> Overall don’t do anything with AI that is ethically wrong to do it by other means (for example, you can replace AI with expensive human labor to do it)
Not sure I understand what you mean -- can you explain this?
If I understand correctly, he's saying "if it would be unethical if executed by a human, it's unethical if done by a machine".
Depending on where you draw the line here, things like military applications, insurance price calculation, or drawing convincing Photoshops of famous people in the nude could all be in that category.
There's a difference between "unethical to make a human do" and "create an unethical result": e.g. there are ethical benefits from the automation of risky jobs. The obvious one in ML and computers is image recognition of horrific images.
(Granted, there are then additional issues of where and how you deploy that technology.)
Not sure I understand what you mean -- can you explain this?