Will AI be dangerous, safe or neutral