Last month, an amateur photographer with the pseudonym "Ibreakphotos" decided to perform an experiment on his Samsung smartphone to find out how a feature called "spatial zoom", launched in 2020, which claims a zoom rate of 100x actually works.
The South Korean manufacturer has used images of the Moon of sparkling clarity to promote this feature. "Ibreakphotos" took his own pictures of the star, blurry and without details, letting his phone add craters.
The built-in AI software used data from its "training" on many other photos of the Moon to add details where there were none.
"The photos of the Moon taken by Samsung are fake," the amateur photographer concluded on the Reddit forum, leading many people to wonder if the shots they take are really their own and if they can even be considered photographs.
"I wouldn't say I'm happy with the use of AI in cameras, but I'm okay as long as it's clear what each processing channel actually does," Ibreakphotos told AFP, asking not to use its real name.
Samsung has defended its technology, saying it doesn't "overlay" images and users can turn off the feature.
The company isn't alone in the race to integrate AI into its smartphone cameras: Google's Pixel devices and Apple's iPhone have been marketing such features since 2016.
Not "man-made"
But the AI debate isn't limited to amateurs on discussion forums: professional bodies are sounding the alarm.
According to Michael Pritchard of the British Royal Society of Photography, the industry is awash with AI, whether it's cameras or software like Photoshop. "This automation is increasingly blurring the lines between a photograph and a work of art," he told AFP.
The nature of AI is different from previous innovations because the technology can learn and bring new elements beyond those recorded by film or sensor. This opens up perspectives but also "fundamental challenges to redefine what photography is and how a photograph is +real+", he adds.
Professional photographers are concerned about the rise of AI tools that generate entirely new images from text, such as DALL-E 2 or Midjourney © Lionel BONAVENTURE / AFP / Archives
What concerns professional photographers the most, however, is the rise of AI tools that generate entirely new images from text, such as DALL-E 2, Midjourney (recently used for a viral photo of the pope in a down jacket) and Stable Diffusion.
"This is not an author's work" and, "in many cases it is based on the use of training datasets of unlicensed works," notes Nick Dunmur, a member of the British Association of Photographers, while the phenomenon has already given rise to lawsuits in the United States and Europe.
Jos Avery, an American amateur photographer who recently tricked thousands of people by posting stunning portraits on Instagram... with Midjourney, disagrees, pointing out that the creation of his images often took many hours.
"AI will not lead to the death of photography," he said. A prediction shared by Michael Pritchard, who reminds us that photography has endured from the daguerreotype to the digital age and that photographers have always met technical challenges.
© 2023 AFP