Adobe Stock Guidelines for AI Imagery

 

I received an email last week from Adobe's stock division where I have in the past placed a number of my photos.  The email in question laid out guidelines for submitting AI generated imagery.  While they were no doubt meant to be helpful, I found them to be more confusing than anything else.  I had no problem with the requirement to submit a model release if a given image was intended to represent an actual person (presumably still living since it would be rather difficult to obtain a signature from a deceased individual), but the next requirement definitely left me scratching my head.  It read:

"If the generative AI content was not based on a real person, but it visually appears to resemble a person, then you must submit a property release confirming that  you have all the property rights in the content."

The question here is who is to sign the release since there is no actual person or property being shown in the generated image?  I'm really not sure and considering how little money is involved in stock photography I don't think it's worth retaining an attorney to get a legal opinion.  Better to simply submit actual photos, as I have in the past, that do not require any form of release, such as photos of flowers.  Much as I enjoy working with AI imagery, I'd rather do that than expose myself to any form of legal liability.

The Times Issues Warning on AI

 

By now I'm used to reading in the tabloid press dire warnings of AI wiping out humanity if left unchecked, but it's on a whole new level when an article in the staid UK publication The Times adds its voice to the chorus.  The article is actually a report on a briefing given Parliament's science and technology committee by researchers from Oxford University on the potential dangers posed by AI if not regulated and tightly overseen.  One doctoral student, Michael Cohen, warned bluntly of AI's existential threat to humanity:

"With superhuman AI there is a particular risk that is of a different sort of class, which is... it could kill everyone."

Scary stuff all right, but it must be remembered that there is no proof that such a threat actually exists.  Right now it's no more than a frightening theory.

Shutterstock AI Imaging

 

Shutterstock had in the past prohibited creators from uploading AI generated images to its platform.  Whether or not that still holds true, the stock agency has partnered with Open AI, developer of DALL-E 2, to introduce its own AI image generator that will allow anyone with an account to generate images and submit them for licensing without ever leaving the site.

I have actually never submitted any of my stock photos to Shutterstock, though I do have an account, preferring to limit my submissions to Adobe Stock if only for convenience sake.  Although I have sold some images from the Adobe site there is actually so little money in stock photography - and yes, I've read those accounts of photographers who earn six-digit incomes from it but have never been able to figure out how they manage it unless they work at it night and day without ever sleeping - that it never seemed worth my while to make an all out effort and begin submitting to  multiple venues.  Given my interest in AI imaging, however, the opportunity to make use of Shutterstock's own generator seems too good an opportunity to pass up; and I most likely will start using it in the near future once I finish working on my current project (a novel fully illustrated with AI imagery) and then license the images I derive therefrom on Shutterstock.  I don't think there's much to lose, though I will first certainly read the fine print to determine exactly what rights I'm signing over to Shutterstock by so doing.

Great Service from On1 Tech Support

 

A big shout or to On 1's tech support team for the excellent service they've provided me.  Briefly, I had recently purchased from Clever Photographer their bundle of Landscape Photograph Overlays (of which I'll write more in a subsequent post), but when I went to import the files into Photo RAW 2023 using Manage Extras from the File menu the app repeatedly crashed.  Rather than brushing me off on the excuse I was using some other company's product, On 1's tech support stuck with me through a series of emails in which they suggested various remedies.  When none of those worked they requested I forward to them samples of the files that had caused the crash.  After analyzing them, they informed me the problem was that some of the files were in CMYK color mode rather than RGB and that it was this that had caused the app to crash.  This was something I never would have figured out on my own and I'm very grateful to tech support for having gone way beyond the call of duty to help solve my problem.

Microsoft Invests in Open AI

Older companies often react more slowly to innovation than those that have only recently launched their businesses.  But not Microsoft, at least not in this case.  The venerable behemoth that helped launch the original personal computer revolution has obviously grasped the enormous potential offered by AI and is investing heavily in Open AI, creator not only of DALL-E 2, one of the web's premiere AI imaging apps, but also of ChatGPT, the app that has created an internet craze with its ability to hold informative conversations with users in response to their questions rather than simply provide links as do most traditional search engines.  According to a recent ABC news article, Microsoft is making a "multiyear, multibillion dollar investment" in the AI startup.  This is in addition to the one billion MS previously invested in the company in 2019.  Not only will MS now be able to incorporate AI features into MS Office and its other products, but the investment has the potential to make the company's Bing search engine far more competitive with Google Chrome than is currently the case.

