@james@mstdn.starnix.network
This is always the problem with neural networks — what you train them on.
I remember when Google had this CAPTCHA when you had to enter two words, one of the words was meaningful — it was checked against the expected result, the other one was used to train their OCR, which was later used to digitize books. And 4chan users trained it to recognize half the words as n-word 🤣
@atomicpoet