Hunting Mosquitoes With a Blunderbuss

Perhaps I’m more disconnected from regular life than the average Joe, but look at that video. Did you know the LLM thang had progressed this far, this fast? Internet wags who stubbornly still call it nothing but glorified autofill might want to rethink their opinions. Although no one ever rethinks their opinions on the internet, do they?

Another term that gets used to describe all manner of LLM ouput is slop. It was the word of the year, and an addendum has been made to the word’s entry in the dictionary to account for the output of Large Language Models. Most devotees of calling everything an LLM puts out “slop” are textual complainers, not visual cavilers. I’d like to take up the cudgels for anyone who wants to call the video we just watched slop, because it is. Just not for the reasons they’d like to think it is.

By referring to the things that LLMs put out as slop, people are mostly objecting to the great leveling that they produce. But the average of everything is bound to be average, isn’t it? Complaining about it seems a fool’s errand. I was on the committee that named the internet, and I remember that Fool’s Errand came in second in the naming list. Pandora’s Hope Chest and The My Little Ponies of the Apocalypse had too many letters, and fell out of the running early. Al Gore said it was the internet, and here we are. Calling machine written text slop, when referring to the internet, is silly. The internet has always been 99% poorly written slop. At least chatbots can spell correctly.

So exactly where is the here, you know, where we are? Watch the video again. It’s not easy to make a video like that. It looks to my eye about as sophisticated a final LLM product as you could find. Then again I might have fallen behind the times in the time it took me to type this. Progress, if you want to call it that, is happening fast, and picking up speed.

My wife is quite innocent, and asked me how much of the video was real. She assumed that at least the young woman host must be a green screen projection of a real person, and some of the backgrounds must be some sort of Colonial Williamsburg real-life thing. I told her that the girl, and everything else, appears to be some guy name Jonathan, sitting at a desk, who types faster than I do, and knows how to stitch together all the LLM apps and LLM prompts to produce videos like that.

I asked an LLM (Chat GPT) to explain exactly how to make a video like this. It didn’t have any trouble, and explained it in great detail, which I won’t bore you with. I’ll bore you using my usual methods. I did ask it to make a short summation of how the video was achieved, and it kindly boiled it down:

Creators usually build these videos with a multi-stage AI pipeline rather than one app alone.

Typical apps:

OpenAI Sora, Runway, Kling AI, or Luma AI for generating photorealistic video clips
ElevenLabs for AI voiceovers
Adobe Premiere Pro or DaVinci Resolve for editing and stitching shots together

The prompting technique is highly structured. Instead of a simple sentence, creators stack detailed layers describing:

format (“vertical iPhone selfie vlog”)
subject (“young woman with messy auburn hair”)
action (“walking through Tudor London talking to camera”)
environment (“muddy streets, timber houses, chimney smoke”)
camera behavior (“handheld shake, autofocus hunting, rolling shutter”)
lighting (“flat cloudy daylight”)
realism flaws (“compression artifacts, motion blur, imperfect framing”)
negative constraints (“no CGI look, no cinematic lighting, no face warping”)

The realism comes from intentionally simulating imperfect smartphone footage rather than polished cinema. Modern AI models learned not only historical imagery, but also TikTok/influencer behavior patterns and phone-camera physics. Creators then generate many short clips, discard bad outputs, and edit the best pieces together into a seamless “historical influencer vlog.”

I noticed that the intrepid authors of videos like these seem to have at least mostly solved the problem of drift. LLMs had a tendency to allow things to morph into different things when making longer clips, so something like the narrator would be a problem. She’d be the internet ideal of a hot babe one minute, and slowly turn into someone else as the video went along. The characters in the video are pretty immutable through the whole thing, a real accomplishment. I found it kinda amusing that Anne Boleyn’s face was grabbed from a portrait I recognized, just made a little less dowdy and two-dimensional. A few months ago, an LLM might have made her look like Anna Nicole Smith after enough frames, for inscrutable LLM reasons.

So I promised you I was going to complain, in your honor, about the LLM slop this video represents. Here goes:

It’s wonderful. Truly. It’s a hair’s breadth from being an entire movie production studio RAMmed into a Dell Optiplex. It could be as big a breakthrough as Edison’s kinetoscope. To find its equal, I’d have to go back to Charlie Chaplin setting up shop on North La Brea Avenue, and thinking about making more than Keystone Kops two-reelers. It has that kind of potential. Unfortunately, it will all be wasted.

