First, I like this approach better, with an up-front discussion about the experiment you intend to pursue. So that's good!

Whereas I recognize that AI is likely here to stay, and will become increasingly enmeshed in the social web, I still have some reservations about its use in the context of a proof-of-brain community.

As @steemchiller points out, this might require some stringent testing and fine tuning.

My thought on anything AI on Steemit would be some very clear "marking," some of which might need to be written into the code base.... along with some fundamental rules for its usage.

For example, all AI comments might be shown on a pale grey background and marked as AI — much the same way "promoted content" is marked on other social sites. Next, it might behoove us to have — as a broad based community rule — that all comments/content deemed AI either will have all its rewards burned, or simply have an automatic "decline rewards" setting.

Where adjusting the code base comes into play would be in the form of adding an "option" of sorts for community members/curators to tag AI content as "AI" as a means to trap anything AI passed off as POB, whether deliberately or accidentally... maybe a little dropdown option that allows "mark as AI," or something.

I also agree with the sentiment that all AI generated comment systems should include an "opt out" option.