It's not the em-dashes. It's the dumb-asses.
AI's Id-10-T error
Tropes are interesting.
A topic has ‘made it’ culturally when creative work about starts relying on tropes, and you’ve already seen tropes about AI.
Note: I’ve been focused on a freelance project that’s due today. So, the next post in the series about YouTube changes will be delayed. That covers how Shorts are f#cking small creators. It’s guaranteed to make you mad, so stay tuned!
That image relies on both Seinfeld tropes and AI tropes, for example.
I just realized I can hear that image. I can feel it in my fillings. Perhaps that’s something no AI will ever experience: The full cultural (and literal) resonance of Jerry Seinfeld’s voice.
Woe to those who dare write “it’s not x—it’s y” and face the angry mob.
The problem with these tropes, and the pitchfork-waving that ensues, is that they’re not all that reliable for discriminating between AI and non-AI writing.
Which is not some new thing I just discovered. Plenty of people have been saying this for a while.
Wait. Imagine breaking up with someone, giving them the old “it’s not you—it’s me,” and getting accused of using AI. That would be hilarious.
The list of AI ‘tells’ has also gotten quite extensive… expansive? Whatever, there’s a bunch now:
I used to teach writing. These AI "patterns" didn’t come out of nowhere.
The hallmarks of AI writing are what, for decades, we've told undergrads to do for a grade.
It's what we taught the D-students to do, to improve their writing to a C-student level, so we wouldn’t have to see them again next semester.
We were ‘rescuing’ them from their dumb-assery, so they could produce writing that was good enough (barely) for the purpose.
It didn’t make sense to teach those students nuance. They didn’t know enough basics to understand that. So, we just gave them rules of thumb to ‘class-up’ their trash.
We taught them how to put lipstick on their pig before turning it in.
Well, guess what? Those kids graduated and got jobs.
In the working world, the writing characteristics we taught those D-students merely so they could get by, became hallmarks of writing produced by an educated professional.
That is, not a ‘professional writer’ but someone who 'knows how to write professionally’, i.e., for or within the context of their profession.
How many times have you (if you’ve done any hiring) seen ‘proficient in writing professionally’ or some version of that phrase on a resume or cover letter? Dead giveaway that this is not a professional writer.
Everyone working in marketing — those who don't specialize in writing, anyway — was a ‘rescued’ D-student.
Some of them will be mad about that sentence, because they were never taught about hyperbole.
That’s one of the many nuances we had to skip over to keep them off of academic probation.
Now, if you know how AI 'learns', it's completely unsurprising that its writing shares characteristics with middling student writers. AI constructs heuristics (rules) from examples it's been given.
AI was trained to write professionally by studying examples of what it was told was 'professional writing.' This was not examples of writing created by professional writers. It was writing created within a professional context. Most of which is not created by writers, just professionals.
When you ask AI to write something for a professional context, you get style based on what it's learned 'professional writing' should be, so...
AI writing is based on rules derived from the work of C-students who've entered the working world.
AI is more precise than people in a specific way: It’s better (more consistent) at following rules. And because AI is 'better' at following rules, all the C-student writing characteristics get amplified.
It’s “shit in, shit out” writ large.
We showed AI what it ought to do, just like we showed the D-students. Since neither knows any better (neither understands nuance), they now think what they’ve seen is ‘good’.
And, obviously, more of whatever makes something good makes it even better!
What began as, ‘how about you add an em-dash so that this nightmare of a sentence becomes mildly understandable?’ turned into ‘em-dashes = good!’
Pre-AI, there were a lot of professional writing tasks where some dumb-ass should have hired a writer but thought "nah, Jones went to college — we'll just have him do it!"
And the quality was good enough (barely) for the purpose… but every writer knew that wasn’t written by a non-writer.
Post-AI, there are a lot of professional writing tasks where some dumb-ass should hire a writer but thinks "nah, AI is good — we'll just have AI do it!"
And the quality is good enough (barely) for the purpose… it’s just more obvious to non-writers that it was written by a non-writer.
What’s happening now is analogous to…
“The Greatest C-Student Story Ever Told.”
(Note: this actually happened, I was not the professor or student in question.)
A student was arguing with a professor about their grade. They believed their paper deserved better than a C. The professor explained why their grade was what it was, pointing out the characteristics of their paper that determined the grade.
The student continued to argue, saying, “my friend’s paper got a B and I can’t see any difference between their paper and mine!”
The professor, frustrated that the conversation had carried on for nearly an hour, replied, “well, if you could tell the difference between a B-paper and a C-paper, you wouldn’t write a fucking C-paper!”
(Seriously. This happened.)
Many people calling out AI writing now think they’ve learned that difference.
They haven’t. The quality of writing from non-professionals has not changed as a result of AI. There’s just more mediocrity in it.
It’s no better. It’s no worse. It’s still trash.
The only thing that’s really changed is who’s writing the trash, and how often they remember that em-dashes exist.
Like most technological problems, it’s what tech support refers to as an “Id-10-T” error.
Yes, that’s a bit of a boomery joke.
I worked tech support for over a decade and things that sound like jokes just become jargon. You’re grateful for shorthand expressions when it’s the 100th time you’ve had to deal with the same incompetent executive.
And that same executive, who had trouble understanding how to right-click, is now using AI or commanding others to.
The idiot behind the keyboard doesn’t understand the insanely powerful tool in front of them. They’ve just been told that the difference between a D-grade executive and a C-grade executive is “the power of AI.” They don’t have the basic understanding needed to comprehend the nuances.
So, when you see a piece of writing that you think might be produced by AI, maybe it was just made by a C-student. And if it is AI, it’s actually the result of a complicated series of interactions between dumb-asses: students, executives, marketers, and… well, writing teachers.
Me included.
I hate when people gate-keep knowledge about media and content creation. That disadvantages beginners who need that knowledge the most. Keeping the AutoTea substack free is a huge priority for me, but it does take time to create these posts. So, if you appreciate what I do here, considering buying a tea for AutoTea!
I feel like with the constant updates and tweaks, everyone's style of writing is going to have a "that sounds like AI" moment at some point or another.