Wow. What a great way to intensify biases and put them in a feedback loop.

I wonder if this will make companies try to create ai that actually undestands stuff and doesnt need to be trained in a way they currently are. Because eventually this will pollute ai’s training to garbage.