Cowbee, im sorry, but as you yourself say this isn't a settled question at all. There is some common ground we can agree on, but after that there's many different assertions one can make, all with decent reasoning behind. Those assertions can then be related to in a few reasonable ways as well.
What we can agree on is that most gen-ai right now is not made with locally run models using renewable energy and the data used to train those models does come from massive data centers. There are people using locally run models and there's probably someone doing something renewable and there's some who have their own smaller models that dont rely on data centres, but they are quite far from being the norm.

So unless otherwise stated I assume it is an average user of an average gen-ai.
Then the other assertion is that people find it easier to dismiss AI-slop - easier even than some meme, which means AI slop makes for poor agitprop. At this point there is plenty to be pointed towards going for and against this assertion, so its not settled in any real way. As such, I doubt we are able to come to a consensus at the moment.
Another assertion is the one I made about normalisation. If there is anything that can be solved thru discussion at the moment it might be that, but honestly it isn't the big reason im against it, since normalisation is only bad under current circumstances in my eyes.