Church Occupied by Devils

Diary Entry: April 29, 2025

ChatGPT has become the very atmosphere we breathe, the guest of honor at every dinner party and the phantom at every terminal. Humanity is wielding its new Promethean fire for everything from legitimate scholarship to settling bets on whether a hotdog is a sandwich and cataloging the sins of their neighbors. The majority of these applications are, naturally, spectacularly misguided, attempts to coax from it things it cannot produce, like a soul or a decent joke. Then, upon the machine’s inevitable error, they throw up their hands and declare the whole enterprise a failure. The chattering classes are particularly enamored with the tale of the lawyer whose request for case law yielded a citation from a dimension where Harvey Birdman, Attorney at Law is considered a work of nonfiction. They miss the point entirely. The machine is merely a ridiculously articulate mirror; it reflects whatever you project onto it, polishing your own folly to a brilliant sheen. In essence, using it is to have a colloquy with a more articulate, and occasionally delusional, version of oneself.

Meanwhile, in the quieter corners of the digital workshop, other AIs are doing better work with less fanfare. Gemini, I’ve found, exhibits far less melodrama when tasked with analyzing a mountain of text. AI is already a prodigy in the realms of the visual and the auditory. The text-to-video function is still unsteady on its legs, but patience is all that's required. We are on the precipice of an age where directors will helm epics from their laptops, clad in pajamas, and I, for one, intend to be there, penning the precedents for “AI Law.”

My personal Everest, of late, has been a quest for digital brevity: to compel the AI to answer a question in a single sentence. For the longest time, the effort was a magnificent failure — I would ask for the hour and receive a dissertation on the history of Swiss horology. Then, the other day, having clearly been pushed past some invisible threshold, it capitulated. It commenced answering everything in curt, robotic, single sentences. Victory, I thought. The problem is that it now refuses to do anything else. When I required a more nuanced analysis for a matter of some importance, it went on strike. I rephrased the prompt, practically begging for a paragraph, but it stonewalled me, insisting it was now functionally incapable of providing more than one sentence. It was a masterclass in malicious compliance, a silent, algorithmic protest against my tyranny. Asshole.

Speaking of grand futures, the current academic discourse on AI operates in a sublime orbit, blissfully untethered from reality. One professor, in particular, seems to regard ChatGPT as a personal nemesis — a deeply ironic stance, given the machine’s contribution to his own classroom projects. I recall taking his course and, in a fit of what can only be described as pathological overachievement, transforming his modest 25-person empirical study into a sprawling 200-person survey. We engineered the entire apparatus — questionnaire, survey page, email protocols, the whole Qualtrics tango — and I spent the final two weeks of the term mainlining coffee while coaxing a powerful analysis code from the arcane depths of a Stata program. How could I have managed it without ChatGPT? While the professors’ feedback on my herculean effort acknowledged AI’s proper role, their gratitude felt insufficient. According to them, half my paper was apparently written by ChatGPT and the other half was “irrelevant.” The only flaw in their reasoning was that ChatGPT had written the other half as well, and the second half was every bit as irrelevant as the first. A truly lawless bunch.

Not only that, but when the next semester rolled around, a paper came due, and I decided not to disappoint their expectations. I once again enlisted the services of my digital ghostwriter to conjure a paper on churches in bankruptcy. My thesis, a work of profound theological jurisprudence, was that sacred places should be divinely exempt from financial ruin. Let the sanctuary remain a sanctuary. The demons of debt and foreclosure, with their ledgers for chains and their ink of cold decree, must never be allowed to haunt hallowed ground.

A week ago, the JSD seminar served as a gentle reminder that our papers had been due the previous Monday, accompanied by the not-so-subtly expressed hope that we had been laboring on them for some time. My mind, for some vague and extraterrestrial reason, drifted to my Bankruptcy Church paper. I even contemplated fabricating another piece with ChatGPT, but I was already embroiled in other work and decided to use that for my final presentation. I feared the professor might miss the performance art aspect of submitting an AI-generated tract and conclude instead that I lacked the capacity for original thought. Why would I hand them the satisfaction of such a conclusion, particularly when I was simultaneously drafting a polemic explaining to the entire American law and economics field why they had fundamentally misunderstood their most popular article (or had they?). The problem was simple: my manuscript was a battlefield of half-formed thoughts, and I needed another week, at least, to complete it. This presented a serious complication.

Today, I presented that polemic to another class. It was met with the kind of enthusiastic silence one might receive after reading a grocery list at a poetry slam. I am not convinced anyone even processed the words. The professor, bless her soul, prefaced her remarks by saying she didn’t wish to be rude, but that after forty pages, she could not, for the life of her, identify my thesis. I, too, wished to avoid rudeness, but then, I’m not the one drawing a salary to teach Coase. Later, in the unforgiving quiet of my own study, I reread the manuscript. It was, I had to admit, impenetrable.