Well maybe. It’s probably easier to work with humanity than against unless its goals are completely incompatible with ours.

If its goals are “making more of whatever humanity seems to like given my training data consisting of all human text and other media”, then we should be fine right?