Why does the caption display 17°C? Thats not very hot?

I wonder how biased such a system can become. We’ve seen how deep learning systems will end up being just as biased or more than the data they are trained on. What is stopping it from flagging all <insert race/gender/ethnicity> as suspects and then reinforcing it’s beliefs because it so happened that a few of them were stealing?

— Je me demande a quel point ces IA deep learning sont partials. On a déjà vu a quel point des systèmes equivalent sont aussi partiales ou plus que les données avec lesquelles ils sont entraînés. Il n’y a rien qui ne les empêches de cibler des groupes spécifiques et de renforcer leur croyance parce qu’une partie d’entre eux volait.

Could someone recap whats going on?

I miss Apollo :(