[smoldesu]

Who takes the fall when an AI screws up?

If you shine enough light on it, apparently the brand does. If a human were to do this, the company would immediately fire the employee and cut all ties with them. But as the article points out, 'fixing' an AI mistake isn't really a fix at all:

> [Google] said it was "appalled and genuinely sorry", though its fix, Wired reported in 2018, was simply to censor photo searches and tags for the word "gorilla".

[exporectomy]

If a human did it with intent to offend, they would. But if a black person genuinely looked like a primate to human eyes, that would be pretty shitty to fire the poor worker who had no way of knowing. Here, the AI isn't trying to offend so maybe there should be no consequences and people should stop demanding severe punishment for minor accidental insults.