Movin' On Up — CIqSi

There is a scene in Star Trek IV: The Voyage Home where Scotty sits down at a 1984 Macintosh and begins typing — fluently, confidently, ten fingers, the works. Scotty is from the 23rd century — the crew has traveled back to 1986 — and the joke is that he's bewildered by the primitive equipment he finds there: he tries to talk to the computer first, gets nothing, then resorts to the keyboard with visible reluctance. What I couldn't stop thinking about is the choice the director made. Scotty types well. A man from the 23rd century, encountering a keyboard for the first time in his life, types like a trained secretary. If he'd hunted and pecked — two fingers, eyes down, the way anyone would attack an unfamiliar instrument — the scene would have been funnier and considerably more plausible.

The precedent was right there in the original series. "A Piece of the Action" — the episode where the crew visits a planet that modeled its entire civilization on a 1920s Chicago mob manual — has Kirk trying to drive a flivver. Flivver is itself a word so archaic it has spent the sixty years since broadcast quietly fading from English, which is its own small joke. Shatner's Kirk is visibly faking it, and the scene works precisely because of the fakery. Nobody expected him to know how to drive a Model T. The gap between confidence and competence is where the comedy lives.

James Doohan almost certainly couldn't type either. He was a male actor in 1960s Canada. Typing was clerical work — for women, for secretaries, for the transcription pool. The confident ten-fingered performance Scotty delivers in The Voyage Home is a performance within a performance: Doohan, who probably never learned to type, playing Scotty, who certainly never learned to type, typing fluently on a machine neither of them had ever seen. The keyboard is a prop. The plausibility is a missed opportunity.

I suspect my own descendants will find typing equally quaint. If I weren't so accustomed to organizing my thoughts through a keyboard, I could easily be dictating this prompt. The keyboard is a transitional technology. Most transitional technologies earn exactly as much nostalgia as they deserve.

The Hierarchy and Its Variable

Every civilization has organized itself around a hierarchy, and every hierarchy has needed a justifying variable. Race, religion, national origin, sex — the list is long, and the structural pattern is consistent: take something partially outside a person's control, dress it as a measure of worth, and sort people accordingly. The modern world flatters itself that it has abandoned the obviously specious ones. What it's kept is ability.

There were probably millions of people with the cognitive equipment to be Leonardo da Vinci. They weren't born in Renaissance Italy. They didn't have access to the Sforza court, or Verrocchio's workshop, or the specific intellectual ferment of 15th-century Florence. The meritocracy, when it rewards demonstrated ability, is measuring something real — but it's also measuring context, and the two are much harder to disentangle than the vocabulary suggests.

"Industrious" and "smart" and "skilled" sound like things you did. They are also — heavily — things that happened to you: where you were born, who had bandwidth to teach you, whether the tools you needed were accessible at the moment you needed them. The achievement and the circumstance arrive bundled together, and we hand out the trophy to the achiever.

When the scarce resource that defines a hierarchy becomes cheap, the hierarchy destabilizes. That is exactly the current situation.

What the Machines Are Taking Over

AI is absorbing the tedious cognitive tasks — the research, the first synthesis, the initial draft, the routine debugging, the literature review — the parts of knowledge work that required training and time but not, to be precise about the word, judgment. The tasks that defined a junior analyst, a capable generalist researcher, a mid-level developer three years ago are increasingly within reach of a well-prompted system.

The backlash against this is running about six months behind the technology. The posts demonstrating how you can "tell" when something was written by AI are demonstrating, almost without exception, the limitations of a model that no longer exists. The dunks on GPT-3's reasoning were already out of date when GPT-4 shipped. The comparison is almost always the same: a poorly prompted legacy model against the most diligent, expert human performance. Even by that unfair standard, the human rarely wins on speed and first-draft density. The inevitable plateau that critics keep forecasting hasn't arrived. There is no particular reason to think it will.

The Rush Concert and Human Slop

The concern over AI-generated content is reasonable. The comparison to human output is less flattering than its proponents assume.

I've been watching mobile phone footage from the Rush Fifty-Something tour — handheld, ambient-lit, shot on consumer hardware and uploaded directly to streaming platforms. Millions of viewers are studying every one of Anika Nilles' fills in footage that is far from broadcast-ready. Nobody is calling it slop. They're calling it access. Fifteen years ago that footage would have been unwatchable. The audience adapted its tolerance because proximity was worth the tradeoff.

YouTube has been an unrelenting torrent of human-generated content for twenty years. So has every social media feed, every comment section, every conference hallway recorded by someone with a phone and a theory. We normalized the human variety entirely and barely noticed. AI output will follow the same arc: tolerated first for what it enables, then normalized, then invisible.

Photoshop meant you couldn't trust photographic evidence. CGI meant you couldn't trust video. Humans have been lying in print for centuries, one character at a time, drawing on the training data of every story their culture told them. A language model generates text the same way: one token at a time, from the current context, through a trained neural network. The mechanism is not new. The scale is.

Libraries have largely gone away. Newspapers are going, which is close enough. Television is destabilizing, balanced reporting first. None of this is unprecedented. William Randolph Hearst's papers helped initiate the Spanish-American War. The rogue media organs causing so much contemporary alarm are the dying successor generation of outlets that were doing the same thing a century earlier with moveable type instead of engagement algorithms. What changes across media transitions is the velocity, not the function.

The Ancient Skill

Which brings me to where I have actually landed.

I'm editing a blog now. More precisely: I'm editing, and the blog is a byproduct. The first draft is produced by a machine. My contribution is the prompt that shapes it, the judgment that accepts or rejects what comes back, the revision that corrects its idioms when they drift from mine. The writing here is getting better — because I'm getting better at generating it. The prompt is the craft. If I weren't so attached to keyboards, I could easily be dictating the prompt. My descendants will find the keyboard as quaint as Scotty found the one on that 1984 Mac.

This is what moving up the stack looks like from the inside. The bottleneck shifts from execution to judgment. The role becomes less about performing the thinking and more about directing it — recognizing what's right, naming what isn't, knowing the difference well enough to instruct a system that will take you at your word.

My ambition is to be as prolific as Mark Twain with perhaps one percent of his skill. The editing is the skill. It is as ancient as writing — Mel Brooks, in History of the World, Part I, placed the first art critic approximately thirty seconds after the first cave painting. Editing is the quieter branch of the same impulse: not the sharp dismissal, but the shaped one. Not criticism. Direction.

Scotty tried to talk to the Mac. The Mac hadn't learned to understand him yet.

It has learned now.

Published by Geordie