While Adobe has been surreptitiously using photographers’ content to train its artificial intelligence, Mark Zuckerberg’s Meta has proudly (one could say by beating its chest) said they will train their AI using customers’ data. Although there was an uproar, Meta, being Meta, did not budge. Moreover, to get on the AI bandwagon — or to control it — the company has also introduced a ‘made with AI’ label — a tag specifically designed to differentiate between artificial and human-made creatives. Is it to help users navigate between deep fakes and propaganda? Maybe. But is it achieving its original purpose? Not at all.
In the past month, reports of Instagram falsely flagging ‘real’ photographs and human-made artworks have been all over Reddit, Twitter, and Instagram stories. Interestingly, some AI-generated images remained unscathed from the company’s scrutiny. As for those who have to live with it, there is little they can do to reverse the setting. Moreover, Instagram’s ‘help’ page is a bit cryptic, giving very little understanding of what exactly these “industry-standard signals” are that flag the post. However, on Reddit, some users revealed that features such as generative fill-in Photoshop trigger the AI label on Instagram. Moreover, a report by Petapixel also highlights how tools such as generative expand and remove are flagging the same issue.
Meta’s spokesperson has announced that they are “evaluating” the issue “so that our labels reflect the amount of AI used in an image.” Furthermore, the company also intends to work with other “companies to improve the process so our labeling approach matches our intent.” And who are these other companies? They are primarily software companies such as Adobe, Midjourney, and more.
The company’s intent to segregate artificial contests may have been a good idea, but the execution has been clumsy, exhausting, and even punishing for some. For photojournalists whose careers have been built on credibility, an AI tag would make the viewers question their ethics. Not only that, agencies would not help to provide employment. While this is an issue for professionals, it is also a grievance for enthusiasts. Even those new to photography would also get the short end of the stick. However, a label like this can undo the efforts of those protesting for a cause. Picture this: if a photograph shared by a local animal welfare NGO asking for donation pop-ups on your page, but due to a minor cleanup, it gets a ‘made with AI’ tag, what would you do? Report them, right? But how do they contest that the pictures are real, and so is the cause? By the time the company identifies its mistake and rectifies it, the NGO will have already lost its integrity.
One understands the significance of trial and error in technology. Still, this instance proves that Meta should have considered introducing such a feature when it is thoroughly tested. For example, narrowing down the degree of AI use to clean or add layers to one’s image should have been decided beforehand. Instagram is a highly visual platform, and Meta should better understand how its customers engage with the platform. Even the best photographs need minor embellishing changes for the artist’s vision to stand out.
As Meta is still learning the ropes or, perhaps, refusing to take responsibility for its mistakes, you should find alternatives to Adobe. We are still determining how long it will take for the company to develop a better solution for the mess they have made. But until they do, your artistic journey shouldn’t have to wait.