Reading Asimov in the days of LLMs

Krystal Guo · June 20, 2026

While the articles in my headspace and in my group chat were entitled “Dispatches from the Possibly Last Days of Human Relevance”¹ and “AI Is a Meteor. Don’t Be a Dinosaur”², I read Asimov’s robot short stories and found this passage:

Your robot takes over the galleys. Soon it, or other robots, would take over the original checking of passages, perhaps even the deduction of conclusions. What would that leave the scholar? One thing only — the barren decisions concerning what orders to give the robot next! I want to save the future generations of the world of scholarship from such a final hell.

While this quote speaks directly to our current discourse about the use of generative AI, this story appeared in 1957. Everything really has all been said before; Emerson wrote “Old and new make the warp and woof of every moment. […] The originals are not original.”; before him, Shakespeare wrote “Show me your image in some antique book”; and even before, Chaucer wrote “And out of olde bokes, in good feyth, / Cometh al this newe science that men lere”.

“Galley Slave” is a courtroom drama. A robot named Easy (or EZ-27) has been leased to a university to proofread galleys, the drudge-work of scholarship. A sociology professor, Simon Ninheimer, sues the robot’s manufacturer, claiming Easy has destroyed his reputation by altering the text of his book to misrepresent the work of his colleagues. The First Law of Robotics forbids Easy from harming a human, and destroying a scholar’s reputation is harm, so how could this have happened? Spoiler alert: the answer, which the robopsychologist Susan Calvin — a recurring character in these robot stories — deduces, is that Ninheimer instructed Easy to do it, to bring the trial about. Ninheimer speaks the quoted passage to Calvin, during their conversation at the close.

Aaronson’s blog post is in response to an unknown OpenAI model solving the unit distance problem.³ Over the past month, I’ve seen two prevalent responses to this news in the mathematical community: morose despondency and shaky denial, both well represented in the comment section. The common cause is that they share Ninheimer’s fear of being reduced to giving prompts. Even worse, the updated fear is that mathematicians will be reduced to checkers and bookkeepers, verifying the logic of AI-written proofs. We might end up being more like Easy than Susan Calvin.

Earlier in the same conversation, Ninheimer compares his predicament to that of artisans, displaced by the Industrial Revolution.

For two hundred and fifty years, the machine has been replacing Man and destroying the handcraftsman. Pottery is spewed out of molds and presses. Works of art have been replaced by identical gimcracks stamped out on a die. Call it progress, if you wish! The artist is restricted to abstractions, confined to the world of ideas. He must design something in mind — and then the machine does the rest.

Barak also writes, in echo, “In particular, the Industrial Revolution cost many skilled artisans their jobs, including handloom weavers, knitters, and shoemakers.” Both Barak and Ninheimer compare this to robots coming for jobs in their own disciplines, but reach opposite conclusions. Barak encourages students — for this is a commencement-season op-ed — to see the AI as a tool, instead of as a competitor.

A common refrain of people trying AI is “yeah, I tried it and it’s awful, it gave this-and-that bad output.” But they didn’t try giving it more context. They didn’t try giving it clearer instructions or asking it to break the task down into more tractable parts first. They didn’t try asking it to analyze its own logs, in order to engineer better prompts. Ninheimer has one interaction with the robot and then reads one book on the topic. He hasn’t grasped the many implications of the First Law — Asimov wrote several volumes of stories about said implications — and that knowledge gap is exactly how Calvin catches him.

TLDR; the panic about AI isn’t new; it’s been done eloquently in 1957, for example. Generative AI is new; our energies would be more effectively spent on learning how to use LLMs, as opposed to voicing panic, in which realm we would not outperform Asimov.