It is yet again the start of a week! Happy Monday if you celebrate.

You’re reading Please Stay Human, the newsletter figuring out how to live in the new and strange world of AI.

And help out another soul: forward this email to somebody you know who’s worried about AI.

People don’t speak AI

At this point, really, it’s more that AI doesn’t speak human. Let’s take a look at this from an almost annoyingly non-technical perspective:

Way back when computers were first created, people had to learn the structured, codified languages used to program them.

Those people were the first programmers.

Then programming languages got better and better. We moved from punch cards to easy-to-understand, structured blocks of text. Instructions that were simple for a human to understand.

People could talk to computers. But now, it’s the other way around.

Today’s AIs, like ChatGPT, are trained on an unimaginably large corpus of text data. Countless lines of human-written text. And when you open up your AI and ask it to write something, it always uses that special tone of voice.

AI-written text shows up everywhere these days. Websites, posts on X, especially posts on LinkedIn, comment sections, spam emails. And every single time I see it, I know it’s there, and it’s annoying. It’s annoying because

  • It’s not realistic. Nobody talks like that.

  • It shows a lack of caring and effort.

I’m only looking at issue A right now: nobody talks like that. Take the famous ChatGPT em dash for an example. It’s supposed to be having a human-like conversation with you. But somehow, it’s not trained to sound like a human would.

Screenshot from some random ChatGPT user

No human would speak in this rhythm or tone, no human would pause like that, in those places. It’s not just the em dash. It’s the cadence that’s off. We hear people speak all day. We consume human-made content.

And the AI slop doesn’t match in rhythm, cadence, tone, or, as it turns out, punctuation. The em dash was almost solely the domain of the fiction author before AI (as far as I’m aware.) But now you can find it in spam emails and bot comments. It’s gaining a bad reputation, too: ChatGPT made it so common that people WON’T STOP posting online about how the em dash is a curse that means something is AI written.

No, it was just uncommon online before now. A lot of good books are full of em dashes if you look.

This is all just a temporary issue. New AI models are built, trained, and released at a hectic pace, and they sound more human every time. Eventually, they will be pleasantly non-annoying.

Or maybe it will never go away:

Edit out the em dash? Wrong.

You could have ChatGPT write you something, then go through and touch it up here and there until there were no em dashes left.

But that is a bad, bad idea.

When you sit down to write a thing - like the piece I wrote above - you think it through. You mess up, go back and fix it. It is an outpouring of thought onto screen.

If the AI does the thinking, and you just touch it up, you’ve changed your role. You’re not the author, you’re the editor. And those are two very different things.

So please stay human.

Microsoft: these jobs are safe from AI

Microsoft did a big study on what industries will be most and least impacted by AI. This is the “least” 40.

Worried about AI taking your job? Try learning rail-track laying or embalming soon.

If you’re an embalmer reading this right now then good for you, hit reply and tell us.

Funeral 🤣

Hey, we’re Please Stay Human. Our goal is to help you face the reality of an AI-powered world without feeling like you have to pick up embalming as a side project.

You can always hit reply and tell us what you think. The emails go right to our inbox.

This email is by humans, for humans, so if you know somebody who’s feeling the loom of AI, please forward this to them.

Every single word in this newsletter was typed by fleshy human fingers on plastic Macbook keyboards. We make a lot of mistakes and fix most of them.

Keep Reading

No posts found