Most conversations about AI danger are about the future. What it might do someday, what could go wrong eventually, what we ...