As if to underline the potential offered by ChatGPT a recent article in the New York Post headlined a news item "AI Bot ChatGPT outperforms students on Wharton MBA exam: professor."  The professor in question, one Christian Terwiesch had made the claim in a paper referenced by Financial Times, and he displayed a marked gift for understatement when he went on to write, "This has important implications for business school education."

Backlash against AI Imaging

 

There was recently an interesting article in Japan Today, the English language news site, regarding the growing backlash among traditional artists against the new wildly popular AI imaging apps, such as Midjourney, they see as a threat to their own livelihood.  The protest has taken the form of lawsuits filed in both the UK and the US by disgruntled artists.  The US filing complained that AI images "compete in the marketplace with the original images.  Until now, when a purchaser seeks a new image 'in the style' of a given artist, they must pay to commission or license an original image from that artist."  The basis of the suit is therefore an argument against the misappropriation of images without due recompense to the artists who created them and in that sense is very similar to the lawsuit recently filed by Getty Images against Stable Diffusion.

While it's entirely understandable that traditional artists should feel entitled to payment when their images are "scraped" without their permission from the internet to become part of an AI imaging app's dataset, I think there's much more to the current backlash than just the monetary damages to the artists' income.  To me, the situation is reminiscent of the controversy that erupted in the nineteenth century following the invention of traditional photography when it was questioned whether the new medium could truly be regarded as an art.  Of course, photography was eventually accorded it's place in the arts and I feel fairly certain that the same will hold true for AI imaging once users discover the medium's ability to create new forms rather than simply copy the conventions of traditional art.

Lunar New Year

 

I hope everyone who observes the lunar new year, most particularly my many Asian friends, will have good health and great good luck in the Year of the Rabbit.

The above AI image was generated by Stable Diffusion.

January On1 Plus Sale

 

I upgraded several months ago to On1 Photo Raw 2023 and have been very pleased with my purchase, especially as I had not been impressed at all with the 2019 version I had previously owned.  On1 has really gotten its act together with the newest version to the point where it may take the place of Luminar Neo as my preferred editing app.  (I still use DXO PhotoLab 6 as my go-to RAW processor and it's from there that I extract the files I then edit in either Photo RAW or Neo.)

Through the end of January, On1 is now offering a 25% discount on its On1 Plus program that includes, among other perks, access to a creative library, a video library, 40+ full length video courses, a bookshelf, and bonus loyalty rewards.  For my own needs, it was the loyalty rewards and access to the creative library with its presets, textures, and other add-ons that finally tempted me to join.  I eventually chose the one-year access option for $44,99 + tax, reduced from $59.99, and will now determine how well the plan works for me over the course of the next twelve months.

Nikon Z 26mm Pancake Lens

Nikon recently announced in an Instagram post that it is currently developing a 26mm f2.8 "pancake" lens for its Z mount cameras.  When mounted on my DX Zfc camera that would be the equivalent of 39mm, a focal length that neatly straddles the wide angle and normal lenses' angles of view.  Adding such a lens to my kit would render my already lightweight Zfc even more compact and even easier to carry with me around the city.  There's no word yet on the price, but if it turns out to be reasonable I may very well end up purchasing the lens.

Getty Images Sues Stability AI

 

Getty Images has brought suit in London against Stability AI, creator of the wildly popular AI image generator Stable Diffusion alleging that the company "copied and processed millions of images protected by copyright."  Part of the problem apparently stems from the fact that Stable Diffusion's training dataset is open source, and therefore there are no safeguards in place that would have prevented a large number of stock photos from having been incorporated into the dataset.  If the accusations prove to be true - and an image posted online that was generated by Stable Diffusion containing the Getty logo certainly gives credence to them - Stability AI may be required to pay huge sums in damages for failure to properly license Getty's images.

Nor is this the only lawsuit Stability AI faces.  Attorney Matthew Butterick has filed a class action lawsuit in US District Court for Northern California, San Francisco Division on behalf of three artists who allege their work has been misappropriated by Stability AI, Deviant Art (creator of DreamUp), and Midjourney.

Both these cases will be closely watched as their disposition must inevitably affect the future of AI generated imagery.

AI Image Copyright

Controversy erupted last year over the copyrighting of images generated by AI; and judging from the popularity of apps such as DALL-E 2 and Stable Diffusion, the dispute is not likely to go away anytime soon.  Briefly, in August 2022 the US Copyright Office, in response to an appeal of a 2019 ruling rejecting a request submitted by one Steven Thaler who had invented an AI imaging app he named Creativity Machine, handed down a decision holding that AI generated images cannot be copyrighted because no human agency is involved in their creation.  Thaler has now taken his case to court and has filed suit with the US District Court for Washington DC in which he basically argues that the Copyright Office's decision should be overturned because it is based on outdated precedents and that copyright law "does not restrict copyright to human-made works, nor does any case law."

