Important Takeaways
- Midjourney is introducing a web application for generative AI picture modification, with initial access restrictions.
- To reduce risks, the business will implement sophisticated AI moderators in addition to improving human moderation.
- The distinction between fully AI-generated content and AI-modified photographs and monitoring alterations is still up for debate.
In order to reduce risks, Midjourney will soon introduce an online application for generative AI image manipulation, increasing user capabilities while restricting initial access.
On Midjourney’s official Discord server, CEO David Holtz declared that the picture editor will be made available “early next week.” Using Midjourney’s AI, the program will allow users to retexture photographs while preserving their shapes and altering details and colors in response to cues. With this change, Midjourney may now allow users to alter pre-existing photographs in addition to producing AI-generated art.
Midjourney will first restrict access to a “subset of the current community” due to the enhanced tool’s potent capabilities. To help stop harassment, Holtz said the company will implement “new, more advanced AI moderators” and improve human moderating.
Holtz stated that the business plans to publish a survey to get user comments and recognized that there is some hesitancy about limiting access to the new function. The results will be utilized to determine the proportion of the user base that will be impacted by various limitations.
Deepfake Risks and Regulations
Trends at Google and Adobe, which have also released comparable editing tools, are reflected in these sophisticated AI functions. But this makes it difficult to tell the difference between photos created from scratch and those altered by AI. For example, Meta has had trouble labeling photos that have been altered by AI, highlighting continuous discussions about tracking these alterations.
Midjourney has not accepted C2PA, which monitors an image’s complete provenance, but last year it adopted the IPTC’s Digital Source Type feature for embedding metadata in AI-generated photos.
There are risks associated with launching editing tools without the necessary protections, such as widespread copyright infringement and the spread of deceptive deepfakes. Businesses and political organizations are becoming increasingly concerned as deepfakes spread on social media, making it harder to discern between fact and misinformation. Given the upcoming US elections, this trend is especially concerning because deepfakes have the potential to sway public opinion and mislead voters, thereby threatening democracy.
Although there is no federal legislation that prohibits deepfakes, 19 US states have passed laws governing the use of generative AI in election communications, and seven more are debating similar legislation.
A service that lets users create AI graphics online without a Discord account was introduced by Midjourney in August. Users can explore sections, make graphics using text prompts, and register with Google or Discord. Following a complimentary trial, subscription plans are offered.
