Key Points
- Adam Mosseri warns that authentic content is becoming scarce on social platforms.
- He suggests shifting focus from creating content to creating uniquely personal content.
- Mosseri proposes fingerprinting real media using existing metadata like EXIF and XMP.
- Current AI tools from Meta, OpenAI, Google Gemini, and Perplexity can produce both polished and imperfect aesthetics.
- A universal standard for content identification could enable filters for human‑generated posts.
- Uncertainty remains about fingerprinting text and audio content.
- Mosseri hopes widespread adoption of fingerprinting will restore trust in social media.
Background
Social media platforms are increasingly saturated with AI‑generated images and videos, making it difficult for users to discern genuine content. This trend has prompted concerns among creators and platform leaders about the erosion of authenticity online.
Mosseri’s Perspective
Instagram chief executive Adam Mosseri observed that “authenticity is fast becoming a scarce resource.” He noted a shift in creator expectations, moving from the question “can you create?” to “can you make something that only you could create?” Mosseri also described a growing preference for unpolished, natural content that contrasts with the highly processed aesthetic promoted by some smartphone manufacturers.
Proposed Approach
Rather than continue to watermark AI‑generated media, Mosseri suggested that platforms should focus on fingerprinting real media. He argued that it may be more practical to label genuine content before it reaches users. Existing metadata standards such as EXIF for photos and XMP for video already embed details about the camera, lens, settings, location, and date, which are difficult to falsify.
Technical Considerations
While image and video metadata provide a clear pathway for fingerprinting, Mosseri acknowledged uncertainty around how to apply similar techniques to human‑written text or audio recordings. He also mentioned the possibility of using author history and previous posts as additional authenticity markers, though the scalability of such methods remains unclear.
Industry Implications
Mosseri highlighted that major technology firms—including Meta, OpenAI, Google Gemini, and Perplexity—are developing AI tools that can generate both highly polished and deliberately imperfect aesthetics. These tools could enable creators to produce content that appears authentic while still being AI‑assisted. Mosseri emphasized the need for a universal standard for content IDs, allowing any platform to offer users a simple filter to view only human‑generated posts.
Looking Ahead
The Instagram chief expressed hope that the coming year will see widespread adoption of fingerprinting methods, restoring utility and trust to social media. He cited an example involving bird photographer Carl Bovis, where viewers struggled to differentiate between AI‑generated and real images, underscoring the current challenge. Mosseri’s vision calls for coordinated industry action to protect authenticity while embracing the creative possibilities of AI.
Source: techradar.com