Distortion of history

Experts deplore “tsunami” of fake photos

"Milestone in early aviation: Wright Brothers achieve flight success" is the description of this AI-generated photo, which is available via Adobe's image marketplace.

(Bild: stock.adobe.com/john (KI-generiert))

A crying mother and child during the Great Depression, a soldier in the Vietnam War - images like these can be seen in pictures currently circulating on online networks. At first glance, they appear to capture real historical moments, but they have been created using artificial intelligence. Experts fear that this could distort history.

For example, a picture circulating on Facebook purports to show the Wright brothers, two pioneers of aviation, after their first powered flight in 1903. A smiling pair of brothers with blonde hair can be seen posing in front of an airplane.

In contrast, real archive photos show two men in slouch hats who do not resemble the blonde couple. The atmosphere of the photographs is also completely different. In the original photos, they are either depicted with serious expressions or are merely seen as distant silhouettes working on the prototype of their airplane.

"Tsunami of fake history"
Why was this fake photo produced? The creator of the fake photo did not respond to an inquiry from the AFP news agency. However, the image did reveal which prompt - i.e. which instruction - he had entered into the AI program. According to this, a "festive" atmosphere was to be created.

"AI has caused a tsunami of fake history, especially in the field of images," laments Dutch historian Jo Hedwig Teeuwissen. Under the pseudonym "Fake History Hunter", she tracks down false claims or prejudices about historical developments in online networks.

"In some cases, these are AI reproductions of actually existing archive photos. That's really strange, especially when the originals are very well known," the scientist marvels. For example, the popular AI image program Midjourney has been used to create forgeries that mimic famous photographs - such as the killing of Lee Harvey Oswald, the alleged assassin of John F. Kennedy, in 1963.

There are also midjourney images that purport to show the explosion of the Hiroshima atomic bomb on August 6, 1945, the invasion of Prague by Warsaw Pact troops in August 1968 or the Colosseum in ancient Rome - none of which are real.

Marina Amaral also criticizes the trend of these fake archive photos generated by AI. The Brazilian artist specializes in colouring old photos. She is concerned about the staging of "events that happened too long ago to be photographed" or moments in history that are "poorly documented".

"The risk is that these false images will be seen as historical facts and that over time this will change our understanding of history and weaken the public's trust in (the image) as a reliable source for studying the past," Amaral believes.

Still to be distinguished
Marina Amaral and Jo Hedwig Teeuwissen are still convinced that, thanks to their expertise and certain errors, they are able to distinguish the real from the fake. According to their findings, AI-generated photos often still have telltale flaws: too many fingers on one hand, missing details - such as the lack of a propeller on the Wright Brothers' plane - or, conversely, an image composition that is too perfect.

Amaral believes that AI cannot imitate certain characteristics of real photographs, such as snapshots of a moment full of emotion. "They lack the human element, the intention, the reason for the photographer's decisions. Although they are visually convincing, these images remain hollow," says the artist.

"Dangerous" view
For Teeuwissen, "real photos are taken of real people and usually something is out of focus, or someone accidentally looks silly, the make-up looks bad and so on." However, the historian estimates that it is "only a matter of time" before the quality of AI images makes forgeries difficult to spot with the naked eye. This is a "dangerous" prospect that threatens to strengthen the power of disinformation.