It will be literal killer app, if the algorithm inadvertently alters faces or car license plates on camera feeds so that the error is not obvious. Today's codecs cause mostly predictable and obvious artifacts, yet there are problems already [1]. Can anyone guarantee that ML algorithms get these crucial details right?<p>[1] <a href="https://www.zdnet.com/article/xerox-scanners-alter-numbers-in-scanned-documents/" rel="nofollow">https://www.zdnet.com/article/xerox-scanners-alter-numbers-i...</a>