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.

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.)