Too much demand, concern about the influx of disinformation generated by its tool? Maybe a bit of both. Midjourney, the research laboratory that offers an eponymous image generation tool, has closed access to its AI to those who used it for free. You must now subscribe to continue generating images from text commands.
A new, much more powerful version
“Due to a combination of extraordinary requests and trial abuse, we are temporarily disabling free trials until our next system enhancements are rolled out”, said David Holz, CEO of Midjourney, in a message posted on the Discord community of the company. If he does not detail what is behind the formula “trial abuse”the man had recently declared, according to the washington posthave difficulty ensuring compliance with the rules, “especially when it comes to portraying real people”. In other messages, the CEO explained that the company had “blocked a bunch of words related to topics in different countries”.
The recent decision to suspend free trials on Midjourney comes after the launch of version 5 of the Generative Image AI. A new version that promised a much wider stylistic range, a much higher image quality, more details, fewer anomalies, or even additional care given to textures. What attract new curious and explode the statistics of use of the platform.
Ultra-realistic (but misleading) images abound online
More concerning, Midjourney has recently been used for the dissemination of many ultra-realistic – but misleading – images on social networks. We have thus seen images of Emmanuel Macron in demonstration, or of Pope Francis equipped with an ultra-fashionable down jacket, go viral.
On Twitter, a post sparked a long debate over whether or not the image was generated by an AI. It showed an elderly protester, bleeding, surrounded by several members of the police. If certain clues allow to certify that this image is created from scratch, only the most seasoned in this field can notice it (absence of insignia of the forces of order, strange shape of the visor, etc.).
We found a first occurrence of this entire image in a Facebook post from March 19, which several elements suggest that it is an image generated by artificial intelligence #AFP 2/… pic.twitter.com/0twzrWLmXi
– AFP Factual ? (@AfpFactual) March 30, 2023
However, the use of these misleading images to spread false information is likely to give a hard time (and a few headaches) to those who specialize in checking the facts…