[pulse7]

Microsoft may be using AI for spam detection and... is - of course - unable to fix it, because you can not debug a neural network (or may need months/years to debug and fix it)... So: WELCOME TO THE NEW WONDERFUL WORLD OF AI where we will not be able to fix things...

[stareatgoats]

I think this is it, too. AI is efficient, but malfunctioning a fair percentage of the time. Are we going to have to accept it just because it has a number of upsides - or are we to demand that errors be brought to 0 before using it on critical systems, like email? Or in critical situations like war for that matter.