>hould we treat them as being harmful because computers make mistakes at a significantly faster rate? Yes. Significantly doesn't really cover it. When you incr

>hould we treat them as being harmful because computers make mistakes at a significantly faster rate? Yes. Significantly doesn't really cover it. When you increase the rate of an activity by many orders of magnitude and put an AI behind it, it becomes a totally different activity, even if the mechanics are notionally the same.