Midjourney intends to provide an enhanced online application that will enable users to utilise the generative AI of Midjourney to alter any submitted photographs from the internet.
Users will also be able to retexture items in photos to “repaint” their colours and details in accordance with captions using the updated tool, which Midjourney CEO David Holz said would be made available “early next week.”
AI picture editing has recently gained a lot of attention. While businesses like Google have introduced sophisticated AI capabilities that do not visually indicate that photographs have been altered by artificial intelligence, platforms like as Meta have struggled to distinguish between images created from scratch using an AI model and those enhanced using AI tools.
Midjourney pledged last year to use the IPTC’s Digital Source Type attribute, a technical standard that adds information to photos to identify them as artificial intelligence (AI)-generated.
Nonetheless, the business is among the few significant AI platforms that hasn’t used C2PA, a metadata technique that tracks the whole provenance of a picture, including the hardware and software that were used to produce it.
Holz said in a message on Midjourney’s official Discord channel that in order to try to minimise abuse, the updated picture tool would initially only be available to a “subset of the current community,” with both “new, more advanced AI moderators” and additional human moderating.
“To be honest, we don’t know exactly how to limit the use of this feature,” he said. In order to decide which people get access first, Midjourney is using a poll to seek community input.
May Also Like: ChatGPT: Everything you need to know about the AI-powered chatbot
If these kinds of editing tools are released without sufficient security measures, there are hazards involved. They may encourage the proliferation of deceptive deepfakes or enable widespread copyright violations.
On social media, deepfakes are catching on like wildfire and making it more difficult to discern between false information and the truth. Most recently, during Hurricane Helene, phoney generative AI pictures of devastation and human misery went viral online.
Data from the deepfake detection company Clarity shows that there have been 900% more deepfakes produced and released this year than there were during the same period last year. It is obviously alarming. According to a recent YouGov survey, 85% of Americans were worried about deceptive deepfakes becoming viral online.
Although there isn’t a federal legislation in the United States that criminalises deepfakes, over ten states have passed laws against AI-assisted impersonation. California’s proposal, which is now blocked, would be the first to provide courts the authority to force deepfake posters to be taken down or face financial penalties.
Using AI responsibly hasn’t exactly been shown by Midjourney. (For instance, it is being sued for allegedly using copyrighted material to train its generative AI models.) However, in the months before the U.S. presidential election, the platform has implemented measures to curb the proliferation of deepfakes, such as filters for political personalities.
