Are Popular Substackers Using AI?

58% of us are concerned about the use of AI, but most of us can't tell what's AI and what's human. Maybe it's time we learned?

Mary Oliver said she doesn’t know exactly what it means to pray, but she knows how to pay attention, how to fall down in the grass, how to kneel in the grass, how to be idle and blessed, how to stroll through the fields. How to watch a grasshopper eat sugar from her hands, jaws moving back and forth instead of up and down.

Doesn't everything die at last, and too soon? she asks.

It’s from Summer Day and she ends with this beloved question.

Tell me, what is it you plan to do
with your one wild and precious life?

I don’t know what I want to do with my one wild and precious life.

Write, I thought.

But I don’t know anymore.

Because now there’s AI and people don’t know the difference. We don’t. We think we do. We pretend we do. And a few people actually do. But mostly, we don’t.

I wrote an article on AI art. Shared twelve photos and asked readers which are AI and which are human. It was meant in fun. But also? To show people how truly bad we are at knowing what’s AI and what’s human. Mostly, people failed abysmally.

It’s even worse with words. Sometimes, people come along and tell me it’s because AI has become “so good” now. They’re wrong. It’s not that AI is that good.

It’s that we’re not paying attention.

It’s that we don’t know what to look for.

And then people write pieces telling us “the signs” of AI, like specific phrases or god forbid, an EM dash and we eat it up but it’s all clickbait. Because AI only uses the phrases it learned from humans.

But there is a way to tell. If you know what to pay attention to. Like Will does…

---

Let me tell you about Will. He’s an author who also writes on Substack.

His most recent book is The Science of Storytelling, in which he says storytelling is the secret to everything. Because if you can tell a good story, his theory goes, the world opens its arms to you. That book became a Sunday Times Bestseller.

And I think that was true.

But the book came out right before ChatGPT.

I want to tell you about a piece he wrote. It’s called “Scamming Substack” and I think you should read it after you finish reading this piece. I’ll link it again at the bottom. I hope it gives you the same “holy crap” moment it gave me.

He said two weeks after starting his Substack, he noticed that many of the bestselling Substacks had an interesting thing in common. He called it “a surface profundity.”

They all had thousands of likes. So he decided to study them, see if he could learn anything. And the more of that style he read, the more of them Substack fed him. Because yay algorithms. It’s what they do. Give you more of what you read.

After a while, it dawned on him that many of those “bestsellers” have a few things in common. Frequency, for example. How are they publishing so often? he wondered. And then it hit him. Oh. Of course. Duh.

So he did a little experiment of his own. He reviewed a bunch of those “bestselling” pieces and came up with a ChatGPT prompt that churned out articles so similar to the “bestsellers,” you’d be hard pressed to know which are AI and which are human.

He’s talking posts with 12K hearts.

He compares actual “bestselling Substack” posts to ChatGPT posts and its chilling.

Something about that almost makes me want to weep. Throw in the towel. Because why the hell are writers spending hours to maybe get 50 or 100 likes or subscribers when someone types in a ChatGPT prompt and gets freaking ten thousand hearts?

He even shared the prompt and I cringed because some people are going to use it to vomit more AI unto the world. Yay more bestsellers.

#tired #sotired

Meanwhile, you and I pour our hearts out and wonder why no one is reading.

But I think there’s merit in sharing his experiment, his observations and the prompt he came up with. Because maybe it will help some people see the real difference between human writing and AI writing.

Because the real “sign” of AI isn’t a freaking EM dash for god’s sakes.

He also shared 5 telltale signs that a writer is using AI and they’re not words, phrases or punctuation. Once you see the signs of AI, you can’t unsee them.

I hope, anyway.

And I think we need our eyes opened. Collectively.

---

I read an article in Wired called The AI Backlash. It said before ChatGPT came out, 38% of people were concerned about the rise of AI in our day to day lives. One year after ChatGPT launched, that number shot up to 58% according to a Pew Research study.

58%. Which means most of us have concerns about the rampant use of AI.

I want to remind you that not all AI is the same. Okay?

Some AI can help us. It can help all of us, help humanity as a whole.

AI can look at medical imaging and pick out cancer cells. At Mount Sinai they’re using AI with 94% accuracy to predict the development of liver, rectum and prostate cancer. So maybe someone you love is diagnosed before stage four. AI can help deaf kids learn to read. It’s being used to halt the decline of endangered species.

But ChatGPT isn’t that.

ChatGPT is generative AI. So is Google Gemini and Claude and Midjourney.

It’s generative AI that people are concerned about and rightfully so. Generative AI amplifies harmful stereotypes. It’s bad for the environment. Every day, ChatGPT uses the same amount of energy as 250,000 average households. More energy than 117 countries. It’s depleting our fresh water supply and costing people their jobs.

If we’re going to destroy the planet, can we please do that trying to save us from ourselves? But no.

We’re destroying the planet and costing people jobs so we can have AI-generated questions on LinkedIn, AI-generated podcasts on Spotify and AI-generated images on microwave noodles at the grocery store.

Last October, Wired said AI Slop is Flooding Medium. To the tune of almost half the content on the site, and that was 8 months ago. That hurts real writers. Financially.

And now it’s here too.

And it’s not floating to the top, we are pushing it there.

Because we don’t know the difference.

But we can learn to tell the difference. So I hope you’ll go read Will’s article, Scamming Substack. And I’d love to know what you think.

To pay attention, this is our endless and proper work.
—Mary Oliver