AI Etiquette.

Recently, a comment on Hackernews inspired me to think about AI Etiquette. The number of times I've seen improper use of AI, people asking me whether I've asked Copilot already, or seeing someone upload pictures of people's faces to commercial AI-tools like ChatGPT without their consent... it's time we set some boundaries and rules for thee.

Feel free to disagree, and please let me know when you do. I'll update this blog post accordingly as I come across new cases.

What (not) to do when using AI.

When sharing isn't caring.

Unless prompted by the other, don't share a complete message from ChatGPT with someone without context, or blatantly copy-paste it somewhere. Just because it hit home for you, or you're amazed at what it's capable of, doesn't mean it will resonate with the receiver the same way. I'll admit I've done this too. And I get it—you're enthusiastic. But as AI language models become a commodity, sharing walls of AI-generated text, especially seemingly purposelessly, feels like sharing slop. If you aren't getting a reply, now you know why.

Don't push people towards using AI.

This is a big one. I'm looking at you, corporates who can't keep up with the market and need to show they're right on top anyway.

It's great that you've seen the light. These Generative AI models nowadays can do absolutely amazing things when you know how to put them to good use. However, everybody has their own way of working. Some may have a preconceived notion that AI will do more harm than good. And that's okay. We may educate people around us, we may increase your AI-literacy, but we may not force people into workflows they aren't comfortable with. Before asking someone whether they've considered using AI to solve a problem, let's think critically whether you think they'd actually want to hear that suggestion from you.

A cat in a tight spot will make weird jumps. That is to say, if you're pushing people to use AI in their daily life without proper education and willingness on their behalf, odds are they'll be misusing AI, sharing things they shouldn't be sharing, perchance uploading highly confidential documents to a commercial party like "Open"AI. At that point you're much farther from home.

^{Sometimes a bit of Dutch creeps into my English still—my apologies.}

Just don't push people towards using Generative AI.

Instead, give people the support they would need, in order to pull themselves in when they see fit.

To generate or not to generate images, that is the question.

Should you use AI-generated imagery? You can, but preferably use it just for decoration purposes. If your AI-generated image should be a focal point (unless the aim is artistic), you should probably use a real photo or human-created artwork, or you should make sure to reiterate the AI-generation of the specific image you're envisioning until it appears flawless. And then still: if you're using AI to generate an image that you'd like people to stare at, you'd best make sure it's clearly distinguishable as being AI. You wouldn't want to deceive anyone, would you?

I've actually been asked, professionally, to create a video using AI where we'd then swap the people's faces with those in management.

Don't do that.

Ever.

_{^{Apparently this is not obvious.}}

We care about your privacy (except I really do).

Every time you send any message to a commercial or otherwise closed-source Generative AI instance, by default you should assume your data is what's making them money. In the European Union we luckily have laws like the GDPR that aim to protect us from data-hungry companies. However, these laws are only as good as the people that follow them.

Privacy is often a weird one to argue about, since there's often the counter-argument of "they can have my data; I have nothing to hide". Let's just say that while you're free to decide for yourself, you have to respect others who think otherwise.

Don't share information with AI-tools that can in any way be linked to a real person. Preferably don't use ChatGPT as a therapist either: this data is incredibly personal and valuable. You may have nothing to hide now, but what if your government has a regime-change? Now you're suddenly an enemy of the state. Thank you very much for sharing your deepest and darkest fears.

Don't assume it's AI-written, just because of em-dashes—like these.

Can you believe it, there are actually people who enjoy writing with fancy symbols like em-dashes. If you're on MacOS, you can quickly type en-dashes (–) with Option + -, or em-dashes (—) with Shift + Option + -. I love 'em. Use 'em all over the place. And yes, I am using them slightly less now that it's a common "sign" that AI was used to write the text. At the same time, as a writer, I won't drastically change my ways purely due to the existence of AI tools. I'll find other ways—like how I've added the "Written by a human"-label at the top of each blog post, right before writing this one.

Looking back, most of these rules I've written effectively come down to understanding people's rights and needs, and just not being a prick. It's not that difficult, really.

Hol' up—you're an AI Engineer, aren't you using AI for your own website here?

Do I use Generative AI for these blog posts? Not to write any sentences, no. To brainstorm and inspire? Absolutely! They're a great solution to writer's block—that is, in my way of working. Just be aware that AI models tend to give similar outputs to similar inputs. It might "think outside the box" here and there, but know that it will likely think outside the box in the same direction as it did before.

The images for my blog posts are often AI-generated. They're not the main focal point—I actually hope you quickly skip past the images and read the text I write instead. I make sure the images look cartoonish, not realistic, not to deceive anyone. Let me know what you think!

As for pushing people towards using AI, this one might need some explaining from me. After all, I am professionally coordinating an AI Centre of Excellence and giving workshops to increase people's AI-literacy across entire organisations. I'm pretty much selling AI wherever I go. I will stand by my principles however: I love teaching people what they can do, but everyone has their own way of working. Of course I do think nearly everyone can benefit from using AI in a way that supports their preferred way of working. But you're always free to say no.

If you ever see me deviating from my own AI Etiquette, let me know; I will do my best to inform you about it beforehand, for instance for this purposely AI-written blog post from 2023. As I come across more elements worth adding to my AI Etiquette, I will be updating this blog post continuously.