AI Generated content

[JAS_OH1]

yup, ai likes it's own version of truth ChatGPT falsely told man he killed his children

That was a horrifying story.

 

[AdaptAndAchieve]

I’ve found with translation it requires more than one go.
You have to translate and then then translate it back again to the original language you started with. That seems to pick up a lot of mistakes and gives you the opportunity to refine it.

Also sticking with the most simple sentence construction you possibly can or even missing out connecting words and just using a sort of shorthand is very effective.

This is exactly what I resorted to when on the russian server. For example instead of a sentence constructed like 'I'm going to town to buy a shirt', I'd make it two sentences like 'I'm going to town. I will buy a shirt'. The translation becomes much more accurate with 2 short statements rather than a sentence.
And as you say - a reverse translation as a sanity check

 

[AdaptAndAchieve]

That was a horrifying story.

Indeed, I can only imagine the mans shock and horror. In my opinion, at present AI is not developed enough to be trusted to publish anything online without human supervision ... but of course it is publishing a lot of things online