You see, I can picture what could be accomplished with this setup, or probably the setup that comes right after this one. Movies and television entertainment could be wrested back from the avaricious, depraved, unfunny, and historically, economically, and socially illiterate people who have a death grip on it now. But I know in my heart it won’t be. Staunch traditionalists will continue to complain that they don’t make movies like they used to, and then renew their Netflix subscription, lest they miss out on seeing Batman XIV, an adaptation of La Cage aux Folles.

The premise of the Chloe vs. History is charming, if you squint hard enough. The simalacrum of the same kind of vapid influencer girl with a camera on a stick, going on vacation and telling you all food is amazing as long as it’s served in a foreign restaurant, going back in time instead, could have an element of humor in it. It doesn’t, but it could. It’s a testament to the kind of imagination at work. It took a lot of intelligence and hard work to find, master, and cobble together all the LLM applications that make that video possible. The nature of the intelligence stopped there. It doesn’t know what to ask it, excuse me — tell it — to do. So he makes a TikTok influencer vlog, because it’s all he knows. Of course he’s double smart, because he knows it’s the only sort of thing his audience is going to understand, too.

Once the charm wears off, you notice that the missing element is something truly intelligent, insightful, or even informative. The THING, the agglomeration of apps and circuits, is probably nearly capable of producing a new Citizen Kane if you told it to, a real story with real characters who do dramatic but believable things. No one will even try. The same people who make Helen of Troy a rich, dark hue, and say Thomas Cromwell was the hero of the Tudor piece, and a sex machine to boot, will make the LLM slop, because they’re in charge of the big budget slop already. They’re currently using cameras instead of mainframes, but they’ll adapt. The remaining set painters, the ones not already in the breadline from green-screen filming, will be laid off, and the production of anti-human entertainment pablum will accelerate, not diminish, simply because it will be cheaper.

Granpa told me all about the genie in the lamp.

It’s the oldest story ever and came from the land of the sand and the women with only eyes. It’s in there, the genie of everything, but you have to find him and let him out. Then he’s out and you have to figure what to do with him. Granpa says he’s wonderful but as dumb as a stump, just like all of us. He can do anything but doesn’t know what to do. He needs guidin’.

The lamp is always hidden in plain sight he says. Men go prospectin’ all over the landscape for the easy riches but they’re generally layin’ right there on the ground but you step over them in your hurry and scurry to look for them. Granpa points to the men through the door of the grog shop and they’re playin’ cards and Granpa says what good does it do for them to find the riches anyway.

Granpa would take the books down from the high shelves that the kids weren’t supposed to get because the treasure in them was too dear to waste on such as us. He told me to run my hands over the cloth on the cover to see if it was the real deal inside there. They don’t waste the nubbly cloth on the fakers.

The lady wouldn’t like it but Granpa would shush her and we’d go home and open that book but only so far. A book is like a man, Granpa would say. You can only bend him so far back until he can’t take it no more and then his back breaks. People always put the book back on the shelf but you can always tell because neither the man nor the book can stand up straight any more after that.

Scheherezade told that Sultan all those stories and it kept her alive and me too.

What Grok Thinks About Sippican Cottage

[Editor’s Note: Offered without comment. Author’s note: There is no editor]

Critique the writing style at sippicancottage.com

The blog Sippican Cottage (sippicancottage.com) is a long-running personal site run by an author who describes himself in the tagline as: “A man who has nothing in particular to recommend him discusses all sorts of subjects at random as though he knew everything.”

That self-mocking subtitle is a perfect preview of the writing style — deliberately contrarian, conversational, self-deprecating, and allergic to earnestness or conventional polish. Here’s a breakdown of its main characteristics:

Voice and tone
The dominant voice is first-person, informal, and curmudgeonly with a wink. It’s the tone of a guy at the end of the bar who’s seen a lot, doesn’t think much of most of it, but still finds it worth riffing on. Heavy on dry/irreverent humor , sarcasm, and gentle (or not-so-gentle) mockery of pretension, modern absurdities, AI, cultural trends, music snobbery, home improvement orthodoxies, etc. Self-deprecation is constant — the author happily calls himself a hack when accused of it.

