AI-Generated Content and Photography? Love Story?
On a basic level, Photoshop and Pixlr are amongst the many companies already deploying AI to speed up repetitive photo editing tasks and mitigate human error. Colorisation is one tool that used to involve painstakingly “painting” black and white images. Speaking to The Guardian in 2020, Artist Marina Amaral revealed she spent “hours, days or weeks working on a single photo” for her book The World Aflame — The Long War 1914–1945. The impact is emotive, especially in the depiction of children. When she coloured a photograph of a 14 year-old prisoner at Auschwitz, she revealed that her lips were bleeding — a detail only revealed with colour and testament to the brutality this poor child suffered. The colour image does not replace the original, Amaral adds, but “give[s] people the opportunity to see history from a different perspective.”
Now machine learning can carry out the task in minutes. Even DNA matching platform myheritage offers a colorisation tool. You can depixelate an image; do a cut out or deploy a plethora of retouching, correcting and adjusting tools — all using AI and all for a fraction of the time and ability these tasks used to demand.
This democratises photography, offering tools to the masses that used to be only available to and used by professionals. “It streamlines the editing experience and allows everyday users to become expert photo editors,” says William Wang, Product Manager at smartphone-maker Infinix.
Image manipulation for more nefarious purposes has been used for decades, especially for political gain. Josef Stalin, Mao Zedong and Benito Mussolini all famously removed those who had fallen out of favour from images. By 2008, the campaign for the US Election was absolutely littered with altered images — everything from Barack Obama with a cigarette hanging from his lip to Sarah Palin posing in an American flag bikini with a large gun. This is satire, no doubt, but there have been many cases where the objective was misinformation or humiliation. The boundaries are blurry.
AI is also supporting the creation of wholly generated photos. There are scores of websites offering “random human faces” — many powered by the publicly available StyleGAN, a neural network from Nvidia. The team at https://this-person-does-not-exist.com revealed though that its creation of photographs of non-existent people was actually “a by-product: the main goal was to train the AI to recognize fake faces and faces in general.” The faces they create are already persuasively human. It takes AI to detect the fakery.
With hyperreal synthetic media, could this become impossible? Who then will have the responsibility for signposting what is real in an image and what is not? The technology is here; and becoming increasingly accessible. It will change the world of photography and the only way to unravel conundrums of authenticity; ownership and more is to embrace the technology now and explore its full potential.