Hacker News new | ask | show | jobs
by d3vnull 348 days ago
What's keeping anyone from using a trustworthy camera to take a trustworthy picture of a screen displaying an AI generated image?

And that's the low tech version. Taking a trustworthy camera and plugging the CMOS sensor connector into an LLM that generates its output as RAW data that feeds into the sensor isn't that hard if you want to make extra sure you're not detected.

Such a system might even cause more harm than good by giving people the impression that real/AI images can actually be certified.

1 comments

Because it will be a picture of a screen not reality? The hypothetical camera knows the scene it's taking is a flat surface or not, and puts that detail into the signature.