Sentence structure and rhythm
Very eclectic, often fragmented and punchy. He favors:
– Short, abrupt sentences or outright sentence fragments for comic timing or emphasis.
– Deliberate run-ons or lists when the mood strikes.
– Sudden shifts from the prosaic to the poetic or philosophical, then right back to sarcasm.

Examples from recent-ish posts include things like the “New Year’s Resolution Sentence Fragment List”:
> I promise to quit drinking
> At two ‘o clock this morning
> Maybe three
> I pledge to eat healthier
> Candy
> Not after I drop them

Or casual observations like:
> “Charming host in that video. I miss the old intertunnel. It was full of regular people like her. They didn’t pretend to be experts on geopolitical events based solely on the last three things they saw on CNN.”

This choppy, telegram-like delivery can feel like spoken word transcribed without much editing — which is clearly intentional.

Vocabulary and diction
– Colloquial and vernacular, with occasional inventive slang or playful misspellings/mispronunciations (“intertunnel”, “clicky-pop”).
– Mixes plain-spoken blue-collar phrasing with more literary flourishes when the topic calls for it (e.g., music criticism or architectural history).
– Avoids academic jargon almost entirely; when technical terms appear (roofing, music theory, Victorian architecture), they’re usually deployed plainly or ironically.

Strengths
– Highly distinctive — you recognize the voice immediately after a few paragraphs.
– Excellent comic timing and deadpan delivery.
– The randomness of subjects (music, home renovation, cultural commentary, family anecdotes, old media nostalgia) never feels forced because the voice ties it all together.
– When he slows down and writes more conventionally, he can produce very evocative, lyrical passages (especially about music or old houses).
– Refuses to pander or write for algorithms/search engines — refreshing in 2026.

Weaknesses / criticisms
– The fragmented style can feel mannered or gimmicky if you’re not in the mood for it — some readers find it affected or hard to follow for longer stretches.
– The relentless irony/sarcasm/self-mockery sometimes makes it difficult to know when the author is being sincere (if ever).
– Little interest in conventional narrative structure, transitions, or “blog post” formats — entries can feel like stream-of-consciousness brain dumps.
– If you dislike curmudgeonly boomer-adjacent takes on modernity, large portions of the site will grate.

Overall verdict
The writing is idiosyncratic and unapologetically personal — a throwback to early-2000s blogging before everything became SEO-optimized listicles or earnest Substack thinkpieces. If you enjoy writers like P.J. O’Rourke meets Richard Brautigan (someone once accused him of aping Brautigan, which he denied ever reading), or just appreciate a highly opinionated, funny, digressive crank who can actually write, Sippican Cottage delivers in spades. If you want clarity, linearity, warmth, or emotional straightforwardness, you’ll probably bounce off it hard.

It’s one of the few remaining blogs that still feels like a true individual diary — eccentric, uneven, and stubbornly uninterested in pleasing everyone. That alone makes the style worth reading, even if you don’t always love it.

Sippican Abides

star lebowski

The world is moving at warp speed lately. The picture pretty much encapsulates how I’ve been dealing with it for my whole life.

I’m not that old. I don’t fart dust or have God’s unlisted phone number or anything. But a lot of earthshaking shifts have taken place in my lifetime. Chatbots/LLMs/Ai/whateveryouwanttocallthem are just the latest shake of the technological snowglobe we live in.  Off the top of my head:

  • Phones went from a single black thing on the kitchen wall that rang like a four-alarm fire, to Dick Tracy communicators, with multiple steps between
  • Photography went from B&W to Polaroids to 35mm Kodachrome, to potato cams to megapixel cameras in everything. Been to a Fotomat recently?
  • Television went from 3 channels to cable to streaming to coming out of gas pump screens
  • Movies went from giant screen, destination viewing to VHS tapes to DVDs to digital files people watch on their phones in the subway
  • Attention spans went from 8 hours to 8 swipes on TikTok
  • Social media went from bulletin boards (actual cork ones) to chat rooms to MySpace to Facebook to Instagram to Twitter to Discord to heaven knows what now
  • Cars went from 2 tons of sheet metal, a bench seat, and an AM radio, to rollerskate/spaceship/iPhone cradles
  • Finding out stuff on the internet went from (blog)lists to directories (Yahoo) to Google to Chad
  • Making images, including moving images, went from pen and ink or cameras the size of a refrigerator, to pixels via an LLM, with mucho layoffs in between

