Currently, a significant issue being discussed regarding AI content is determining who generated the content, what data was used, and whether it has been compromised. This concern becomes even more pressing as deep fakes, forged images, and model theft become increasingly easy. The concept of PoA (Proof of Attribution) proposed by OpenLedger is specifically developed to address this problem.

PoA is a form of "AI provenance" certification. Whether it is text, images, audio, or video, any content generated by AI can automatically carry a watermark through PoA technology that indicates:

Who generated it + How it was generated + Generation time

This watermark is not directly visible but can be verified in a digital form. In simpler terms, when you encounter any content that appears to be AI-generated, whether it is an article, image, or audio/video, you can use this technology to check whether it is indeed AI-generated, which large model was used, and even specify the version of the model, such as GPT 3.0.

This application significantly enhances the ability to discern the current proliferation of AI. For instance, articles on Twitter or suggestive photos can clearly indicate whether they are real or AI-generated.

The action of applying this "watermark" is the technology of Infini-gram. This technology essentially functions as a content-level fingerprint system that can embed a unique identifier without affecting the quality of the image or text. Moreover, this identifier is resistant to compression and editing, ensuring it cannot be removed, much like a signature at the code level.

Using PoA + Infini-gram technology is akin to issuing a "digital birth certificate" for every piece of AI-generated content. Under this mechanism, the output of AI models is no longer untraceable; instead, every piece of content generated by AI can leave an identifier, making it clear whether it is AI-generated. This particularly enhances the competitiveness of content creators in the market.

Therefore, I personally hope that @OpenledgerHQ's PoA approach can be applied to more large models, enabling more people to easily identify AI-generated products.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。