>This is the same semantic argument of "Guns don't hurt people". Of course, guns hurt people. A physical object can be harmful without a human being to give it intent. It's a straw man argument that delves into the meaning of words than actually addressing if something is actually harmful or not.
Given the proposition that 'objects' harm people without human involvement, that does not in any way improve the null-hypothesis of wanting to prevent harm. If an earthquake or tsunami hits we do not shrug it off as an act of nature or god. We have an intent to prevent the disaster even if no human was involved. The comparison to guns is an ill fitted one. A gun that no man knows about cannot hurt someone. A person still has to wield it, compared to AI, people are definitely the ones wielding the weapon. Weapons, like AI is a tool. whereas guns have a very clear target and modus operandi, AI has no such thing. AI is by definition innocent of every crime in the book since you cannot demonstrate it to have morality in the first place. Which is a problem if someone has used it to commit crimes. Obviously AI doesn't kill people, the people using AI want someone dead.
In the context of the article, it means we have to take responsibility for what the AI creates regardless of its intent, precisely because there might or might not be someone behind with intent.