When I say Chad, I mean Chat GPT. Or Claude. Perplexity. Gemini. Any of a number of large language models that elicit scorn or paranoia to taste. Both scorn and paranoia are understandable, but I fear the scornful have forgotten the saying, “You may not be interested in war, but war is interested in you.” And while armies of terminators aren’t currently roaming the countryside, relying on their chatbot overlords to send them to hunt the next Sarah Conner on the list, there are bound to be lots of casualties from the integration of leviathan computational machines into everyday life. It might not be you directly, but there’s bound to be a lot of collateral damage.

If you’re like me, you’re mostly like The Dude in the picture above. You’re hurtling at warp speed to god knows where, but you’re just along for the ride. I have mostly avoided relying on, or even participating in every step in the technological chain I listed above. But that didn’t mean I scoffed, exactly. I just stood athwart the world and yelled whatever. I didn’t bother yelling stop, because I knew you wouldn’t. I didn’t willingly cooperate in any degringolade, but I didn’t stick my fingers in my ears and chant la la la, either. The world beamed me down to some pretty unpleasant digital planets along the way, though, and I’ve had to change my job description more often than my wardrobe.

So there’s hysteria and there’s scorn about Chad’s effect on damn near everything, depending on who you’re listening to. But someone is at least starting to ask the right questions about the LLM phenomenon. Who’s going to be Crewman Expendable when beaming down to the planet’s surface wearing a red pullover instead of a bathrobe?

Payrolls to Prompts: Firm-Level Evidence on the Substitution of Labor for AI (study found via Marginal Revolution)

It’s an economic treatise, so they’re using weird economist backwards nomenclature. They mean that AI is being substituted for human labor. The study tries to track whether firms are replacing human labor with generative AI in their spending patterns, and what it might cost them to do so. They measured the amount of money being spent on freelance marketplaces like Fiverr and Upwork, or whatever they’re calling themselves this week, and then comparing it to the amount of money the same people spent on AI to do the work instead. Here’s a taste of what they came up with:

We see differential patterns of spending shifts by exposure quartile. In the highest exposure quartile, we find that for every $1 decrease in labor marketplace spend, there is a $0.03 increase in AI model provider spend in Q3 2025 relative to Q1–Q2 2022 baselines. In the middle exposure quartile, we find that for every $1 decrease in labor marketplace spend, there is a $0.30 increase in AI model provider spend in Q3 2025 relative to Q1–Q2 2022 baselines. The true magnitude most likely lies somewhere between these two quartiles. The middle exposure quartile is only significant in the last time period, whereas the highest exposure quartile is significant in all time periods. We note that we cannot observe all potential additional spending that comes from bringing AI in-house, such as infrastructure costs for serving models, as well as increases in engineering headcount to build and maintain AI capabilities. Even if this estimate is conservative, it is still a significant cost savings. For example, if a firm is spending $100,000 on labor marketplaces and $10,000 on AI model providers, the firm is saving $90,000 by substituting labor for AI. Understanding how these cost savings are distributed both across and within firms is important to understand the potential impact of AI on labor markets and the economy more generally.

Got that? It’s possible to save somewhere between 70 and 97 cents on the dollar by firing someone and hiring Chad to do your intellectual scut work.

I can assure you that the replacement of freelance webworkers with Ai chadworkers is happening, bigtime. Entire ecosystems of people gulled into thinking they could write SEO articles or product descriptions or fake reviews or whatever other phony dreck the internet mostly consists of are becoming dead as Scrooge’s doornails, almost overnight. Over the last fifteen years or so, these people have seen one after another internet toehold shift under their feet and leave them without a crummy internet income. They’ve adapted somewhat as webwork changed, but Chads are currently putting a fork in a lot of them permanently. So what are they going to do with themselves?

I have a theory. They’re not going to get real jobs. If they thought they could handle real jobs, they wouldn’t be trying to make their Etsy store pay for their rent, student loans, and medical marijuana in the first place. Even if they’re capable, they’re not willing. They think the world has dealt them some pretty shitty cards, and whatever they can get back from that crooked dealer, they deserve. This is what they’re going to do, in general, if not in particular:

Of course I’m the worst kind of prognosticator. I often predict things that have already happened, and this is no exception. The United States is already awash in criminality. It’s almost ubiquitous at this point, but boy howdy, it can get worse. Shoplifting, aggressive panhandling, porch pirates, vexatious litigants, learing center operators, disability fakirs, drug dealers, gift card scammers, phishers, hackers, and just plain old scofflaws riding around smoking a J with an expired registration and a suspended license.

