Japanese Versus American Victory Claims (5 Viewers)

[Irregualr23]

So much about aviation is covered incorrectly in the media that AI tends to retrieve the incorrect stuff as gospel. I've spent a lifetime around aircraft / aviation. Hardly any media civerage about aviation is anywhere near true. Quite a few of the books used as references by AI are nothing more than collections of stories with the attendant embellishment that always creeps in when stories are told.

When I have used AI to get things about WWII aviation, it is about 40-60% accurate if I'm lucky. Many times I'm not so lucky. AI doesn't sift through text for accuracy; it sifts for published contant regardless of accuracy. It basically outputs what it finds written in text that is avaiable for perusal.

It's kind of like a self-driving car. The "goodness" of the AI driver is completely dependent upon the programmer or team that writes the software, and they're only as good as the things they anticipate might happen in traffic. If anything happens they have not anticipated, the result is a toss up as to whether or not the response is good. Witness the stupid things done by self-driving taxis fairly recently.

Free AI tools simply tend to make up stuff, using sota model and anything above gpt5.6 and fable5 mostly get rid off the hallucination you see in older models, and they are very useful for going over large quantities of images and ww2 orals in batches and always linking evidence with original source rather than make up any fact by their own.

AI is definitely doing better on reading and understanding context than those who doesn't understand what researcher's writing based on losses reported on each side rather than pure claims.