Complex Numbers

Machine Learning

In which The New Yorker discovers science fiction.

What a Sixty-Five-Year-Old Book Teaches Us About A.I.  
Rereading an oddly resonant—and prescient—consideration of how computation affects learning.

Pick a random selection of science fiction from the past—I don't mean space opera, which is fine in its own right, but actual attempts to write about the intersection of society and advancing technology—and you'll find that some of them are, in hindsight, "oddly... prescient," while the vast majority are like "well, that never happened."

Neural networks have become shockingly good at generating natural-sounding text, on almost any subject. If I were a student, I’d be thrilled—let a chatbot write that five-page paper on Hamlet’s indecision!—but if I were a teacher I’d have mixed feelings.

If I were a teacher, I'd be like, "hey [chatbot], write a syllabus for a sophomore English class aimed at mid-range students."

On the one hand, the quality of student essays is about to go through the roof.

Yeah, except, well, see Jayngle Bells's rant about that sort of thing, here: "AI Detectors are Horseshit" 

Luckily for us, thoughtful people long ago anticipated the rise of artificial intelligence and wrestled with some of the thornier issues.

Substitute pretty much any technological advancement for "artificial intelligence," and "thoughtful people long ago" anticipated it and thought about some of the possible consequences. That's basically the definition of science fiction.

But I don't expect anyone at TNY to understand that.

Their book—the third in what was eventually a fifteen-part series—is “Danny Dunn and the Homework Machine.”

Just including this bit so we all can see what they're talking about.

There follows several paragraphs of meanderings about the setup, plot, and characterization, before getting to the stuff that is really relevant to the topic. Typical TNY. Still, it's worth glancing over so you know what he's talking about later.

I hesitate to give away too much of the plot, but (spoiler alert!) two mean boys in their class, one of whom is jealous of Irene’s interest in Danny, watch them through a window and tattle to Miss Arnold.

Oh, no, wouldn't want to spoil a plotline from 65 years ago. (Spoiler alert: the author pretty much gives away too much of the plot.)

Incidentally, Rosebud was the sled.

He points out that Danny, in order to program Minny to do his homework, had to do the equivalent of even more homework, much of it quite advanced. (“Gosh, it—it somehow doesn’t seem fair,” Danny says.)

I must have read these books as a kid. I have a vague memory of them, anyway. But it may explain why, whenever I have to do anything more than once, I search for a way to automate it, and often spend more time crafting an Excel spreadsheet or some code than I would have spent on the projects.

“Danny Dunn and the Homework Machine” is ostensibly about computers, but it also makes an argument about homework.

And yet, there is still homework. I'm of the considered opinion that grade-school level homework has the primary purpose of making the kids leave their parents alone for a few precious minutes in the evenings.

The article spends an inordinate (to me) amount of time arguing about homework in general and not on the ethical implications of AI, but the main point remains: people have been discussing this sort of thing since long before it was technologically feasible.

Just like with the rest of science fiction.