I would be interested in seeing AI created content as an experiment.

I would say that to be "ethical" you should disclose up front in each post "The following was created by an artificial intelligence"

If the content is upvote worthy, it should get upvoted, just like human posts. If the content violates TOS it should be flagged, just like human posts.

Many companies have tried to implement AIs that produce interactive content. One of the biggest problems with the "learning" engines is that they can be "taught" to respond to whatever is popular ... some have even become overly sexualized teen-bots in a matter of weeks.