyrighted image that got scooped up in the training set (as one among many 100s of thousands) has not tangibly increased the revenue of the company and you have not lost revenue. Society is too litigous already … we need to relax on some things.

That said - I do become more uncomfortable if the AI produces artificial images that bear close resemblance to existing copyrighted/commercial work (i.e., so that the generated image can be tied to a specific source image). That said, I am not very profit-oriented myself, so if I recognised one of my images in an AI creation, I would be more amused than upset…

I don’t disagree with your points, but once again… this point is the pirates’ argument.

This comes under the definition of fair use. These all reproduce and credit small, usually unmodified portions of copyrighted material. The AI image trawlers take everything and do not credit anything.

This is like saying there is no point voting in an election because one vote won’t change anything. The revenue of the company is based on the aggregate training set. No, removing one image, or even all of one individual’s images won’t change a lot, but if we’re arguing “in the large” then all people should be accorded the courtesy of not using their material without permission. I think it was Flickr who were talking about providing specific licensing options to allow AI training use. In that case, it’s perfectly fine.

And here we arrive at the crux. You see your own indifference to the profit motive as a justification for these companies who are very much driven by profit. You can be very sure that if these companies sell their results, the price you pay for those results will be quite different for “personal use” than if you make a profit from using them. This is a common licensing/pricing concept, usually seen as “personal” versus “business” licenses.


“I don’t disagree with your points, but once again… this point is the pirates’ argument.”
No, it’s not. When ‘pirates’ take content, they do so with the intent of sharing it as an alternative to purchasing it. If a photo ends up in a training set, it is not diverting someone from purchasing it.

" This comes under the definition of fair use. These all reproduce and credit small, usually unmodified portions of copyrighted material. The AI image trawlers take everything and do not credit anything ."
Yes, I and think trawling the internet for massive training sets should be fair use. I’m not even sure how you would go about crediting individual images from among many 100s of thousands. And they’re not even reproducing or displaying individual images [it becomes problematic when they are].

“This is like saying there is no point voting in an election because one vote won’t change anything”
My point is that you have not been tangibly harmed if your image is scooped up. The election analogy does not apply (though yes, you have been tangibly harmed if you are prevented from entering the voting both).

" [quote=“DocNo, post:27, topic:34177”]
That said, I am not very profit-oriented myself

“And here we arrive at the crux. You see your own indifference to the profit motive as a justification for these companies who are very much driven by profit”"

And no, you misunderstand my point. I’m referring to the person who worries one of their images may be scooped up by an AI trawl, not the companies training AI on such images. Of course, I understand companies are in it for the profit. I run my own quite profitable company myself. But I will often do out-of-scope work for clients because I see myself as a professional and I value my business relationships. So, as I said, I would not be bothered to see something resembling one of my images generated by an AI engine. To repeat: I would not be losing anything as a result

Milan, what is your real point here?

All knowledge/“know how” tends to be cumulative. So is AI. That´s why I would be very surprised if AI will get all that harnessed in the future legally. It´s a too valuable force to boost productivity to fuck with.

Immaterial rights are an interesting problem though and in the US they are facing a new era where their whole already pretty oversized legal system might get very focused on milking the software industry of money in endless trials over immaterial rights. Maybe this will restrict the AI-industry far more than the so far pretty confused attempts by both the American authorities and EU to domesticate the AI-development.

Anders Håkan Lans, born November 2, 1947 in Enskede parish in Stockholm,[1] is a Swedish inventor that worked for [Försvarets forskningsanstalt] in Sweden (Försvarets forskningsanstalt – Wikipedia) (FOA). While working at FOA, he developed a color graphics system for computers, for which he personally obtained patent rights. For many years, various disputes have been current in US courts regarding this patent.

For ten years (1985–1995), Håkan Lans pursued a lawsuit against Hitachi regarding the infringement of his American patents on graphics cards. The parties agreed to a settlement, which contributed to IBM signing a license agreement.

