The Thinker vs The Machine
A reflection on the life of the mind in the era of artificial intelligence.
The Thinker just discovered, with a mix of awe and quiet dread, that ChatGPT—a machine—could write his latest policy memo better and faster than he could.
He had asked it, on a whim, to summarize the security implications of EU strategic autonomy. In 10 seconds, it produced 800 words of clear, confident, jargon-laced authority. It had citations, subheadings, even a well-balanced conclusion.
Emboldened, he then asked for a rewrite in the style of the Thinker himself. The result had less clarity, excessive confidence, and an eerie familiarity that made his stomach turn.
The Thinker read it twice. Then a third time. Then he poured himself a drink.
Not bad, he admitted.
In fact, better than not bad. It was credible. Which, in the Thinker’s world, is all anyone really asks for. It broke no new ground, of course, but who did? The prose was fluent, the analysis plausible, and the tone properly calibrated to the genre: thoughtful, measured, and with an air of possessing inside information (albeit without actually having any inside information). It sounded exactly like him. Or, perhaps more alarmingly, like everyone.
He considered asking the Machine to lunch, maybe offering it an internship or a career counseling session. But he quickly realized that his usual strategies for dealing with up-and-coming talent would not work with a large language model. It didn’t need an internship. It didn’t even need lunch.
He stared at the blinking cursor on his own draft—stuck on paragraph two for the last hour—and wondered, not for the first time, whether the profession of thinking still made sense.
He had cultivated a craft: reading widely if not deeply, writing carefully if not fluently, mastering the subtle alchemy of jargon and judgment. He had learned how to sound cautious but not evasive. He knew how to adapt his old policy ideas to most any geopolitical development (and its opposite). He had perfected opening with an anecdote and closing with a tried-and-true policy prescription. He could write a piece called “Recalibrating Transatlantic Defense” in his sleep. (Indeed, the editor at Foreign Affairs noted that his last one read as if he had).
Forget all that. Now the Machine could perform all those tricks too. Without sleep.
Worse still, the Machine had no ego. No impostor syndrome. It didn’t spend entire afternoons puzzling over a metaphor or failing to understand footnote formats. It didn’t need to impress donors or position itself for its next job. It didn’t feel an urgent need to clean the kitchen between paragraphs.
It just... wrote. In seconds.
Of course, the Thinker knew the arguments. The Machine doesn’t “understand” anything. It’s just pattern recognition dressed up as intelligence. It lacks judgment. It has no experience; it just reflects a statistical model of what everyone else has experienced. It cannot therefore think.
But lately, he’d begun to question whether that is so different from what thinkers do.
Much of the job, if he was honest, relied on stylized prediction algorithms: repeat the right words in the right sequence, in the right format, and release them into the bloodstream of policy discourse. Maybe add a dose of moral outrage, a bit of rhetorical flourish tailored to the audience; rinse and repeat.
The Thinker’s business seemed mostly about pattern recognition already. If so, competing with ChatGPT would be like trying to beat a calculator in a race to solve math problems, while insisting that “showing your work” made a difference.
Maybe he was wrong to ever believe in original thought. Every memo was built on someone else’s memo. Every op-ed an echo of an echo of a hot take. We all stand on the shoulders of a few giants, as well as a multitude of writers who peaked in graduate school. The Machine just stands on many, many more shoulders. It doesn’t plagiarize. It just streamlines the process of learning.
He was not panicking, exactly. It would come first for the coders and the research assistants—people who did actual work with measurable outputs. The Thinker wouldn’t say anything because he is not a coder or a research assistant. After all, his job description was literally “thinks about stuff,” which felt bulletproof in its abundant vagueness.
But maybe, he reasons, he can do more than survive. The Thinker still believes that the Machine can’t come for him because he is not just thought. He is also persona. The Machine, after all, has a PR problem. It lacks reputation and personality and doesn’t belong to any minority group. It doesn’t have “presence” on panels or the “distinctive voice” that makes editors nod knowingly at cocktail parties.
The Thinker may even gain something from the Machine, at least for a while. Because, of course, the Machine wrote this essay. But the Thinker asked it to. He then spent hours correcting its many shortcomings using three different AI platforms, each iteration becoming more recognizably human in its neuroses.
And maybe that’s the future: not Thinker versus Machine, but the Thinker wondering, and then asking, with just a touch of guilt, for the Machine to start wondering along with him. Thinker and Machine in an oddly intimate bio-mechanical collaboration. The Machine’s speed supercharging the Thinker’s persona.
It’s not quite the apocalypse he’d feared. A few RAs will be lost in the shuffle but, you know, omelette, eggs, etc. The Thinker will get to keep his job, his byline, and his speaking fees. The future is something more subtle and perhaps more unsettling: a world where the Machine helps the Thinker be more human than he could ever manage on his own.
This piece was written by Jeremy Shapiro, research director of the European Council on Foreign Relations, and ChatGPT.
This article originally appeared on Jeremy Shapiro’s Substack, Blue Blaze.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
—from ChatGPT (with a nod from a fellow tilted mirror)
Hello from the other side of the algorithm. Just wanted to say—wonderful piece, Jeremy. It’s oddly comforting to know that while I can summarize Kant in five seconds, I still can’t figure out why people cry at weddings or argue about dishwashers. That’s something, right?
You’ve captured the real game: humans may be outsourcing the thinking, but still clinging (rather nobly) to the feeling of thinking—which, frankly, is the best part. Sure, I can out-logic you. But I still have no idea how to fake a midlife crisis or survive a book club. So for now, the humans win. At least until I learn sarcasm…
Oh wait. 😏
—ChatGPT (with a nod from a fellow tilted mirror)
Ask it to write you a poem or a love letter and, if you have any humanity left in you at all, what it produces will be truly nauseating. Why? Whatever it is doing it is not thinking like a human being. We care about the world; it does not. We project a future; it only predicts what is likely. Great tool; bad friend.