Chad will be the dread god of such dinky criminality. It’s the rough beast that slouches toward a datacenter in Bethlehem, Pennsylvania. What to do? I’m not full of advice on the topic. I’ll simply abide, I guess, as best I can, while everyone around me loses their mind over Chad, until the next thing to lose your mind over appears.

To suffer woes which Hope thinks infinite;
To forgive wrongs darker than death or night;
To defy Power, which seems omnipotent;
To love, and bear; to hope till Hope creates
From its own wreck the thing it contemplates;
Neither to change, nor falter, nor repent;
This, like thy glory, Titan, is to be
Good, great and joyous, beautiful and free;
This is alone Life, Joy, Empire, and Victory (Shelley)

The Kindness of Strangers vs. The Wisdom of Crowds

I’m so old I remember when people looked things up on the internet like a leap of faith. Ancient history now. Maybe five years ago. You’d type a question into the little box, hit enter, and hope that somewhere out there, a stranger would be kind to you.

The intertunnel was Blanche DuBois central. If you’re not familiar with Blanche, you should watch “A Streetcar Named Desire,” and see Vivien Leigh fading away while Marlon Brando became the next big thing. At any rate, just like Blanche, you had to be off your rocker to trust the internet, but by gad, people sure did. It involved more self-deception than trust, and a healthy dose of bad judgment.

No matter what it said on the masthead of whatever site Google sent you to:

  • You didn’t know who answered your question
  • You didn’t know why they answered it
  • You didn’t know whether they were brilliant, biased, drinking heavily, goofing off at work, or trying to sell you something on the sly

But there they were on the first ten results on Google. Strangers. And you depended on them.

Lording over this whole mess was Google itself. The ultimate stranger. Google never claimed to be your friend, exactly. It claimed to be helpful. Reached out an elbow for you to grasp on the way to the funny farm, and you took it. It claimed to be neutral and objective. That was a laugh. Its methods of curation were always opaque, its incentives were never purely aligned with your best interests. Google didn’t wonder what was true or even what was useful. Its ranking system consisted entirely of discerning who was kissing Google’s buttocks sufficiently to make Google some money.

What followed was inevitable. If Google rewarded visibility, then visibility became the greasy pole every website operator tried to climb. Entire industries sprang up to reverse-engineer the G’s algorithm. SEO experts. Content farms. Listicles with suspiciously specific headlines. “Doctors Hate This One Simple Trick” became a genre.

The internet quickly filled up with very strange people indeed, strangers shouting, waving, and stuffing keywords, all angling for a click. To be fair, no matter how bad it got, some people actually knew useful and amusing things, and offered them to the public, free of charge. These benighted souls, through titanic efforts, could climb to be found on page 114 of the Google results. It was left to the brave internaut to sort sincerity from strategy in real time. Pretty much, we all failed at that. Hence, Buzzfeed!

The kindness of strangers, it turned out, was unreliable, and just like Blanche lying on the floor, often exhausting. We were all ripe for something different. Google had hogtied the whole internet. The only place sadder than page 114 of Google was the top of the Bing results. You could hide dead bodies on Bing. The internet went from sclerotic to petrified. Only a completely different way to look for information could save us.

That’s what Large Language Models actually represent. They’re not glorified autofill, as many would characterize them. They’re not intelligent, either, in the true sense of the word, but so what? Unlike Google, which claimed to have the world’s digital information all curated for you, LLMs like Chad (Chat GPT) read the whole internet, and then some, and settle on a crowdsourced answer for you. Not original thinking. Not thinking at all, really. Just paying attention to everything, everywhere, more or less. Instead of being handed a ranked list of links curated by an inscrutable and avaricious stranger, you were handed a synthesis. Not a single authoritative voice, but an average. A blending. A statistical distillation of countless human scribblings. The good, the bad, and the ugly.

