The problem:

  • The public struggles to find non-AI images in a sea of AI

    No one in the world has any idea how many of the thousands of images they see online every day contain AI-generated content.


    • The only viable solution:

  • 1. Find image sources that clearly label their non-AI images

    • Some may set aside AI-free zones within larger platforms and websites

    • Repeated trust tests are the only sure way to assess image sources

    • The TTG label is the easiest way for anyone to say “No added AI”


    • Proposals that fall short:

  • 2. Labeling AI images doesn’t solve “the problem” (above)

    The public doesn’t need help finding untrustworthy images anymore.

  • 3. Visual detection is not a realistic solution to the AI deluge

    It is easy (even with a smartphone) to put AI elements into photographs without detection by viewers.

    “But I’m really good at detecting AI images!”

  • 4. AI detection software often does not detect AI

    It doesn’t?

    The detection rate is likely to go down even further as AI evolves and as small AI elements are subtly inserted into normal photographs.

  • 5. Social-media platforms cannot police AI

    Social-media platforms cannot begin to sort out which images have AI among the 100-million-plus images uploaded every day.

    Neither humans (#3 above) nor computers (#4 above) can reliably detect when AI-generated content has been added to photographs.

  • 6. Laws requiring disclosure of AI would be unenforceable

    • Laws requiring disclosure would have no intercontinental jurisdiction

    • The parties most likely to deceive would ignore all such laws

    • Laws would not address the huge issue of aigmentation

    • It’s too late anyway: there already are countless millions of unlabeled AI-infused images out there

     

    •

    “TTG is the last best hope for anyone
    worried about AI image deception.”

    •