Overall agree, although regarding > large enough AI models cannot be "race-blind" because if you remove race as a feature, they will be able to infer it anyway

Overall agree, although regarding > large enough AI models cannot be "race-blind" because if you remove race as a feature, they will be able to infer it anyways from proxy features In theory using a gradient reversal layer and an adversarial classifier you could do just that, to an extent. It could hurt the model's performance, which is exactly your point (should we ignore features with signal because they can be used to discriminate.)