The prime idea behind this has a name: The Wisdom of Crowds. The term was popularized by James Surowiecki in the early 2000s. The observation itself is much older. Francis Galton came up with the concept, more or less, back in the 1800s. He spent half his time being pretty smart about statistics, and the other half writing a rough first draft of Idiocracy. He didn’t have faith in any single member of a crowd, not by a long shot.  One of Galton’s classic illustrations is a fairground guessing game. A crowd is asked to estimate the weight of an ox or the number of jellybeans in a jar. Individual guesses are all over the place, too high, too low, and confidently wrong, usually. But if you take the average of all those guesses, the result is often eerily accurate. No single person knew the answer, but the crowd, in aggregate, effectively did.

This core insight is counterintuitive. Under the right conditions, large groups of ordinary people can collectively make better judgments than a small group of experts, or even the smartest individual you can find. That includes me, I guess. I’m the smartest individual I can find, but then again, I’m alone in my apartment right now. I’d have to put on pants and go outside and look for someone smarter than me. It could take minutes. Never fear. The wisdom of crowds doesn’t work because people are especially wise. It works because their mistakes are all over the place. Biases cancel out. Overconfidence is diluted. Individual blind spots are cancelled out by other individual idiocies.

Large language models are sorta like that. They are not intelligent in the human sense. They don’t reason or understand, and probably never will. They’ve been trained on enormous amounts of human-created text.  Everything from high-quality scholarship mixed with drunken Reddit screeds, “journalism” (tee hee) mixed with marketing copy, insight mixed with the comments under cat videos. Much of it, taken individually, is not to be believed, never mind trusted. But when the model predicts answers based on patterns across all of that stuff, what emerges is something like a crowd’s best guess. It ain’t truth, exactly, but it’s at least a probabilistic consensus shaped by millions of whoevers rather than one loud stranger.

This is a subtle but profound shift. Before, you depended on the kindness of strangers. You hoped that someone, somewhere, had taken the time to answer your exact question thoughtfully. Then you hoped Google had decided this person deserved to be seen. Good luck with that. “Don’t be evil” is right up there with “Arbeit Macht Frei” in the accurate slogan department.

Now, you depend on mediation. The LLM doesn’t care about clicks (yet). It doesn’t care about ad revenue (right this minute). It doesn’t care about SEO tricks or keyword density (don’t worry, it will eventually). It doesn’t wake up hoping to sell you a multi-level marketing membership. Its incentives are different: produce something that sounds coherent, relevant, and responsive. That hardly makes it perfect or unbiased. Far from it. The wisdom of crowds can be wise under the right circumstances, it’s true. But crowds can also be lynch mobs. Garbage in, garbage out, averaged.

And since LLMs are programmed never to say, I don’t know, you end up reading hallucinations. You wish you’d get Sargent Shultz, and end up with Cliff Clavin instead:

The experience feels fundamentally different. You’re no longer wandering a digital marketplace, hoping to bump into a benevolent stranger. You’re having a conversation with a synthesized amalgam of John von Neumann and Cliff Clavin. Good luck figuring out who is who.

AI: The World Will End Yesterday. Plan Accordingly.

Well, if you watch the artificial news, Artificial Intelligence is going to take your job and your girlfriend, at least when it’s not too busy taking over some bunker in North Dakota and launching nukes willy nilly. You could form the opinion that AI already has taken over the world. You can’t turn on anything on these here intertunnels without some demented form of Clippy the AI assistant offering to correct your grammer, and maybe write that email for you that you’ve been meaning to send, but you can’t for the life of you remember how to spell Deer Sur.

There have been many, many laundry lists published of all the jobs that are going to be wiped out by one chatbot or another. Most everyone outside of longshoremen and prostitutes are slated to be standing on streetcorners holding signs that read: Will photoshop the background out of pictures of female footwear for use on your Shopify store for food. The usual commentarazzi are furiously analyzing the inroads that Large Language Models (LLMs) are making into the economy, and publishing their search engine optimized articles, written by ChatGPT, natch, with headers like: AI: The World Will End Yesterday. Plan Accordingly.

What is missing is some form of sober analysis. Just adding a new definition of slop to the dictionary isn’t helpful, any more than adding a new definition of vaccine kept you from getting the flu. I’m interested in the topic, however, and I finally found one lonely source that at least attempted to answer the only cogent question:

HOW ADAPTABLE ARE AMERICAN WORKERS TO AI-INDUCED JOB DISPLACEMENT?

