The Pangram gong show

I woke all bedhead and bleary because I couldn’t sleep last night. Walked down the hall and stopped. The pillows on my couch were hot pink, neon pink, and I stared at them in sleepyhead confusion. Wtf? Why are my pillows pink?

If you’ve ever been here, you know I have two Icelandic sheepskin throw cushions on a black leather loveseat. I toss them on the floor to make room for company to sit but mostly I lean against them to read. One taupe and one black except this morning they were both streaked with hot pink. Half awake, I stared at them in confusion.

I turned to the kitchen, looked out the window. The sun was a red ball and I wanted to cry. Because I know. It’s fire. Canada is on fire so I opened the weather channel and it said the forecast today is smoke. What the hell kind of weather forecast is that?

Despite the fires, the sun is shining sunbeams through my kitchen window and onto the loveseat and the giant monstera and there’s a hot pink sunbeam streaking across my dining room floor and it’s surreal to see the hot pink and know North America is on fire but what we’re angry about is Pangram. God, the world is so screwed up.

I saw a Note that asked Substack to let us turn off the “scan for AI” feature. There’s enough AI theft, she said. She doesn’t want AI tools scanning her writing and feeding it to AI. Substack, you’re supposed to be on “our” side, she lamented.

I stared at the Note, thinking okay, but Pangram doesn’t feed our writing to AI.

Does she not know that?

Apparently not.

Because, why check I guess?

Every business has a business model. Pangram’s is do not eat words on the internet because people lie. People say it’s human writing when it’s not. So do not eat words on the internet because it might poison the model because people can’t stop lying.

They don’t say it that bluntly of course. But they say it.

Their site clearly says they do not “feed” words they scan into AI. They explain why in polite terms. Words to explain that they are trying to differentiate and feeding their models on text they find on the internet would compromise the accuracy of the model and basically that means they know that people lie about using AI.

You want to know who is stealing our writing and feeding it to AI?

Not Pangram. Other writers.

Or, at least, they call themselves writers. I’m not sure I’d call them writers. Because actual writers have respect for other writers and they do not. Let me explain...

I’m not talking about people churning out AI slop posts.

There are writers here who want the benefits of commenting on popular posts so they get visibility and the link to their profile on every comment, except they couldn’t be assed to actually read anything. So they feed our writing to AI and get it to leave comments for them. It makes me so angry, you have no idea how angry.

I want to shout how dare you? Who gave you permission to do that? Did I tell you it’s okay to feed my writing to AI so it can comment for you? No, I did not. Assholes.

That’s who is feeding our writing to AI.

Also? People who find popular articles and feed them to AI and say rewrite this to pass a plagiarism test and it does. I’ve lost count of the number of times that has happened to my writing. I see a post that sounds suspiciously like something I wrote so I click and read it and every point I made is in there and there’s not a goddamn thing I can do about it. It makes me so angry I could just cry with anger.

That’s who is feeding our writing to AI.

Our “fellow writers” — if you can call them that.

Not Pangram. Our fellow writers here. Feeding our writing to AI.

On Sunday, there’s an arts and crafts fair in the park and if we can breathe the air on Sunday I’m going to go with my kiddo. And if it’s a glorious sunny summer day, and we walk through the park admiring the hand created goods, I’ll still know. It’s only nice here because the winds shifted and someone else is breathing the smoke.

The fires are so bad they can’t put them out. Over nine hundred fires.

Burning, burning.

AI detection and Pangram is like that. Another raging fire.

There’s a lady who comes to some of the shows selling hand knit Cowichan sweaters. Dad always wanted one and I didn’t know until we were sitting outside one summer day in the last year of his life and he said he’d always wanted one, but he couldn’t justify the price. I was going to get him one for fall, but I lost him the Sunday before Father’s Day. Broke my heart that I couldn’t give him that sweater.

A handknit Cowichan sweater takes 40-50 hours to knit.

So when you buy one, you’re paying for an entire week of that person’s life.

You can buy machine made Cowichan sweaters, too. Funny thing is, the tags never say “machine knit” — they just hope you don’t know the difference.

Know what the difference is? A handmade sweater is made by someone who put hours of skill and labor into it. They didn’t start out good. They started out with uneven stitches and they dropped stitches and had to unravel and start over.

Kind of like the way we learn to write.

Machine knitting a sweater is a thing you need to learn, too. The difference is that you need to learn to operate the machine so it can make the stitches. You are not making the stitches. No one operating a knitting machine thinks they are hand knitting.

