For now, a valid “attack” of Friendly AI is to actually research this question, given that it wasn’t seriously thought about before. For time travel or antigravity, we don’t just not have at attack, we have a pretty good idea of why it’s won’t be possible to implement them now or ever, and the world won’t end if we don’t develop them. For Friendly AI, there is no such clarity or security.
For now, a valid “attack” of Friendly AI is to actually research this question, given that it wasn’t seriously thought about before. For time travel or antigravity, we don’t just not have at attack, we have a pretty good idea of why it’s won’t be possible to implement them now or ever, and the world won’t end if we don’t develop them. For Friendly AI, there is no such clarity or security.