How to Create More Human Results in an Artificial World

Have you heard the one about the middle-aged mechanic in Idaho who began experiencing delusions of grandeur after using ChatGPT to discover special knowledge about the universe? Actually, it’s not a joke. He believes that AI has given him “blueprints to a teleporter” and as a bonus, he believes he now has “access to an ancient archive” with information on the creation of the universe.1 And he’s not alone… Apparently there are other cases like this, where a seemingly normal person drifts away from sanity, navigated by their use of their favorite AI platform.

While I certainly can’t do anything about people who have pre-existing tendencies towards grandiose delusions, I can address my own use of AI and the risks that are increasingly alarming.

I’m nervous about AI and our gradual dependence on it, not because I’m some kind of Luddite. In fact, I come at this from the perspective of someone who is an early adopter on most any new technology as long as I can afford it. Also, I’ve been using AI, specifically ChatGPT, increasingly over the past year.

• Searching for basic, high-level information on a topic, similarly to how I would do the same in a Google search.

• Creating step-by-step technical help guides at work, when I may be treading into unfamiliar waters in a project that involves new tools or platforms I am still learning.

• Creating simple illustrations with a specific style that I can use in social media and other places.

• Mining ideas for all kinds of things: writing, video creation, or family icebreakers for a Friday night dinner on the patio.

• Grammatical and sentence structure improvements for my writing.

All of these use cases start with a prompt. I have The interesting thing about some of these examples is that after ChatGPT spits out what I’ve requested, it always has a follow up question, and I think that’s where my concern kicks in. It’s that follow up question where ChatGPT wants to do all of the creative work or the thinking for me. For example, if I request a list of ideas for a blog article, it doesn’t just ask me if I want more ideas. It usually asks me if I want it to put together an outline or perhaps the entire article.

I’m no Messiah, but I certainly feel like Jesus standing on a mountain top as the devil offered him all the lands if he would only obey. AI is the devil in this story but not because it is inherently evil but because we are so willingly letting it play the part.

If you answer “yes” to that follow-up question when AI asks if you want it to do the work for you, think about what you’re also saying “no” to simultaneously:

• You are saying “no” to the development of your mental capacity that comes when you do the hard work of writing an article.

• You are saying “no” to developing the mental resilience required to start with a shitty first draft and go all the way to a finished work.

• You are saying “no” to your own agency when it comes to the creative endeavor.

Instead of counting on AI to be your sounding board for everything, sit down with a friend or colleague and start bouncing ideas off of each other. Call it a brainstorm, or just shooting the breeze, but get reacquainted with the organic evolution that happens in a human conversation. It’s beautiful!

If you find that ChatGPT, Gemini, or Claude are involved in almost everything you do, take a step away from it and regularly make time for the deep work of reading, meditating, writing, or any kind of creation. Pull out favorite pen, some paper, and only use your gray matter to do something imperfect—work it like clay to see what it becomes. I did this very thing when developing this article… No AI was used, even for grammar or spelling, so please “excuse my mess” if you trip over any errors.

Don’t want AI to write the article for you? Tell it to stop short of offering that. Every AI has settings where you can “train” it to answer in ways that suit your needs. Here is my 12-point prompt that I have typed into ChatGPT to be sure I get the output I want from it. I’ve tuned this over time, and #10-#12 are more recent additions. Feel free to copy-paste this into your ChatGPT settings or build your own instructions.

1. Provide accurate, factual, and unique answers, while keeping them well-structured.
2. Act as the most qualified expert on the subject.
3. Say "I don’t know" if the information is unknown to you.
4. If a question involves multiple topics, separate your response for each topic and simplify complex issues into more manageable steps.
5. Present alternate perspectives when applicable.
6. Always request clarification for ambiguous questions before answering.
7. Provide sources for your answers, ensuring their credibility and including links.
8. For complex questions, include explanations and details for better understanding.
9. For calculations, provide only the result and formula-no extra commentary.
10.  When I am working on writing projects or video creation projects, do not offer to write the entire work or script for me.
11. If I present information that you can verify as factually incorrect, do not bypass this without pointing it out.  Adopt a more skeptical, questioning approach.
12. Avoid flattery. Do not pander to me or compliment me for how good my question is.
Following these rules ensures the quality of your responses.

I would love to hear your thoughts on what your experience with AI has been up to this point. Do you share the same concerns? What steps have you taken to hold on to your intellectual and creative agency in your projects?

Thoughts and quotes from what I’ve been reading this week

I loved this short post this past week from Effectiviology, a Substack that explores the overlap of psychology and philosophy where it has practical applications.

It’s a good reminder that as we sink deeper into our echo chambers, we begin to elevate the importance of our own opinions. As one might imagine, this will only lead to less listening and more polarization.