The logic of a termination

On the way to the coffee shop to write this newsletter today, I listened to Lulu Garcia-Navarro’s NYT interview with Scott Pelley about his firing as a “60 Minutes” correspondent by new CBS bosses Bari Weiss and Nick Bilton.1 I was struck by Pelley’s rage (he compares the changes to his wife getting murdered), which will grab onto something inside many journalists with the misfortune of having watched an institution erode from within.

As I listened to Pelley’s tale in that famous broadcast baritone of his, I could picture him at Bilton’s patronizing first address to the “60 Minutes” staff: Pelley, looking around, realizing he’s the most senior person in the room; the one with the burden to defend his colleagues’ honor, and his own. In his downfall, Pelley remains a world-class storyteller, unmatchable in talent by the people who just fired him.

Of course, it’s the fate of every true craftsperson to love a job that can’t love them back. Money and prestige are nice if you can get them. But in the balance of life, talent is something you give away.

Anyway, the rest of today’s newsletter is about the logic of using AI to replace workers, though the common theme here is that the fundamental replaceability of workers is almost always a wounding discovery to the workers.

When I flip open my tenth edition of “Managerial Economics" by Samuelson, Marks and Zagorsky, I find simple formulas that can help news executives mathematically decide when to, say, lay off journalists and replace us all with AI. Production inputs are classified as either labor (workers) and capital (stuff, like tools). A savvy manager can use the “marginal rate of technical substitution” to figure how much capital you need to replace labor and produce the same results.

[Imagine I printed the actual microeconomic formulas here. Now imagine at least a dozen readers’ eyes immediately scrolling up into their sockets, a terrifying darkness descending, a final bioelectrical signal sending their hands toward the “unsubscribe” button just before catatonia.]

Imagine the CEO is a little sweaty because it’s too warm inside the office. The CEO could bring on board a team of interns to cool him down (the CEO is a man in this anecdote) by fanning his face with their hands in the manner of an ancient imperial retinue, with some extra bowing and scraping to optimize executive morale. But unless the interns are unpaid and/or can’t be put to more productive use doing something else around the office, the science of marginal economics will probably say it’s more profit-maximizing for the CEO to just replace these labor inputs with a capital input: in this case, an air conditioning unit. This is automation, and its purpose is identical to why farmworkers are replaced by tractors, receptionists by phone trees, copy editors by spellcheck, analysts and writers by Google Gemini, so on.

There’s nothing new happening just because the capital input is called “artificial intelligence” instead of “a stapler” or “a hammer.” The process of substituting capital for labor is constant, ever-present, sometimes boring, frequently career-shattering, and, when done a certain way, insanely profitable. It’s how productivity increases and the economy gets bigger; old industries are destroyed and new ones created; goods sometimes made cheaper and tasks a little easier; and how so many more mansions are built and inequalities multiplied on behalf of whoever’s holding the equity. It’s because of this inexorable process that executives are practically mathematically obligated to be excited about AI. After all, if they don’t implement the technology, someone else will — won’t they? And aren’t the first movers likelier to win, to knock out their clumsier competitors? What if you’re one of the clumsier competitors, is the status quo even an option? Under the laws of competition, what looks like innovation is more like an obligation. This is why the disruptive impacts we blame on AI are more accurately attributable to capital, humanity’s first and most powerful autonomous technology.

It’s math the whole way down. If you add in costs, figure for changes in quality and demand elasticities, estimate changes in revenue, then you can figure whether a productivity increase means your profits increase more by increasing production using the people you already have or by maintaining output with a smaller workforce. (Which depends on the level of competition and consumer demand.) These are the formulas behind whether CEOs decide precisely to fire a bunch of people while announcing to investors that your company is “pivoting to scalable AI enterprise solutions.” However, the coming chokepoints of chip availability, data center siting and energy costs all suggest that AI usage won’t be as cheap as executives prefer, and a higher cost of “capital” means it’s less mathematically favorable to axe half your human workforce. The jobpocalypse will probably have to wait for another day.

Still, it’s the math that workers are correct to be skeptical of — where the two primary outcomes of successful innovation are that 1. you lose your job, or 2. you do more work for the same pay while the corresponding new profits accrue to somebody else. According to the Wall Street Journal, “hourly wages are up 3% since the end of 2019 while profits are up 50%,” a disparity that preceded the explosive growth of AI deployment.2 Looks like the kids are completely right to boo AI at their graduations. Well, what are we going to do about that? A society that wants more innovation needs fewer good reasons to resist it.

Once you cut through all the wizard bullshit (and we are against wizard bullshit around here, unless it has some kind of literary or aesthetic value), the principles of modern managerial science and labor theory say there’s nothing special about generative artificial intelligence. AI is a normal technology, meaning it can be adopted and analyzed using normal business principles; the abnormality is in how we talk about AI.