[thomasjoulin]

That's why I mention biases, which is a concern as AI becomes more and more ubiquitous. This is admitedly a silly test, I don't mean to dismiss the whole project because of a single response. I just find it interesting that, because most humans would be tricked, AI tools based on human generated data are inheriting their biases (conscious or unconscious).

Imagine if (or when) these tools were used to make more serious decisions, like hiring or sentencing:

For example, if an hiring AI disregards a female candidate over a male candidate with the same experience for an attorney role because statistically the male candidates fits the role more even if resumes are otherwise similar.

Or a sentencing AI infering crime is more likely to be committed by some groups, purely because those groups are currently over-represented in the prison population...