The soil that AI grows in is permeated by the outlook that everything is performance. The mission is not to get correct answers; it is to deliver answers with style. Some of them will be right, some wrong, some neither. The style is the thing. That, if anything, is what the customers will be willing to pay for.

What's damning about this is that such an error was both not caught by the researchers and then not corrected after it was raised as an error. Is this the kind of conduct that should be acceptable amongst AI researchers who want to stake a claim on their work being authoritative?

> this can lead to headaches

A relative worked at home doing medical transcription. It was not unheard of for physicians to make up non-existent body parts, non-existent drugs, etc. Part of her job was to correct their terminology.

She has since been replaced by cheaper off-shore transcribers and by speech to text applications. It's not clear how terminology is corrected now.