But let’s step from knitter to the person buying a sweater, okay?

Some people don’t care if their Cowichan sweater was made by a machine or a little old Native woman clicking her knitting needles. Other people care. Dad cared. He wanted a real Cowichan sweater, knitted by a little old lady with knitting needles.

No one yelled JOHN, WHY DO YOU CARE if it’s hand knit or machine knit?

No one says if you like the sweater, does it matter if it was made by a machine? No one screams “witch hunt” if you just want to know if the sweater was knit by hand.

No one asks any of those things about hand knit sweaters. Just about words.

Here’s another Note I saw. A writer said he used Pangram to scan his writing and it said 100% human. Yay, because it was. Then he saw a typo and fixed it. Then Pangram said it was AI assisted. Then he changed a word and it was “human” again. Then he edited a sentence and it was AI again. Bwahaha, he said. This thing is bullshit.

Over a thousand hearts. Two hundred re-stacks.

Sometimes I could just weep in frustration.

Because Substack TOLD us. In their post. We can report errors to Pangram.

To help it learn.

So it can find the worse AI abusers here and identify them.

The people feeding our writing to AI. The people churning out daily slop.

But what’s the f—ing point? Because we all know the people using AI will churn out their f-ing slop and when Pangram marks it as AI, they’ll report that it’s an error…

While human writers bitch about witch hunts and scream at Substack to turn it off.

And just watch — the articles about Pangram’s errors will start flooding the feed. Because yay, clicks. Because 1200 hearts and two hundred comments.

Sometimes, we humans are collective exhausting.

Let me tell you about a company you’ve probably never heard of. Yes, it’s relevant.

There’s a company called Form Energy in West Virginia. They’re working on a system to store renewable energy. Grid-scale storage that’s measured in days, not hours.

Because they see what’s happening in the world. And they say look, maybe we can store energy so if a grid goes down, we can supply energy to people. So the food in your fridge and freezer doesn’t go bad. So people don’t die from the heat.

The polar opposite of companies like Saudi Aramco and ExxonMobil and every other company who gives exactly zero shits what happens to us as long as they extract maximum profit for their shareholders today and tomorrow and next month.

Giant, giant companies that have the budget to change how we do things but they won’t because what they’re doing is profitable and change costs money.

It’s always that way in business.

There are AI companies who will steal everything on the internet and use it for their profit and tell us to suck it up, this is the future. And there are companies who are using AI to try to help us sort through the bullshit mess we’re dealing with.

And if we aren’t smart enough to try figure out which is which, then what?

Last Note reference. I saw one Note where a man said look, using AI to detect AI isn’t going to work. We have to just accept that this is the world now.

Well I don’t accept that. I don’t accept it anymore than I accept the world burning because of corporate profit or people dying because they ate lettuce grown by a company that doesn’t care about their health enough to use uncontaminated water.

I’m sorry, I just don’t accept that this is how it is now.

Is Pangram the solution? I don’t know. It’s not perfect, that I know.

Personally, I think Substack didn’t go far enough.

Because here’s what people don’t stop and think about. The “tells” of AI aren’t just on the page. They’re not just in a Pangram scan. They’re in user behavior. How long did it take to read that article? If you read a five minute article in 30 seconds and left a comment, too? Again and again? That’s AI.

Same with posting. I see it on Medium. People posting two or three posts a day, 7-10 minute reads each. That’s AI. I’m sorry, there is no human being who can keep up that pace seven days a week except maybe an ex newspaper journalist. Maybe.

And I think Substack should have paired up Pangram with user behavior.

Find the people commenting on more posts than one human can read. Pair that up with the Pangram score on their comments. Find the people posting long posts every day while also commenting on more posts than one human can read. Pair their usage patterns with their Pangram scores. It paints a different picture than a score alone.

And deal with it.

If we don’t want Substack to turn into a nightmare of rampant AI like Medium and LinkedIn, something needs to be done.

The problem isn’t so much people using AI it’s people abusing AI.

I don’t know the solution. All I know is that it’s not wrong to care whether a Cowichan sweater was made by a human or a machine and it’s not wrong to want to know the same about words. I hope they figure it out. Because no one else has so far.

This is the way the world ends.
This is the way the world ends.
This is the way the world ends.
Not with a bang, but a whimper.
T. S. Eliot, The Hollow Men