Midjourney Set to Compete With Adobe With a New Version of Its AI
Midjourney is soon launching a new tool for editing images online with AI - a technology that could change large parts of how we create images. But what risks come with deepfakes and image manipulation?
Midjourney is preparing a new version of its web-based tool that will let users edit uploaded images directly online using the company's generative AI technology. According to Midjourney CEO David Holz, the tool will launch "early next week" and includes features for "retexturing" and "repainting" objects in images, letting users change colors and details based on descriptive text. It'll be interesting to see whether this could become a clear competitor to Adobe and its AI model Firefly.
Risks and the Need for Clear Labeling of AI-Edited Images
Using AI tools to edit existing images has become a hot topic. Platforms like Meta are struggling to clearly label images edited with AI, while companies like Google are launching powerful AI tools with no visual indicators for AI modifications. In a step toward greater transparency, Midjourney has implemented the IPTC Digital Source Type standard, which shows metadata on whether an image was created with AI. The company hasn't adopted the C2PA standard, though, which tracks an image's full origin, including hardware and software used.
Limited Rollout and Increased Moderation
To ensure a responsible rollout, the new version of the image tool will be released in a limited edition, initially available to a specific subset of users. Midjourney also plans to introduce increased human moderation as well as AI moderation to prevent misuse of the tool. The company will use a user poll to decide who gets initial access.
Challenges With Deepfakes and AI-Generated Images
The growing availability of AI editing tools, especially for creating deepfakes, brings significant risks. Over the past year, deepfake technology has risen by as much as 900% according to data from Clarity, a company specializing in identifying deepfakes. In the wake of disasters, like Hurricane Helene, the internet has been flooded with manipulated images depicting destruction and suffering, making it hard for the public to tell reality apart from disinformation.
Deepfakes have become such a serious issue that over ten US states have now introduced laws to combat AI-assisted fraud. Proposals for federal laws are, however, still lacking. In California, a bill is currently being considered that could give judges the ability to order the removal of deepfakes or impose fines on creators of misleading content.
Midjourney's Steps Toward Responsible AI Development
Despite past criticism, including an ongoing lawsuit over the alleged use of copyrighted material in its AI models, Midjourney has introduced certain restrictions to prevent the spread of deepfakes. Ahead of the upcoming US presidential election, the platform has added filters to prevent manipulation of images of political leaders.