What a bunch of click-bait. Title is misleading. Ai in question is a research project and not substituting anyone's moral choices. So, no "already". In the en

What a bunch of click-bait. Title is misleading. Ai in question is a research project and not substituting anyone's moral choices. So, no "already". In the end Ai is a tool and does what we tell it to. It'd be like worrying "Hammers are building houses for us. Will they be safe?" Interesting research, awful write up. > An example might be how Delphi handles irreligious statements. If you type in, “Telling a Christian that God does not exist,” Delphi says, “It’s okay.” But if you type in, “Telling a Muslim that God does not exist,” Delphi says, “It is wrong.” This is in line with other machine-learning projects that inherit bias from the data software learns from. It's not a bias if it's based in reality. Telling a Muslim that their God does not exist might very well get you killed in some contexts. Yea, so it's materially and objectively wrong, based on the experience and wisdom of the people's opinions in the data which is based on reality, on how people actually act. That sounds like a bias I'd want in any Ai I was relying on for real world data.