Something something AI and Substack

If you’re invested in this platform, then you’ve probably come across this post.

The 5,000+ likes and 1,000+ comments probably tell you this is getting a lot of attention. Why wouldn’t it? After all, this is one of the founders of Substack talking about what’s changing on this very platform. The comments have been rather mixed.

Since these AI platforms can often get a false positive, there is an option for you to state how you made your content. I’m not entirely sure how that works. I presume they want your homework. If you’ve written something of elaborate non-fiction, you will need to provide the research matter on the topic. If you created some art, I presume you need to provide some details on how your art was created.

You’ll probably see that repeated often, and platforms like Substack are likely going to use the view that a few false positives is okay.

There is research to say that GPT-based detection is biased towards non-native English speakers. You could also say that non-English speakers will have a higher degree of AI usage, if only because they’re writing on a platform that has a predominantly English-speaking audience.

Either way you look at it, this is going to be detrimental to those who don’t have English as their first language.

There is some evidence of AI misclassification, likely because the natural writing styles of some neurodivergent individuals mirror the statistical patterns AI detectors look for.

No AI tool is perfect, and that’s something that’s going to be the case for as long as we resort to large language models. However, regardless of that, the nuance is simply not there. Chris Best himself has previously taken a post that was critical of Substack, claimed that it was generated by AI, and summarily dismissed every single point of that post.

If the people you disagree with have some chance of being flagged as somewhat AI-generated, that’s just as easily an opportunity to completely dismiss both the post and the author. We have people dragging up tweets from a decade ago with little to no context and suggesting cancellation. No one is going to look at whether Panagram can mistakenly classify AI text. They’ll simply look at the fact that it has, and that is enough.

So what does that mean for those who have had content posted that has been deemed AI? Most likely, they’ll lose their subscriber base, and if not that, they’re going to have to deal with the harassment that comes with it.

I think it’s very easy to tell if something is AI-written. I have definitely used it to write some sarcastic posts. I used AI to write pretty much all of this.

I did it because I wanted to sound like all those subscription and growth hacking peddlers. And AI writing sounds just like that. I’d be really curious who didn’t pick up on that. But I’m not going to chastise them. I’ve just spent far too long on LinkedIn and in the tech industry to not be able to spot it.

And so little energy combatting the hate speech on this platform. If you want to see some of the vitriolic comments that I’ve come across, I wrote a short post on that.

That’s all I have to say on the topic. I’m currently nursing a very bad wrist injury, and I have used AI dictation software to write this. I’m only human.

Some of you might have noticed that I haven’t been on Substack Notes. I truly have lost a lot of my interest on this platform, and I’m putting a lot more of that energy into ramping up my own blog on my own website. Stay tuned, I’ll drop a link. Hopefully when I’m ready to use a mouse again.