Meta works to identify and label AI-generated images on Facebook, Instagram, and Threads, and is striving to expose “people and organizations that actively seek to deceive the public.” Masu.
Images created using Meta’s AI image tools are already labeled as AI, but Nick Clegg, the company’s global president, stated in a blog post on Tuesday that the company’s competing services will start labeling AI-generated images.
Meta’s AI images already have metadata and an invisible watermark indicating that the image was created by AI. The company has partnered with Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock to work on AI image generators, according to Clegg.
Clegg said, “As the line between human content and synthetic content becomes blurred, people want to know where the line is.”
He added, “People often encounter AI-generated content for the first time, and our users appreciate the transparency around this new technology. It’s important to let people know that it was created using AI.”
Clegg mentioned that the labeling feature is being developed and will be rolled out to all languages in the coming months.
He also stated that the company will add more prominent labels on images, videos, or audio that are “digitally created or altered” and “have a particularly high risk of materially misleading the public.”
Additionally, the company is working to develop technology to automatically detect AI-generated content, even when the content lacks invisible markers or has been removed.
“This work is particularly important because the online space is likely to become increasingly hostile in the coming years,” Mr Clegg said.
He concluded, “People and organizations actively trying to deceive people with AI-generated content will find ways to circumvent the safeguards in place to detect it. Our industry and society as a whole must continue to find ways to stay ahead of the curve.”
AI deepfakes have already become an issue in the US presidential election cycle, with examples of AI-generated deepfakes used to dissuade voters in the New Hampshire Democratic primary.
Australia’s Nine News also faced criticism for altering an image broadcast on the evening news that exposed Victorian Animal Justice Party MP Georgie Purcell’s belly button and altered her chest, using Adobe’s AI image tools.
Source: www.theguardian.com
