Last Updated: 9 November 2024
According to TechCrunch, Midjourney is set to launch an enhanced web tool early next week that will allow users to modify any image uploaded from the internet using its advanced generative AI capabilities. Announced by Midjourney's CEO, David Holtz, this new feature enables users to alter the texture, colors, and details of objects in images based on provided captions, making image editing more intuitive and powerful.
The conversation around editing existing images with AI has been heating up. While platforms like Meta are grappling with how to differentiate between images altered by AI tools and those created entirely from scratch using AI, companies like Google have introduced robust AI features that offer no visual cues to indicate AI modifications.
Last year, Midjourney committed to incorporating the IPTC’s Digital Source Type property. This technical standard embeds metadata into images to signify they've been generated by AI.
Despite this, Midjourney remains one of the few major AI platforms that hasn't adopted C2PA, a metadata technology that traces an image's complete history, including the equipment and software involved in its creation.
In a message on Midjourney’s official Discord server, Holtz mentioned that the upgraded image tool will initially be available to a "subset of the current community." This rollout will feature heightened human moderation alongside "new, more advanced AI moderators" to help prevent misuse.
“Honestly, we’re not sure how to precisely restrict deployment of this feature”
David Holtz, CEO & Co-Founder, MidJourney
Midjourney is seeking community feedback through a poll to decide which users will gain early access.
There are inherent risks in releasing editing tools without proper safeguards. Such tools could potentially lead to widespread copyright violations or contribute to the dissemination of misleading deepfakes.
Deepfakes are increasingly prevalent across social media, blurring the lines between truth and disinformation.
Clarity, a deepfake detection firm, reported a 900% increase in deepfakes created and shared this year compared to the same period last year. This surge is understandably causing concern; a recent YouGov survey revealed that 85% of Americans are worried about deceptive deepfakes spreading online.
Although there is no federal law against deepfakes in the U.S., over ten states have enacted legislation targeting AI-assisted impersonation.
California's proposed law—currently stalled—would be the first to grant judges the power to order the removal of deepfakes or impose financial penalties on those responsible for posting them.
Midjourney’s track record regarding responsible AI deployment has not been spotless. The platform is currently facing a lawsuit over alleged use of copyrighted content to train its generative AI models. However, in recent months, Midjourney has made efforts to curb the spread of deepfakes, including implementing filters for political figures ahead of the U.S. presidential election.
