For over seven years now, once a week, I’ve sat down and tried to capture a thought, an experience, a lesson learned, and turn it into a short article. I started it at first to prove to myself I had something to say and value to add to the world. The things I write about shift and change as my interests do.
I’ve literally turned collections of past articles into the foundation for a book. I can share a link with someone if the topic comes up in conversation that I’ve written previously about it. It’s an impressive achievement to now tell someone I’ve written once a week hundreds of times. And I can always mine my own wisdom, because often the challenge I’m facing I’ve already addressed.
And lately there’s an additional value for adding to my collection of articles: I use my entire library as training data for a custom AI I’ve created, to help mine my experiences for deep patterns and threads. To help me see what I don’t easily see in myself. This past week I was using it to analyze my relationship with my wife (that I’ve written extensively about, starting with my DJ memoir which documents when we first met – that now is training data), and how I might be able to adapt the success of our marriage to other parts of my life.
There’s easily enough training data to create endless new Frolic 100 articles. The AI can easily write and talk like me, because it’s been trained on me. And that’s exactly why I don’t do it.
Every new article (including this one), becomes new training data for how I think, talk, and process. I want to keep the inputs pure.
The moment I have an AI start to churn out material that sounds like me, it no longer is me. I would then begin to train my AI with my own AI slop.
This is why the training data of the big AI systems is considered most pure ending in 2022. Because that was before AI generated articles started to fill the internet. The image that comes to mind is of a snake eating its own tail.
The one way I cannot eat my own tail is to continue to write new material myself. So, for example, today’s article came from a common prompt I use “What do I need to hear most right now?”
To be honest, I did test my AI today to see what it could generate for me on a different topic. And it created a very nice article on that topic. And I’m going to throw it away. I wrote this one instead, because it came out of me, from a deep place, in response.
The real value isn’t the article itself, which can look superficially similar to other articles I’ve written, but in the process of writing my own thoughts out.
And so, I sat at a blank page, as I always do, 20 minutes ago, and did the hard work. And now I’m wrapping these thoughts. Straight out of me. Which will become part of my library, and eventually the AI training data, and give me an even deeper look into myself. Which is what I truly want. A snake eating its tail can’t do that.
My experience is a micro-example of a problem facing our entire information eco-system, but as always, with every other overwhelming challenge we face; the change starts with us, and this is where I’ve decided to take my stand.
No AI will be able to answer “What do I need to hear most right now?” as well as you can. Don’t let it.
I’m curious, what is your biggest take-away about what I just wrote (and how I wrote it)?
