On Wednesday, Midjourney detected “botnet-like” activity that appeared to be Stability employees attempting to collect prompt and image pairs in bulk, resulting in all employees at image synthesis rival Stability AI being suspended by the company. banned from the service indefinitely. Midjourney supporter Nick St. Pierre tweeted about the announcement, which was made through Midjourney's official Discord channel.
Prompts are written instructions that generative AI models like Midjourney and Stability AI's Stable Diffusion 3 (SD3) use to synthesize images (for example, “Cat in car with beer can”). Prompt-image pairs can be useful for training and fine-tuning competing AI image generation models.
Bot activity that occurred around midnight on March 2nd caused a commercial image generation service to be shut down for 24 hours. Midjourney has linked multiple paid accounts with employees of the Stability AI data team who are looking to “obtain prompt and image pairs.” Midjourney then made the decision to indefinitely ban all Stability AI employees from the service. The company also announced a new policy stating that “aggressive automation or service suspension will result in the expulsion of all responsible employees of the company.”
Mary Sue's Siobhan Ball said that companies like Midjourney, which built AI image synthesis models using training data collected from the internet without permission, are worried that their material will be collected. He said it was ironic that people were so sensitive to this. “It turns out that generative AI companies don't like stealing or scrapping images. Cue the world's smallest violin.”
For a monthly subscription fee, Midjourney users get access to an AI image generator that turns written prompts into gorgeous computer-generated images. The bots that create these were trained on millions of works of art created by humans. This is an act that is claimed to be disrespectful to the artist. “Words cannot express how inhumane it is to see my name used over 20,000 times on Midjourney,” artist Jinna Chan wrote in her recent viral tweet. “My life's work and who I am has been reduced to meaningless fodder for a commercial image slot machine.”
stability responds
Immediately after news of the ban broke, Stability AI CEO Emad Mostak said he was looking into it and insisted that whatever happened was unintentional. He also said it would be great if Midjourney contacted him directly. In a response to X, Midjourney CEO David Holtz wrote, “We have sent information that will be helpful to our internal investigation.”
In a text message exchange with Ars Technica, Mostaque said, “I checked and there were no scraped images there, but there was a bot run by a team member who was collecting prompts for a personal project. .I don't know how that will turn out.'' This will cause the gallery site to go down, and we apologize if that happens. Midjourney is amazing. ”
And Mostak says his company doesn't need midjourney data anyway. “We have been using synthetic and other data considering that SD3 is superior to all other models,” he wrote in his X. In a conversation with Ars, Mostak similarly wanted to compare his company's data collection technology to that of its rivals. “We only scrape the ones that have a proper robots.txt and are permissive,” says Mostaque. “And I also completely opted out. [Stable Diffusion 3] and Stable Cascade, which leverages the work done by Spawning. ”
When asked about the recent relationship between Stability and Midjourney, Mostak downplayed the rivalry. “There's really no overlap, but we're doing well,” he told Ars, highlighting the important connection in their history. “I funded Midjourney and [them] Can be covered by cash grants [Nvidia] A100 is a beta version. ”