Hacker News new | ask | show | jobs
by tiagod 1770 days ago
> In the current iteration they are not using AI or anything to identify "bad" content. They are matching hashes for files on on your device against known bad hashes/content.

This is misleading as these are not what most technical people would think of as a hash. These are "perceptual hashes" that can compute a distance, instead of a boolean equal/different, leaving space for false-positives instead of extremely rare hash collisions.

https://www.hackerfactor.com/blog/index.php?/archives/929-On... https://rentafounder.com/the-problem-with-perceptual-hashes/