The work builds the muscle.
I love AI. I love exploring what it can do, seeing how it can help my life, multiplying my outputs by ten-fold (no exaggeration).
But as I used AI more and more, I realised that it was shortcutting a crucial part of the process, and making me worse off for it.
If you've watched the Matrix, you'll be familiar with a scene in the first film where Neo gets his introduction to the Matrix—what it can do, how it can be manipulated. One of the things that he does is he learns martial arts; well, the correct thing to say would be that he has martial arts downloaded into his brain.
This is much the same as what AI does for us—it may not download anything physically into the structure of our grey matter, but it does deliver things to us on a silver platter. And what I discovered was that this was creating a crucial gap in my mind.
Part of learning anything is not the end result of having learned that thing, but the process of learning it.
I was blessed with a great sense of direction. When I get to a new city, I like to know how to get around and to understand the map of the city in my head, so I'll go for a walk or a run. After a few days, I'll have learned the major routes to and from where I am—and then I'll explore the side streets. The turns I never took, the alleys that look like shortcuts. I often take wrong turns or get lost in unexpected dead ends, but also, I learn how to get around with a map fairly quickly.
The same is true for all types of learning—the dead ends and misunderstandings are just as important as the final piece of "correct" knowledge. In fact, it's precisely those dead ends and mistakes that contextualise the knowledge and move us from fragile amateurs, whose knowledge is extremely situation-specific, to experts, whose understanding of the topic transcends any specific context and who can manipulate the information and use it at will. A chess master is greater than a chess amateur, even one who has memorised 1000 opening moves, because the master can be creative with her knowledge, can see things that the amateur can't see.
It's not enough to know how to play chess; one has to have spent years playing chess to become a master.
This is the problem with AI—it creates fragile amateurs, people who mistake the knowing of facts for expertise.
But AI is just a tool, and its usefulness depends less on what it is than how we use it. Some people use it in a fairly shallow way—"write me viral hooks". Others, though, understand that it works best as a thinking partner. Because that's what great about AI—it's like an extremely knowledgeable friend who you can talk to for hours about anything and who will never get tired of your questions. So instead of asking to be creative with writing (which it is, but only mildly, and only if you hurl abuse at it), you can ask it to help you be creative at thinking, suggesting angles you may have missed, or questions you may not have thought to ask.
One area I've used it for to great effect is building reading lists on certain topics. I wanted to study communications and human performance to a level few humans alive had studied it to, and so I put together a reading list of about twenty five books and asked it critique the list. Then, together, we iterated and iterated (in conjunction with different AIs—Claude, ChatGPT, Gemini) until I ended up with three separate lists of ~50 books each. And those are just my first 9-12 months of reading. Alone, it would have taken me hours or days (that I simply don't have the luxury of taking at the moment) to put that list together, and I'm still the one reading the list, processing it, and applying it to my life and work.