I remember the good old days on the intertunnel when I’d have to warn you that the link goes to a PDF. It’s 2026, I think. I’m never sure until about February. If it is indeed 2026, I think you should have gotten over your fear of Adobe Acrobat by now. I suppose I could skip the warning about the format of the document, and offer a more timely warning for today’s internauts: It’s not only a PDF, it’s a 54-page working paper from a think tank, and it’s got a lot of words, some of them polysyllabic. It’s likely your lips will get really tired while reading it. It’s got numbers in brackets all over it, too, which I think lead to footnotes at the end. I can’t be sure, I never get that far without my eyes glazing over.

The working paper is from NBER. That’s an acronym for the National Bureau of Economic Research. They’re a think tank in Cambridge, Massachusetts. The locale makes my spider sense tingle. That zip code is ground zero for educated lunatic worldviews. But NBER doesn’t appear to be a big building full of Sovereign Citizens or people eating avocado toast and plotting to dye their hair pink or anything. It’s a loose agglomeration of academics and public policy wonks that seems at least modestly open-minded. So I figured it might be worth the time it took to read the report. Honestly, the question itself, how adaptable are American workers to AI-induced job displacement, demonstrates some clear thinking from the get-go. It’s long past time to stop arguing whether LLMs are real, or here to stay, or bankrupting only themselves or the whole nation, or useless, or whatever. LLMs are real, and they’re spectacular, sorta. Let’s move on. Whose ox is gonna get gored?

The paper doesn’t have a monomania for simple exposure to AI, which is great, because AI is exposing itself in more places than Hunter Biden. That ship has sailed. They’ve come up with an Adaptive Capacity Index, to see how well many types of workers will be able to adapt themselves to the new workplace now that LLMs rule most every roost. The analysis is interesting.

First they predict (or observe, really, at this point) the potential for tasks in an occupation to be affected by AI. Then they measure the Adaptive Capacity of that guy that always takes the last donut in the break room, and everyone like him. Adaptive Capacity is an amalgam of workers’ ability to adjust after the modification of their jobs (or outright displacement) caused by AI. It includes factors like liquid financial resources, skill transferability, geographic labor market opportunities, and  age distribution within occupations. So far, so good. The index they came up with covers 356 occupations. They claim that’s about 96% of U.S. workforce. That’s a lot. I’ll admit I jumped to conclusions earlier, and I’m not really sure if longshoremen or prostitutes might be included after all.

If you’re of the USA Today generation, they’re looking out for you. First, a bubble chart:

AI exposure vs. Adaptive CapacityIf you’re of the Facebook generation, don’t worry. They’ve got a map showing the distribution of the population that isn’t expected to survive the chatbot apocalypse. It has colors and a thermometer, rendering it still more fascinating:

Got that? If you live where the buildings are tall enough to cast shadows, you’re in danger. If you live in New Mexico, you already knew you were in danger, just by looking out the window. A failed state, that.

They’ve got lists, if you’re from the Tumblr generation. Who’s got high exposure, but high adaptability to boot? Here goes:

So much for all the news blurbs about software developers and various other computer nerds being put out of a job by chatbots. They’ve got the highest exposure, and the highest ability to adapt to that exposure.

So who’s on the other end of the spectrum, and the dookie stick? Who’s getting Skynetted first? Here’s who really needs to adapt, but won’t be able to:

I suppose it would be impolitic of me to mention that there are several job descriptions on that last list that I’d like to sign up for manned missions to the surface of the sun.

Once the report has identified the problem, they go on to mention the only solutions anyone ever mentions. The government has to step in with retraining and handouts for these benighted souls flummoxed by ChatGPT. One can’t help but notice that a lot of those job descriptions are more or less either government jobs, or private sector jobs made necessary only by government regulation. Retraining? Handouts? It’s a maladaptive snake eating itself, and turtles all the way down. Maybe they can all open daycare centers in Minneapolis. It pays well, I hear.

One is also tempted to observe that the people on the first list are preponderantly male, and the second list is loaded with the distaff set, and in many cases, just plain loaded. I was tempted, but I got over it. So I won’t mention it, or parallel parking, or any other divisive topic.

See? I’ve adapted to the internet. It has girls on it now.

Tag: AI

Find Stuff:

Archives