Artificial Polemics / Questions still causing trouble
Questions worth asking, arguments worth having, and ideas I’m not entirely sure I believe yet.
“I’m going to need another glass of wine before I agree with that” is a perfectly respectable way to respond to an argument. This is where the polemics live before they’ve had enough of the juice.
Part parking lot, part waiting room, part launching pad.
Working questions and examined arguments by Glen Cantrell, alongside the essays in Artificial Polemics.
Questions without essays attached. The capital letters below are working titles; they appear to have more confidence than I do.
A real DNF reviewer found my writing “prefabricated.” In the essay that followed, I suggested that the real AI tell might be beige prose. An excellent observation, right up until I remembered I also write prose. The next examination begins by asking Claude, “Is my prose beige?” I’m not sure I want the answer. Naturally, I’ve given the question its own address.
The essay is now published: Beige →
Working title: THE MACHINES ARE BECOMING SELF AWARE
I’ve spent enough time worrying that I sound like a machine. Are the machines now being taught to worry about it too? Claude offered a guess about its own em-dash habits, which is an interesting starting point and rather poor evidence. The investigation here is into documented instructions and tuning intended to steer models away from stereotypical AI prose, including the much-persecuted em dash. The title is getting ahead of the evidence. Whether any of this deserves the words “self aware” is part of the question, not a finding.
Working title: AI MADE ME DO IT
I ask for a cleaner sentence and get back a slightly different conviction. A qualification disappears. A suspicion acquires a necktie and becomes a position. At what point does editorial assistance change the argument a reader thinks I’m making, or the beliefs they reasonably attribute to me? “I approved the edit” remains an inconveniently good objection.
These have essays attached. That means I’ve spent some words on them, which is a considerably smaller achievement than settling them.
Research, criticism, copyediting, writing: an enormous range of activities collapsed into one accusation. Where do we draw the line, and can we defend where we put it?
A stranger in Vermont, a book club made of machines, and an elegant little ferry I had to dismantle. Making something becomes more useful when I can also afford to discover what’s wrong with it.
“Bless your heart” travels through a blizzard with a generator, a hug, and a casserole dish. The words stay put. The meaning refuses.
A detector produces a number. A human decides what that number is allowed to do to somebody. I would like considerably more scrutiny between those two events.
I made myself easy to find, and the junk mail did its homework. Scattered public details become a portrait convincing enough to sell me something. Discoverability apparently comes without an audience selector.
My first DNF reviewer had a legitimate complaint about patterns in my prose and an uncertain suspicion about AI. The literary shortcomings may be entirely mine. I’ve spent years developing them.
AI, Write Like an Eighth Grader. Or a Robot. Possibly Both. →
I reserve the right to change my mind. Ideally after thinking about it. Or more wine.