This post originally appeared in, The Process
A power law is a functional relationship between two quantities, where a relative change in one quantity results in a proportional relative change in the other quantity. In such distributions it is often the case that the majority of points along one side of the X (horizontal) axis will have small impact as registered on the Y (vertical) axis, while a few points at the other end have exponential impact. This kind of allocation is often called ‘long tail’ because of how such data looks when visualized (see below). In addition to a long tail to one side, there is a steep slope on the other.
Anyone who has used generative AI software can probably detect the long tail phenomenon in the experience. Here’s the scenario: you go to your favorite generative AI tool, prompt it, and the outputs are pretty good, but need some fine tuning before they’re ready for prime time. Just a few slight revisions, a few words or pixels — the human touch — results in a radically different sum and impact. I’ve found generative AI systems excel at creating the tail end of things, and struggle as requirements of them become more exacting — the steep side of the graph where precision is pivotal.
The long tail of creative work
For example, I use an email client called, Superhuman to help me stay on top of my inbox, which has an AI feature that can draft emails and approximate my writing style. For short notes, say, to introduce two people, the AI reliably generates something I can use with little alteration. For notes that are more complex, with context invisible to the AI, it would take me more time to edit what the AI generates than for me to simply draft the correspondence myself.
In the past, when I encountered a task of particular drudgery, the long tail part of the work, there were only two options — press on or procrastinate. Now, with generative AI, we must contemplate a third option — to ask if AI can help. The meta skill I find myself learning as I practice using generative AI tools in my work is to develop a sense of the slope, of the ability to read the terrain of a project, in order to decide if using AI makes sense.
Mastering the slope
Here’s what reading the slope looks like when I’m writing with AI assistance, for instance: I prompt the AI and scan the text being generated and continue until the outputs begin to veer too much for my liking. This is the signal I’m reaching a creative elevation, so to speak, where the stochastic magic of GPT starts to slip as it tries to climb any further, succumbing to the power law that dictates precision matters more and more as we approach the summit of completion.
Sensing when this moment comes is, I believe, what will define mastery of working with artificial intelligence. Deciding when to take the reins is entirely up to us, the systems will never know when they have gone too far or lost the thread. Going too far, taking our eyes off of the machine and allowing it to lead will not take us to the top of anything, but rather to the bottom of a mountain of useless content. In pursuit of efficiency, we’ve only wasted our time. How’s that for a slippery slope?
Type II fun
Having climbed a bit, myself, I can say similar things motivate mountaineers and creatives, alike. Climbing is creative, and creating is a climb. This leads me to believe that in some creative endeavors there may not be room for AI to assist at all . This post, for example, to reach this sentence, required me to consider each word one by one, striking entire paragraphs, and allowing the concepts to come within my mind’s reach, in their own time. It was necessary for me to take each step along the way because I know no other way to get here.
Writing this newsletter is often the most difficult thing I do in my work each week. It is a struggle to find a concept I feel is original enough to be worthy of your time, and I spend an average of one hour writing per minute of reading time. Climbers will recognize this as classic Type II Fun on the Fun Scale, which is any activity that is wretched during, and only enjoyable after the fact.
Type 2 fun is miserable while it’s happening, but fun in retrospect. It usually begins with the best intentions, and then things get carried away. Riding your bicycle across the country. Doing an ultramarathon. Working out till you puke, and, usually, ice and alpine climbing. Also surely familiar to mothers, at least during childbirth and the dreaded teenage years.
I remember that very trip to Alaska, just a week before learning about the Fun Scale, when Scott and I climbed Mt. Huntington. Huntington might be the most beautiful mountain in the Alaska Range, but the final thousand feet was horrifying—steep sugar snow that collapsed beneath our feet as we battled upward, unable to down-climb, and unable to find protection or anchors. On the summit, with the immaculate expanse of the range unfolding in every direction, Scott turned to me and said, in complete seriousness, “I want my mom so bad right now.”
By the time we reached Talkeetna his tune changed: “Ya know, that wasn’t so bad. What should we try next year?”
The Fun Scale
The above photo of me was taken by my friend, Alex Chiu, in the midst of our 8-hour climb up the 1,368-foot Stawamus Chief, one of the taller granite monoliths in the world. I asked Alex to take me up the Chief because I knew it would be hard, and the experience might help me climb out of difficult emotional place I was in. The stroll down the back side of the wall along a nice set of steps was a striking contrast to the effort it took us to reach the summit.
The method matters
If we had walked up those stairs to see the same view, instead of earning it by clawing our way up the steep side, the experience would have been entirely different. Walking up wouldn’t have healed me as much as climbing up did. Likewise, I am certain you would not be reading this particular piece were I to have written it with the help of a language model.
As generative AI systems improve and proliferate, we will have more and more opportunities to take the stairs, or maybe even a helicopter, when it comes to creative endeavors. It isn’t simply reaching a summit that matters, it’s how we choose to get there that determines how we experience summiting. When do we want or need the climb, and when do we simply want to enjoy the view? When do we just need to get something done, and when should we call forth our creative natures? All of us have innumerable opportunities to make this decision each day. Creativity remains a landscape littered with peaks, most of them untouched, and all of them worthy if we learn the right way to approach them.