However, Dell, Hewlett-Packard and Compaq used the technology without a license. The companies were sued in 1997 but Lans lost this case due to his American lawyers suing the companies in the wrong name. They had sued them in Håkan Lan’s name instead of in the name of his wholly-owned joint stock company Uniboard. Uniboard is the legal owner of the patents according to the US ruling. Lance was ordered on September 6, 2001 to pay all court costs.

In 2005, the decision was re-examined and the verdict was handed down in June 2005. This verdict is the same as the previous one: Håkan Lans lost and was ordered to pay all court costs.

In 2008, Gateway (the company was acquired in 2007 by Acer Inc.) sued Håkan Lans in Nacka District Court to obtain the legal costs that Lans was ordered to pay in the American judgment from 2005. Nacka District Court decided that Håkan Lans was obliged to do this. The judgment was appealed and on November 5, 2010, Svea Hovrätt decided that the Nacka District Court cannot order Håkan Lans to pay court costs according to the American judgment without trying the underlying legal issues. Nacka district court decided in September 2013 that Håkan Lan’s company Uniboard must bear the legal costs.[7] The judgment was upheld by the Court of Appeal in June 2014, when it did not consider that Håkan Lans had succeeded in showing that any errors had been committed in the American verdict.[8]

Håkan Lans has also sued his lawyers. In April 2012, Lans settled with these. Lance commented that “the parties are satisfied”.[9]

**My point showing these texts from Wikipedia is to high light that it´s not all that easy even if you seem to have a pretty solid case to take on big international corporations in US courts. Who really dares to take risks like that that might cost you multi miljons if you lose. These facts makes it pretty safe for big companies to steal all sorts of immaterial material from actors with less financial muscles. They steel what they can because they can and they cry crocodile tears when other people like the copycats in southern Chinas in turn steel what these companies might have stolen in the first place from smaller actors that of different reasons failed to protect their immaterial properties.

There have been disputes over even other inventings Håkan Lans invented:

At FOA, Håkan Lans invented a pointing device, a [puck](Puck – Wikipedia(pointing device)), for a digitalizing table. Sometimes it is said that Lans was the one who invented the computer mouse. However, that honor must be attributed to Douglas Engelbart.

Håkan Lans has also designed the communication method STDMA, which is used within AIS and VDL Mode 4 for maritime and air. AIS is a general system, which indicates a ship’s position to other ships in the vicinity, in order to increase the safety of navigation. The VDL Mode 4 standard is used as a substitute for secondary radar in aircraft with equivalent function as AIS. These systems have been adopted as world standards, first for shipping and later also for air traffic.

1 Like

I hope you are laughing when you typed this in jest, as this has shown to be absolute and utter b#llocks.

That is the argument used by so called libertarians to justify screwing people over and nothing more. Many private enteprise that can cut corners will cut corners to increase their margins as has been show repeatedly. Hence the desire for the worst polluters to spend a fortune trying to stop action on the environment, on climate, on product safety. You name it, they have done it.

Could “we” please leave this topic / thread?

It’s going nowhere helpful or productive,
while the OP’s question has been answered since long.

1 Like

Good point. I’ll delete my comments.

Here is something potentially useful for protection that might fit this threat.

What Is Glaze? Samples, Why Does It Work, and Limitations

Glaze is a system designed to protect human artists by disrupting style mimicry. At a high level, Glaze works by understanding the AI models that are training on human art, and using machine learning algorithms, computing a set of minimal changes to artworks, such that it appears unchanged to human eyes, but appears to AI models like a dramatically different art style. For example, human eyes might find a glazed charcoal portrait with a realism style to be unchanged, but an AI model might see the glazed version as a modern abstract style, a la Jackson Pollock. So when someone then prompts the model to generate art mimicking the charcoal artist, they will get something quite different from what they expected.

I could refute, again, every one of your points, but I value my time more than being right. It boils down to this. You don’t mind. Clearly many others do. Therefore they should seek permission. End.