Does artificial general intelligence pose an existential risk?