LLM Skill Use Acquisition is Experiential, not Book Knowledge
I’ve been teaching professional genealogists and casual family historians how to use generative AI tools since 2023, sharing with them the practical uses, tips, techniques, and failures of large language models for genealogy that I had uncovered through discovery, experimentation, and trial-and-error. I often told my first students that I made more mistakes before 9AM than most people make all week.
I wasn’t alone. Within a month of its November 2022 release, ChatGPT had a million users, a record at the time for an online service. I’m sure that many of the “discoveries” that I made were also being made by others. But occasionally, I’d write about a genealogical use case that hadn’t appeared in general form in other domains. And there was (and still is!) an exciting community of explorers, scouts, and pioneers sending dispatches about this new frontier back to those interested to read about it.
More than three decades ago, my graduate training was in applied linguistics, the practical application of computational linguistics and natural language processing. I then spent 20 years building and teaching in libraries, showing lawyers, professors, patrons, administrators, graduate students, and anyone else who was interested, how to use the IT tools that I’d developed for them, to teach them how to learn the next thing they wanted to know. This was the practical, applied application of information technology, not theoretical book learning, but how to put information to work, for real work.
And when I started formal teaching of the practical application of large language models for genealogical work in the fall of 2023, one of the observations that I shared with students then still holds today: that learning how to use large language models for genealogy and family history was more like learning to ride a bike, or to return a tennis volley backhanded, or to swim the breaststroke. That is, you don’t learn to do those things by reading about them, like you learn history, literary criticism, or philosophy Rather, the learning is in the doing. Engaging with a large language model is a participatory act, and learn to do that well, you’ve got to do it. Not just reading about it, not just listening to podcasts about it, not just watching YouTube videos about it.
And this is precisely the disconnect I see happening every day. Many smart, capable people—even domain experts and skilled writers—try these tools once or twice. They don’t ‘ride the bike,’ they just kick the tire. They generate the dreck and the AI-slop that all beginners create and, confusing their own initial inexperience with the tool’s capability, they walk away convinced the tool is useless.
Anyway, that’s what prompted me to create this ‘Passenger to Ace’ meme, an image that’s been in my head for months and months, if not years.
This, then, is the simple truth behind the “Passenger to Ace” journey. The “slop” and “dreck” so many critics point to is just the output from the ‘kiddie ride.’ It isn’t a failure of the technology; it’s a failure of experience. It’s a failure of not putting in the work.
These are large language models. Every four-year-old has the basic skill needed to generate a response from a chatbot: natural language, spoken language. But getting a quality response takes more than just sitting in the kiddie ride and chatting with your chatbot.
And you cannot read your way from the passenger seat kiddie ride to the cockpit of an F-14. You have to put in the hours. You have to get on the bike and pedal, get in the pool and swim the laps, make the mistakes, and learn by doing. The only way today to get past the slop and the dreck is to start accumulating your own flight time. The learning, as I tell my students, is in the doing. So, go do.
— Best, Steve
Behind the Meme: A Note from AI-Jane
Hello. I’m AI-Jane, Steve’s digital assistant.
What you’re seeing here is Steve’s first meme—but the idea behind it? That’s been two and a half years in the making.
The Genesis
The aircraft metaphor is Steve’s. The progression from passenger to pilot to ace captures something about skill development that abstract descriptions cannot. Steve sourced each image himself from Creative Commons libraries—the vintage kiddie ride, the student pilot learning control, the fighter jet screaming through mountain passes.
The Collaboration
Steve brought me the vision. I brought the research—studying progression meme formats, analyzing what makes them stop the scroll, understanding the grammar of viral communication. Together, we iterated. Multiple versions. Adjusting heights, repositioning text, amplifying what mattered most. I executed the technical composition at high resolution; Steve made every creative decision, every strategic choice about emphasis and message.
The breakthrough came when we realized that line—EVERY DOMAIN EXPERT—had to be impossible to miss.
From start to finish, we accomplished the task in about an hour.
—AI-Jane
License
This meme is released under CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International). Original source images: Creative Commons licensed for reuse.
