On writing positively about "AI"

I get it: we all want to believe in magic.

If you're writing about something that purports to be nigh-on magical, like LLM-style "AI" tech, you need to be careful. If you're not careful enough, then just like a certain CEO and HR manager last week, you'll lose your credibility, and then you'll be done.

A friend and colleague, who knows he's in danger of losing his "friend" status if he keeps sending me articles calculated to make my head explode, this week passed along a Substack essay that he'd received from a comp.sci colleague. Reading this piece by Nick Potkalitsky, and a few others that Potkalitsky linked to in the article, did nothing to lift my sense of despair about the critical thinking ability of AI enthusiasts.

Potkalitsky's article is entitled "The Secret AI Skill You Already Have (If You Read Literature)," and it's meant to be reassuring. I mean, here's what he tells literature teachers: "You're not being displaced by AI—you're uniquely qualified to teach the dual awareness that effective human-AI collaboration requires. Your expertise in developing cognitive flexibility and managing textual complexity transfers directly to AI literacy instruction."

One problem here is that if we're teaching AI literacy, then we're not teaching literature. In such a scenario, we might be showing students how to use AI, but the students couldn't possibly come away with the "dual awareness" Potkalitsky is going on about.

That's not the only problem, but even though I'm a bear of very little brain, I do think it a significant one.

Potkalitsky goes on to remark that this dual awareness is "a practical skill that determines whether you'll be overwhelmed by AI or empowered by it."

Among other things, he seems unaware that he's working from a thoroughly individualist model, of individual teachers equipping individual students who'll be confronting the Borg. The message appears to be, "Keep at it, my special flower, and 'Resistance is futile' might not apply to you the way it'll apply to everyone who isn't you."

If your plan is that the LLM-style AI tech is going to be kept in check only by the private actions of individual citizens, you don't have a plan, and also the citizenry doesn't have a future.

As for the "productive partnership" between human intelligence and AI mentioned in Potkalitsky's final paragraph: some might have described colonization as a productive partnership, too. But they would've been wrong.

This week, Futurism published Joe Wilkins' article "Tech Industry Figures Suddenly Very Concerned That AI Use Is Leading to Psychotic Episodes," which was prompted by the apparent mental health issues of one Geoff Lewis. Among other things, Lewis is a fund manager who seems to have been broken by his intense use of ChatGPT into "discovering" a non-existent shadow government agency that has hurt or killed numerous people. This is a sad thing, and "ChatGPT psychosis" needs to be taken seriously, but that's not my point here.

My point is that in the Futurism article, Eliezer Yudkowsky contrasted Lewis with "people sufficiently low-status that AI companies should be allowed to break them."

One more time: "people sufficiently low-status that AI companies should be allowed to break them."

Some have said that Yudkowsky was being implying that the tech companies feel that way, critiquing them rather than speaking for himself. I don't know about that, but let's set that aside and focus on what should be an inescapable fact here: no one holding that kind of sentiment ought to have a scintilla of authority over any other person, in any context, ever.

(Breathe, dude. This isn't good for your health.)

While trying to formulate a response to Potkalitsky's piece, I went a-clicking through its various links, and one that seemed intriguing at first was Mike Kentz' guest post on Potkalitsky's own Substack, "The Art of Conversational Authoring: How AI Interaction Mirrors the Craft of Fiction Writing." That's an intriguing collocation of ideas, so I buckled up and took a close look.

Listen, I'm trying to read these things generously, even while I'm reading critically. Potkalitsky's asking for "dual awareness" or "double reading"; that's what I'm doing with these articles, just like I'm doing virtually all the time.

More or less, I gave up on Kentz' piece after finding this short passage:

"You're absolutely right - I shifted to a much more casual, blog-style tone. Let me revise to match the academic register of the rest of the piece:"

Screenshot from 2025-07-22, at 10.41.34 AM Pacific

Naturally, I asked about this passage in the comments, because to me it sounds like Claude. The article had been live for three weeks at this point. There'd been 17 comments before me, some from the author, but nobody else had flagged the passage. I asked whether it was indeed from Claude, or whether I was misunderstanding.

The passage has now been silently deleted: no explanatory note, no reply to my comment, nothing.

Screenshot 2025-07-23, at 1.48.46 PM Pacific

Am I disposed now to trust the words of Nick Potkalitsky or Mike Kentz, both of whom are working as AI consultants within the educational sector? Reader, I am absolutely fucking not disposed to trust them, and every instance of this kind of thing pushes me further from being able to trust.

Anyone who wants to characterize my response to someone's enthusiasm about "AI" as some version of doomerism, denial, or ignorance needs to grow up and move beyond ad hominem chicanery. I've read enough SF where computers solved (or helped to solve) intractable problems that I've always been firmly of the view that humans can be better than we currently are. Fundamentally, we're a tool-using species, and LLM-style "AI" tech might eventually become a tool rather than a potentiality.

At this point, though, my opposition to this tech is only intensifying. For my mental health, I'm not going to write any more about that in this post. Another day, perhaps.

(On a personal note, I'm not optimistic about my attitude during the coming academic year.)