AI Enshittification on Substack

It's getting fucking harder to turn the damn things off

I do not (and will never) use any AI to write anything (not even a grocery list).

I am never going to use an AI to determine whether or not something was written by or with the assistance of AI (here on Substack or anywhere else).

I do use the crummy spellcheck in Word to help me catch typos which, alas, seem to be increasing the older I get (arthritic fingers one of the causes).¹

I spent a lot of time a couple of decades ago reading the scholarship by composition faculty about the problems (false positives and false negatives and general crappiness) with the “plagiarism detection” programs that were being sold (and are still being sold, probably with more AI) and even more time (unsuccessfully) arguing against university blanket requirements for students to run their work through them.

What I’m reading these days about AI makes² me fairly sure that that things have only gotten worse, especially given the environmental and human costs of the damned data centers.

So when I realized that Substack thought it was a good idea to roll out this AI called Pangram for everybody to use, my immediate response was simply no fucking way.³

Apparently, Substack requires us to actually use the damn program (give it your words for its data horde) in order to block it—on a post by post basis. I will not be doing that.

There is no way to just opt out completely which is right on brand for AIs (hell, I keep turning them off in Zoom and on my phone; but somehow the system keeps shoving AI links back at me; Adobe doesn’t even give me the option, and I have to shut down three different “suggestion” menu boxes every fucking time to read an article).

TL:DR: the act of opting out gives Pangram money and data.

Substack just enabled this program recently. The Substack help/information already has an elaborately detailed page of instructions (which I am NOT linking to) on how to use the damned AI to scan any text on Substack and, as far as I can tell, NOTHING on how to shut the damn thing off.

Karen Smiley, at Ethical AI, which I subscribed to last night, has an excellent post on “Why I won’t be using Pangram AI scoring,” complete with specifics on all the harms caused by the program biases and some useful links, including lots of other excellent substackers who are looking at AI critically, plus others to use to comment to support the request that Substack allow us a single opt-out.

Why I won’t be using Pangram AI scoring

1. Crummy because wow is it wrong and limited and clunky. During my years as a teacher, I developed my own internal/personal list of the “wrong words” that I could recognize on sight when students assumed the spellcheck was correct, and it wasn’t.

2. The ongoing pattern of university administrators wanting to replace faculty with machines started in the 1980s (according to my senior colleagues) when admins decided that the sexy new video technology meant they could video faculty lectures and just keep running the lectures ad infinitum. The more things change . . . .

3. My first “fuck no to AI” post has a link to an article about an AI conference using an AI to evaluate proposals on AI — turns out that AI was Pangram!