The significance of watermarking systems has resurfaced with a new announcement by OpenAI on a new DALL-E 3 watermarking system. This significance grown due to concerns about the reliability of digital media, particularly with the increasing prevalence of generative AI technology. Various manufacturers of cameras such as Sony, Canon and Nikon have applied new camera tech to combat fakes and gain trust.
On the other hand, the AI image generators, they have been making their move to combat fakes too — to follow the standards of the Coalition for Content Provenance and Authenticity (C2PA). OpenAI are rolling out this DALL-E 3 watermarking system to image metadata and they’re just one of the companies that are following this C2PA standard.
What Is the C2PA Standard?
C2PA is an open technical standard enabling publishers, companies, and various entities to include metadata in media for verifying its origin and associated details. Beyond AI-generated images, the C2PA standard is also being embraced by camera manufacturers, news agencies, and other parties to authenticate the source and history (Provenance) of media content.
OpenAI’s New DALL-E 3 Watermarking Implementation Based on the C2PA
Images generated through ChatGPT on the web and via OpenAI’s API, which serves the DALL-E 3 model, will now feature C2PA metadata. This update will extend to all mobile users by February 12th. Users can employ platforms like Content Credentials Verify to verify if an image originated from the underlying DALL-E 3 model through the company’s tools. This inclusion of metadata should indicate that the image was generated using OpenAI’s API or ChatGPT unless the metadata has been deliberately removed.
However, it’s essential to recognise that metadata such as C2PA isn’t a fail-safe solution for provenance concerns. It can be easily removed, whether accidentally or intentionally. For instance, most social media platforms currently strip metadata from uploaded images, and actions like taking a screenshot can also eliminate it. Therefore, the absence of this metadata in an image doesn’t definitively confirm or refute its origin from ChatGPT or OpenAI’s API.
OpenAI maintained that embracing these methods to establish provenance and encouraging users to heed these indicators are crucial steps in enhancing the credibility of digital content.
Also Read: Bing AI and DALL-E 3: The Pinnacle of AI-Generated Imagery
How Does the New DALL-E Watermarking Look Like?
The new DALL-E watermarking will include both an invisible metadata component and a visible CR symbol, which will appear in the top left corner of each image. Besides, individuals have the ability to verify the origin, specifically identifying which AI tool was utilised to create the content, for any image generated using OpenAI’s platforms through platforms such as Content Credentials Verify. Presently, only still images, rather than videos or text, are capable of bearing the watermark.
More on the C2PA
The C2PA, a coalition comprising companies like Adobe and Microsoft, advocate for the adoption of the Content Credentials watermark to ascertain the origin of the content and indicate whether it was created by humans or AI. Adobe introduced a Content Credentials symbol, which OpenAI are incorporating into DALL-E 3 creations. Recently, Meta announced their intention to include tags on AI-generated content across its social media platforms.
Identifying AI-generated content is a focal point of the Biden administration’s executive order on AI. However, as mentioned in the above, watermarking is not foolproof in preventing misinformation. OpenAI highlighted that C2PA’s metadata can be easily removed, either inadvertently or intentionally, especially since most social media platforms routinely strip metadata from uploaded content. The screenshot is a straightforward example of it being foolproof and removing the watermark.
According to OpenAI, they stated out that: “We believe that adopting these methods for establishing provenance and encouraging users to recognise these signals are key to increasing the trustworthiness of digital information.”
More On OpenAI: 8 Best GPTs On The OpenAI Store That Will Save Your Time
The Future of AI-Watermarking
In conclusion, the latest update from OpenAI introducing the new DALL-E 3 watermarking system represents another significant stride in addressing concerns surrounding the origin and authenticity of digital content. By collaborating with initiatives like the C2PA and incorporating Content Credentials symbols, OpenAI aim to empower users to verify whether content was generated by humans or AI.
While this development aligns with the Biden administration’s directive on AI and underscores the importance of establishing content provenance, it’s essential to acknowledge that watermarking alone may not completely mitigate the spread of misinformation. OpenAI acknowledge the limitations of watermarking, emphasising the ease with which metadata can be removed. As the new DALL-E 3 watermarking system will only arrive for the first time in the next coming week, there’s still much more space for improvement to combat the existing flaw.
Nevertheless, embracing these methods to promote transparency and trustworthiness in digital information marks a crucial step forward in navigating the evolving landscape of AI-generated content.
Author Profile
Latest entries
GAMING2024.06.12Top 4 Female Tekken 8 Fighters to Obliterate Your Opponents in Style!- NEWS2024.03.18Elon Musk’s SpaceX Ventures into National Security to Empower Spy Satellite Network for U.S.
- GAMING2024.03.17PS Plus: 7 New Games for March and Beyond
- GAMING2024.03.17Last Epoch Necromancer Builds: All You Need To Know About It
