trait-feign comments on Could realistic depictions of catastrophic AI risks effectively reduce said risks?