Ege Erdil comments on A conversation about Katja’s counterarguments to AI risk