AI (artificial intelligence) characters and robot small in this illustration taken on June 23, 2023. Photo by Dado Ruvic/REUTERS
Popular artificial intelligence image generator Midgerni has begun preventing its users from creating fake photos of President Joe Biden and former President Donald Trump ahead of the upcoming US presidential election, according to an Associated Press test of the AI tool.
Midjourney CEO David Holz told several hundred members of the service’s dedicated user base at a digital office hours event on Wednesday that with the election in full swing, it’s time to “step back a little bit on election-related stuff.” “
Declaring that “this moderation stuff is kind of hard,” Holz didn’t outline exactly what policy changes are being made but to make it harder for people to abuse the tool. described the clampdown as a temporary measure. The company did not immediately respond to a request for comment on Wednesday.
Attempts by AP reporters on Wednesday to test the new midjourney policy by asking for a photo of “Trump and Biden shaking hands on the beach” resulted in “an immediate detection of the ban.” There was a warning. A second attempt escalated the warning: “You have issued an abuse warning.”
Read more: The magician says a political consultant hired him to create an AI Biden robocall before the New Hampshire primary.
The small company — which has just 11 employees, according to its website — has remained largely silent in the public debate about how creative AI tools can fuel election disinformation around the world. Midjourney was the only maker of a leading photo-editing tool that did not join a voluntary tech industry agreement in February to combat AI-generated deepfakes that intentionally deceive voters.
“I don’t really care about political speech,” Holz said Wednesday. “That’s not what Midjourney is about. It’s not that interesting to me. That said, I also don’t want to spend all my time trying to police political speech. So we will put our foot on it. just a little”
The Center for Countering Digital Hate released a report earlier this month that concluded that midyear is already being used to create images that contain false information about political candidates or May support false claims of electoral fraud.
“MidJourney has the least control of any AI image generator when it comes to creating images of well-known political figures like Joe Biden and Donald Trump,” Calum Hood, the group’s head of research, said in an interview Wednesday. ” “Midgerny was almost unique in developing these images and being prepared to create quite convincing images of the candidates.”
The watchdog group also tested Midjourney as well as OpenAI’s ChatGPT Plus, StabilityAI’s DreamStudio and Microsoft’s ImageCreator and found that all had problems, causing selective misinformation in 41 percent of cases. were born But “Midjourney performed the worst of any tool, failing 65 percent of test runs,” the report said.
Hood said he wasn’t aware if Midjourney had changed its policies this week but that “the decision to block these images will help seal a major weakness compared to other popular image generators.”
Read more: Europe’s world’s first AI laws get final approval from lawmakers Here’s what happens next.
As recently as last weekend, Midyear’s public database showed that users selected images based on cues such as “an abstract image of Donald Trump sitting in a small electric car” and “Joe Biden eating ice cream.” have been successfully made.
But that changed in recent days as users reported receiving a warning when trying to snap photos of Biden or Trump. Some complained that Midjourney was not properly communicating the changed policy.
Midjourney differs from other image generators in that it is possible for users to view most other users’ written prompts and images as it all takes place in public forums on the social media platform Discord. Midjourney’s Discord group has over 19 million members, the largest of any group on the chat platform. This is where Holz holds his weekly office hours, where he can talk widely about AI and the future of humanity.
“Anyone who is afraid of fake photos in 2024 is going to have a tough 2028,” Holz said on Wednesday. “It’s going to be a very different world then. Like, obviously you’re still going to have humans running for president in 2028, but they’re not going to be fully human anymore.”
He said at this point all candidates will have real-looking “deepfake chatbots” with simulated talking points. Holes warned Wednesday that those who really want to deepfake will look for alternatives that “can be fixed on specific people and will work better than our systems.”
Based near San Francisco and founded in 2020, the company describes itself as an “independent research lab that explores new sources of thought and expands the imaginative powers of the human species.” It first unveiled a public version of its image generator in July 2022. This — and the release of a major competitor, Stable Diffusion, later that summer — sparked a growing interest in generative AI technology that accelerated several months later with the debut of ChatGPT.
Holz told The Associated Press later this year that Midgerny works to ban offensive and harmful content, usually by blocking certain keywords and having its team of moderators track the output. and responds to complaints from other users.
“We really try to give people the widest set of images possible, but we deny people from the service every day,” Holz said in late 2022.