In recent days, false images of former President Donald Trump mistreated in a demonstration, the Pope in a modern white puffer jacket, Joe Biden dressed as a clown or even Emmanuel Macron as a garbage collector had circulated online. It had all been generated by Midjourney’s artificial intelligence.
American journalist Eliot Higgins, who started the trend, has since been banned from the software. Midjourney has also ended its free trials.
But not all leaders are entitled to the same treatment. For example, if he tries to fabricate an image of Xi Jinping or the Chinese president, chances are he won’t succeed the first time. Midjourney simply banned the creation of it.
“Avoid the drama”
The declared objective: “Avoid drama”, says the CEO of Midjourney, in a message on the Discord platform, broadcast by the Washington Post, “because political satire in China is pretty frowned upon.” The company may simply fear legal repercussions, as long as it remains a small structure, as 01Net suggests.
These fake shots can also be used for political and propaganda purposes, while it is still very difficult to tell a real shot from an AI image.
“The images are more and more realistic and the tools are more and more powerful,” acknowledges the founder of Midjourney. He says that he is working to better detect “objectionable behavior” in the use of his AI, without elaborating. The software no longer allows images that are overly violent and text that is “inherently disrespectful, aggressive, or abusive.”
If its competitor Stable Diffusion sets few limitations, the generative AI Dall-E for its part forbids depicting politicians or major geopolitical events currently underway.
In the case of Midjourney, this censorship does not go down well with its users, who have already found ways around it by using other words to describe the Chinese president. A guide posted on a Taiwanese site even details how to “mock” Xi Jinping using AI.
Source: BFM TV
