Friday, July 19, 2024
- Advertisement -

    Latest Posts

    TikTok Announces New Collaboration For Automatic Labelling Of AI-Generated Content

    TikTok has partnered with the Coalition for Content Provenance and Authenticity (C2PA) to start automatically labelling AI-generated content uploaded from other platforms.

    The company says this will be done by implementing C2PA’s technology to read Content Credentials, which include metadata of content posted on platforms. TikTok has launched the process for images and videos currently, and will be expanding to audio content as well.

    “Over the coming months, we’ll also start attaching Content Credentials to TikTok content, which will remain on content when downloaded. That means that anyone will be able to use C2PA’s Verify tool to help identify AIGC that was made on TikTok and even learn when, where and how the content was made or edited. Other platforms that adopt Content Credentials will be able to automatically label it,” the company stated in its blogpost.

    The Coalition for Content Provenance and Authenticity (C2PA) provides technical standards for “certifying the source and history (or provenance) of media content” for aiding identification and prevention of misinformation online. The coalition is formed through an alliance between Adobe, Arm, Intel, Microsoft and Truepic. Read more about how C2PA works, here.

    How are companies labelling AI-generated content?

    In April, Meta announced that starting from May, it will introduce ‘Made with AI’ labels on AI-generated pieces of content across all its platforms. In February, acknowledging that metadata and self-disclosure are not fool-proof methods to tackle AI related misinformation, Meta had informed that the company is working on developing classifiers that can automatically detect AI-generated content, even if it lacks visible markers. The company shared that their AI Research lab FAIR is developing an invisible watermarking technology called Stable Signature.

    In March, YouTube also rolled out its new tool that will enable creators to disclose whether their content is “meaningfully altered” or synthetically generated and seems realistic. The tool is present in YouTube’s existing upload flow to allow user to label their content easily.

    In December 2023, Microsoft launched ‘Content Credentials as a Service’, which uses the C2PA digital watermarking credentials, to allow election candidates to digitally sign and authenticate content through digital watermarking. The digital watermark will contain details of how, when, and by whom the content was created, edited and also if it was generated by AI.  In August 2023, Google DeepMind launched SynthID tool for watermarking and identifying AI-generated images. Additionally, Google also adds metadata to generative AI images using the IPTC standards.

    Is labelling an effective measure against proliferation of misleading AI content?

    While companies are adopting measures like labelling and watermarking, experts speaking at MediaNama’s ‘Deepfakes and Democracy’ event highlighted that these measures may not be entirely effective in timely identification of AI-generated content on the internet.

    When it comes to authentication based on provenance information, a measure undertaken by TikTok, Gautham Koorma, machine learning engineer and researcher from UC Berkeley, informed that researchers have been able to break every watermarking technique that exists out there.

    “I think we should be using watermarking, but keeping in mind its downside, that it can be easily broken by a sophisticated adversary. […] For a non-sophisticated actor, it might be a little bit more difficult, but this is why the push for detection, because once you have removed that watermark, then what do you do?” Koorma enquired.

    Read about various methods of detecting AI generated content and their effectiveness in MediaNama’s detailed report, here.

    Also Read

    STAY ON TOP OF TECH NEWS: Our daily newsletter with the top story of the day from MediaNama, delivered to your inbox before 9 AM. Click here to sign up today!


    The post TikTok Announces New Collaboration For Automatic Labelling Of AI-Generated Content appeared first on MediaNama.

    Latest Posts

    - Advertisement -

    Don't Miss

    Stay in touch

    To be updated with all the latest news, offers and special announcements.