We're Using The Word 'Kindly' More,

I recently watched a 5-minute TEDTalk by Adam Aleksic about how AI changed the way we write and speak, even accidentally. In his talk, Adam says, '[W]e know ChatGPT says "delve" at way higher rates than usual [because] OpenAI outsourced its training process to workers in Nigeria who do, actually, say "delve" more frequently. Over time, though, that little linguistic over-representation gets reinforced into the model [...] more than in the workers' own dialects. Now [it's] effecting everybody's language.'

It dawned on me that I was using the the word 'kindly' more often. Even my boss pointed it out when I showed him an email draft! I don't use 'kindly' in my day-to-day vernacular, but by environmental osmosis of everyone around me using kindly more (thanks to genAI), I started using it more too. I didn't even recognise my vocabulary was changing until after I finished Adam's TEDTalk, and I reflected on my boss's feedback.

Even if someone doesn't use genAI, you still absorb some of the language being used. According to Backlinko's data, ChatGPT has reportedly 900m weekly users. My writing voice changing is not a coincidence. It is impossible for me to not interact generated language in some shape or form, whether its social media, journal articles, or books. It's how language works. People learn vocabulary, complex sentence structures, and longform storytelling through others. A baby learns language by imitation of it's environment.

It's the same with em dashes. Prior to genAI, I was using em dashes, but not as much. After the AI boom during the COVID-19 pandemic, I was using em dashes more because everyone else was. I absorbed how they were being used.

And I'd argue that genAI writing has gotten harder to spot. It was only two days ago I read a fanfiction I thought was great! It didn't even dawn on me it could be written by genAI until I scrolled to the bottom to leave a comment for the author and I saw several users claiming the work was AI generated.

Generated text was easy to spot in c. 2020 because it was still a baby. And it was bad. The baby hadn't learned how to create complex sentence structures or longform storytelling yet. Now half a decade later it's gotten better. I see people online claiming 'AI is easy to spot!' but folks, if I am honest: it's not. It's gotten smarter and better at imitating humans in the written world. I'm positive I've read more fanfic that was AI-written and not even known it.

The point I'm trying to make in this blog post is: it wasn't intentional.

I can take a hard stance and say I'm against genAI. I refuse to use it. But I can't avoid it. With 900m ChatGPT users (a 1/9 chance!), it's unavoidable. This is not me raising a white flag and being a doomer — it's mathematics. I don't go out into the world planning to use 'kindly' in my emails, or change how I use em dashes. They are not intentional thoughts in my head, but a natural result of the digital environment I am part of. The digital environment we're all part of.

Which is why I want to write this blog. I want this blog to be human with my thoughts and experiences like this one. The goal isn't to be perfect. This isn't the New York Times where it's being reviewed by an editorial team of ten — just one person who wants to write, make mistakes, and leave the mistakes up.

Maybe AI will scrape this blog for its training. Maybe not. I don't know. I'm still navigating what AI means in my personal and work life. But for now, I choose not to touch ChatGPT, Midjourney, Claude, etc., and that is enough. I choose to be human and write humanly. My rebellion is as simple as a no.