A few weeks ago I did a Shirt Show episode, and one of the things we talked about was keeping a human in the loop. Because ultimately, a junior separator learns by failing. You fail over and over again, and then you get better at what you’re doing. °FAI: Human in the loop AI, Keep the Failure.

The index job I set up wrong

I remember one of the very first index jobs I set up. As I ordered the screens in the order they were listed, then I ordered the inks in the order they were listed. I figured the screen room would look at the screen number and find the ink that went with it out of the list. I didn’t understand the entire process all the way to the end.

When I got to press, there were screens with the wrong inks in them. A bunch of them. I created that mess. I went, “oh, I misunderstood,” and I learned. It was a stupid mistake. And it was mistakes exactly like that, over the course of years, that kept me learning and kept me improving.

If AI eliminates the challenge, it eliminates the learning

I read something the other day: “Writing is hard because the writer is learning. If AI eliminates the challenge, it also eliminates the learning.”

People talk about AI removing humans. But taking the human out is taking out the place where we fail. And when we remove where we fail, we narrow our possibilities and our training ground for life. Because it’s not just writing and separations where you fail and then learn.

It’s everything. It’s growing up, relationships, friendships, employee experiences, running a business, driving a car, traveling and finding new foods you like. All of it hinges on positive experiences that let you grow and negative ones that let you learn.

Everyone is having their own AI experience

What I see in AI is the potential to augment and enhance whatever you choose to apply it to. That’s an individual choice. You can live a wonderful life without it. You can live a wonderful life with it in very targeted, specific places. And you can let it infiltrate every corner of your life and still have a wonderful life. Individual choice, all of it.

But take away a human’s ability to fail completely, and with it goes creativity and innovation and inspiration and a richness that artificial anything cannot replace.

And honestly, if you are interested in building an AI system, if you take away the ability of a human to fail, what’s the AI left to learn?

Do I use AI to help me write? Yes.

In Fahrenheit, I do. In Throwback Thursday, I do. I don’t use it beyond that, and I use it for very specific reasons. On Throwback Thursday I’m learning how to build more complex RAG systems. On Fahrenheit I’m learning how to translate complex, variable-driven coded projects into two paragraphs that deliver information to an audience on a Friday. I’m learning how to use an AI to convey a message, which is the same thing as a writer learning how to write through failure.

These things can be equal and completely different at the same time. I can say a writer uses writing as a training ground for becoming a better writer. I can say I use AI as a training ground for conveying a message. And I can say AI replacing human writing is a dangerous, terrible long-term goal, and I can say I use AI to help me write on certain projects. It’s okay for us to hold all of these conflicting, polarizing opinions about one thing. It’s healthy, in fact.

Human in the loop AI

In the beginning, when you don’t know anything, you fail. As you experience the failure and repeat, you start developing an intuition, a really valuable skill that is only developed this way.

This is why it’s important to learn AI. Healthy skepticism, and intuition grounded in reality comes from repetitive use and repetitive failures.

Taking the human out of the loop doesn’t just remove the labor. It removes the place where the learning happens. Be highly skeptical, highly curious, and grounded in intuition that comes with experience, all at the same time.

Tools used:

• Claude (drafting and voice matching for °F AI on Fridays)
• RAG build on Throwback Thursday (learning retrieval, one failure at a time)