I won't buy a device with cameras slapped all over it either. But anyone fearmongering over this has not thought through what it would take to do this at scale and really consider how this is worthwhile over existing tracking methods. The amount of compute required to squeeze useful targeting data out of hours of video is just not cost-effective compared to the plentiful data from your activity on a phone or laptop. In all your examples, the same signal would be visible from online activity that is practically guaranteed to be there.