Startup idea: API to verify media (real vs. AI-generated)

3 points by indieept 2y ago ↗ HN

With the surge of AI-generated content we are currently witnessing, and recent misuses of this technology to influence people and spread misinformation, I think it would be interesting to build an API-based solution to verify if online media (pictures, videos) is real or AI-generated. This would ideally become the reference in terms of trustworthiness, and every media company would double-check the content before using it online. Anyone building this / any thoughts?

3 comments

[ 3.5 ms ] story [ 12.5 ms ] thread

Michelangelo11 2y ago ↗

How exactly would you do this?

whelp_24 2y ago ↗

I think ClosedAi wrote something to detect if a peice of text was AI generated, failure rate was so high, they took it down. The problem is that is can't be reliably detected. Remember, chatgpt beat the Turing test (at least for a length of time).

If you create a api to detect generated media, you can use it to improve your generator.

verdverm 2y ago ↗

This will be a game of cat and mouse, much like search and SEO spam, or cyber security.

My belief is that a real solution will require changes at the hardware level, to incorporate hardware level attestation that "cannot be hacked" at the software level.