Post #371856

RE: https://iosdev.space/@sanguish/116116806565276449 This is basically what I‘m afraid of whenever I see someone using genAI for accounting or anything else where correctness matters and people tell me „no, I checked the results, and they were good!“ @sanguish@iosdev.space initially thought the results were correct (but was rightfully skeptical). He had to do a full inventory by hand to find the many errors. The errors you expect from software or humans are not the ones genAI makes, so you won’t know to look for them. You have to do a full separate calculation and compare the final result. And you‘ll have to do it again each time to be sure, as the output isn’t deterministic. So at that point, what time does it save?