The question in my mind is whether the creation of AI generated images does not in fact involve at least some form of human agency since it is obviously a human who initially prompts the AI in order to obtain from it the desired image.  In this respect, it should be noted that Thaler in his initial filing listed Creativity Machine as the creator of the work and listed himself only as the claimant.  I'm  not sure I agree with this since it seems somewhat akin to listing my camera as the creator of my work when I go to register photographs.  The bottom line to me is that unless it has achieved singularity - and neither Thaler nor anyone else is going that far - an AI app can no more generate images entirely on its own any more than one of my Nikons can take photos without my first having pressed the shutter button.

In any event, it will be interesting to see how the court case eventually plays out.

DALL-E Inventor Aditya Ramesh

 

I recently came across an interesting interview in Venture Beat with Aditya Ramesh, the inventor of Open AI's DALL-E.  In it, Ramesh claims to have been surprised by the remarkable popularity of his extraordinary imaging app, and I don't think anyone can blame him.  A mere two years ago the very concept of an AI imaging app was unheard of, and there was no way of knowing what a sensation DALL-E would become when once released.  Still, the promise it held out was irresistible - that one had only to type in a text description in order to see an pictorial representation generated almost before one's eyes.  No wonder so many, including myself, rushed to put their email addresses on the waiting list in order to be among the first to use it.  And no wonder Open AI is now considering a tender offer that would place its value at $29 billion.

Of most interest to me was Ramesh's description of how images are generated.  He noted that images are generated from scratch, beginning with a blurry outline, rather than by cut & pasting from the huge inventory of scraped images in its database.  This was of particular importance to me since I strongly believe in respecting other artists' intellectual property rights just as I would want my own to be.  I  feel I can rest easier now knowing that I am not inadvertently appropriating anyone else's images or parts thereof.

As far as DALL-E 2 itself is concerned, I've worked with it and admire what it can do, but I've actually found the images generated by Stability AI's Stable Diffusion to be more aesthetically pleasing.  I realize, of course, that this is a very subjective choice but I nevertheless find myself working far more often with Stable Diffusion than with DALL-E 2.

Luminar Neo Face Recovery

 

In recent months Skylum has begun releasing a series of extensions to Luminar Neo that are essentially add-ons enabling optional functions within the app.  As previously reported, I've purchased several of these extensions and found them quite useful in my work.

While recently browsing blog posts and You Tube videos, I've found a bit of controversy has erupted over two of these Luminar extensions - Upscale (which I do not own) and Supersharp (which I do have installed on my computer), both of which feature face recovery options.  The problem seems to be that the extensions do not actually enhance facial features as advertised but instead replace them with features taken from other faces.  This seems a somewhat bizarre claim, but after having seen examples posted on the web I would have to agree that this is in fact exactly what the extensions are doing.  While the final results are generally quite attractive, the faces themselves have been altered to the point that their features no longer fully correspond to their real life counterparts, though of course a close resemblance still remains.

To me, none of this is really a problem since I rarely photograph portraits these days and in any event have other apps, such as Topaz Labs Photo AI, that perform face recovery quite well.  I can understand, however, why professional portrait photographers and even casual users may be upset with the results of their Neo edits.  On the other hand, some subjects may actually prefer these stylized versions.  When in the past I did shoot portraits I found my clients almost always preferred an idealized and more flattering version of themselves to an exact depiction.

Phase Detection Autofocus

 

Panasonic has just released two new full frame mirrorless cameras, the S5II and S5IIX, and though I have no plans to buy either one I was intrigued by one new feature they share - Phase Detection Autofocus.  To be honest, I had never previously heard of PDAF and had to check online to understand exactly what it offers.  On the most basic level, according to a Wikipedia article, "Phase detection (PD) is achieved by dividing the incoming light into pairs and comparing them."  According to its news release, Panasonic has taken PDAF a step further in its new cameras by introducing what it terms "Phase Hybrid Autofocus" that increases the number of AF points on the sensor and thus improves on the cameras' tracking ability.

PDAF, as implemented by Panasonic, appears a valuable new tool for its camera customers, and I am curious to see if it is included in the MFT cameras the company will release in the future.  In spite of firmware updates, the tracking ability of my Lumix G9 has always been somewhat problematic.

Happy New Year!!

I hope everyone has a wonderful New Year in 2023 and that all your dreams and wishes come true.  My own resolution is to finish writing my novel And What If, one of the first if not the very first work of fiction to be fully illustrated with AI generated imagery.