Big Change Here In Substackland!

A really positive step towards transparency

Over on my much larger Substack I’ve had an “AI Statement” sitting on the main page for quite some time. It’s a policy document, right there with a couple of others. On that Substack I sell subscriptions, and it seems to me that as a seller I have an obligation to clearly communicate to buyers exactly what it is that they are buying.

In essence, my statement declares that all of my writing is 100% human. That I don’t use AI for anything.

Indeed it would be patently absurd for me to use AI in my writing, because I write not only to communicate, but to learn whatever subject I am writing about. Using AI would be completely counterproductive to that goal.

Plus, if a ‘creative writer’ is using AI to ‘write’ well, then that ‘writer’ isn’t a writer at all. Indeed, if the AI use isn’t disclosed, and the work is claimed to have been written by him or her, that’s fraud. Harmless fraud I suppose, if the work is given away for free. Not harmless if the work is sold to unsuspecting buyers who are led to believe that it is a human written work.

That’s my position.

But, not everyone thinks like me.

And that’s A-OK.

If someone wants to do their ‘writing’ with AI, that’s no skin off of me. The ‘writing’ will kind of suck, it certainly won’t be original, but hey, not my problem. Other people should do their thing. See, that’s the Libertarian in me.

But, what I do want to see, is transparency.

Honesty.

If someone ‘wrote’ something with AI, that’s fine with me. But be honest about it. Don’t try to hide it. Think AI writing with Large Language Models is the future? OK, then mark your ‘writing’ as such and defend your position. That’s honest and honorable. Hiding isn’t. Deception isn’t.

Photo by New York Said on Unsplash

I read my AI statement again today, and I think I should add a little something to it. I probably won’t, because I’m lazy and I don’t think anyone reads it anyway, but I probably should.

I hate promoting things, but I try to force myself to do a fair amount of it.

I shamelessly self promote my own writing, primarily over on Facebook. (OK, this is a lie, I do the promotion, but it isn’t without a twinge of shame.)

But I also promote other writers. Over on Facebook, but not only there. Here too on Substack, both in “Chat” and “Notes.” When I encounter something I like by someone else, I do my best to remember to promote it. Because other writers need promotion of their work just as badly as I need promotion of mine. And I know how very hard it is to convince others to share one’s written work.

My AI Statement should include the fact that I will not knowingly promote writing that I believe to be AI generated. Assuming the work is without disclosure (as almost all AI ‘writing’ is) doing so would make me complicit in a fraud.

So, I won’t knowingly do it.

And I should probably put that in the Statement somewhere.

But, again, I’m lazy, so I probably won’t get around to it.

I do remember writing it somewhere in the past though, somewhere lost in the vast mountain of essays I’ve written through the years.

The trouble is, of course, discovering for oneself what is AI generated and what isn’t AI generated.

There are tells that most writers know, and if the work is polished to the point of dullness, on and on. But one of the easiest ways I’ve found is by looking at volume.

I write fast. I write really fast. Much of my work for my State’s Legislature for a couple of decades was writing, and the volume was huge. I learned how to write fast.

So when I see a writer cranking out essays or books faster than ten of me could ever do, well that’s a pretty damn good indication that AI is the thing doing all that work. A dozen essays a day? Yep, that’s AI. A half a dozen books in a month? Yep, that’s AI too.

But, here’s my trouble…

Yep, I think that I’m pretty good at spotting AI. Both through the marks of artificial ‘writing’ and with some ‘writers’ through their utterly absurd volume of output.

I feel bad though. I worry that I could be wrong. I fear that maybe that fellow I know really did write that essay, and that I should be helping him by promoting it. What if I’m not doing so because of an error on my part?

This week, Substack (the company) to the rescue!

I started writing here on Substack a really long time ago in Digital Years. (I think of Digital Years sort of like Dog Years, the time really flies!)

Back when I started, there wasn’t much here on this platform. As a writer you came here, typed up an essay and hit send. It emailed that essay out to subscribers and archived it on a simplistic website. That was it.

Then things changed.

Recommendations were added. Chat was added. Notes were added. The app became a thing.

And none of these things seemed as if they were really all that helpful to writers. (Except maybe Recommendations.) It seemed as if in some really big ways the Substack company was becoming less about writers and more about becoming the next big social media thing.

In a way, that was sad. But in another, really important way, it simply acknowledged the fact that Substack must be able to compete in today’s digital world. The company had to embrace elements of social media in a social media dominated world.

This week though the company made a big improvement that I think truly will be beneficial to writers.

It contracted with the AI detection company Pangram to allow, not require, but allow readers to run things published here on Substack through Pangram’s AI writing detection service. Think of it as AI ratting out AI, pretty cool.

This is beneficial to writers, not because I think AI ‘creative writing’ should be banned or some nonsense like that, but because it will bring about transparency. It will help readers separate writers from pretend writers. And that is very good for all the real writers out there.

It’ll be beneficial to readers too. Because this is a paid platform. Much is made about buying subscriptions here in order to support writers. And that is a very important thing. But one wants to support writers, not people posting AI generated slop while claiming to actually be ‘writers.’

And it’s really beneficial for me. Because now I’ll be able to double check writing that I suspect is AI generated.

I won’t rely completely on Pangram, because I know that false positives are a thing. But, if I already believe or suspect that an essay has been AI generated I’ll be able to use Pangram to help confirm that. And that will be extremely helpful to me in my goal of not promoting AI ‘writing.’

One final thing that I should mention. As with most everything here on Substack, the Pangram check thing is voluntary. Readers certainly don’t have to use it, and writers can turn it off so that it won’t be available to check their posts. Of course the fact that a ‘writer’ has turned the feature off speaks volumes.

And of course, one can wander on over to Pangram and create an account. Thereby gaining the ability to check anything the heart desires. I did that myself earlier this morning.