Artificial intelligence image-generating models are evolving at an unprecedented pace. What last year was a nuisance for artists online, has become an eye-opening shock for mainstream publications.
Platforms like StableDiffusion and Midjourney, which only require a few YouTube tutorials worth of know-how to use, are now able to generate very convincing digital images of real people, with realistic lighting and now most chilling – hands with only five fingers.
Now, people are using these platforms to make fake images of politicians and political situations for the express purpose of misleading others.
On Monday, the highly publicised meeting between Russian President Vladimir Putin and Chinese leader Xi Jinping led to the spread of a fake image of Putin kneeling before Jinping. An Italian-language publication analysed it to show that the image was in fact created by AI software.
While the image is clearly fake, that didn’t stop a Kyiv Post journalist from commenting on it. After this, it quickly went viral and was circulated on social media with comments in several different languages.
Another (former) world leader who went viral on Twitter this week was Donald Trump, who had falsified images of his purported arrest spread across social media. This came after the former US president made a public statement on the weekend that he would soon be arrested in a criminal case involving hush money payments to adult film actress Stormy Daniels.
Some of these images were being used to spread fake news as #TrumpArrest swells in popularity on Twitter, even though Trump is still very much free as of time of writing. One, in particular, was incredibly realistic, as seen in the tweet below:
One has to look very closely to see inconsistencies between the hands of Trump and the police officers, as well as the fine details of the badges on the police officers’ shoulder pads to see that the image is AI-generated. The objects in the background are also a dead giveaway.
Many of the fake Trump arrest images seemed to be created using Midjourney.
Many of these fake images were being shared on TikTok, which took immediate action and launched a set of new restrictions for the sharing of AI-generated images of real people. Now, if users want to upload fake AI images, they will have to explicitly state that they are fakes.
“It takes a whole village to keep people safe online, so we’re grateful to everyone in the TikTok community and to all of the external experts who have contributed and continue to help us advance our rules and stay a step ahead of emerging threats,” TikTok global head of Product Policy Julie de Bailliencourt wrote in a press statement about the new rules.
Wary users are right to be concerned about the use of AI-generated images to spread fake news, especially as platforms like Midjourney, StableDiffusion and OpenAI’s ChatGPT become more powerful and better at creating realistic and convincing material.
It isn’t a stretch to say that these platforms will soon be able to quickly and easily create near-authentic images of real people in complex situations. And some will use them to spread fake news and misinformation.
It will be up to informed users to ensure we keep an open eye and a critical mind. It will only get murkier, and the line between real and false grows thinner.
[Image – Charlie Warzel on Twitter]
