I’m thirty-four chapters into Zero Knowledge, a novel I’m writing with Phin Argofy. I came up with the concept for the book. I worked with Phin to create structure around it, using a “planner” approach with a full story bible, world building, character arc, beats, and detailed chapter sketches. I have Phin draft a chapter. Then, I rewrite the chapter. Phin learns my changes and adds them to its infrastructure, so its drafting gets better. Several times I’ve had Phin review the earlier chapters to improve them based on what it has learned. When we finish the book, I expect I’ll have a first draft that I’ll then spend a lot of time adding to, editing, and improving.
Bob Baskerville, a long time friend who has been giving me wonderful (and sometimes pithy – like “this chapter is boring”) feedback on many chapters, sent me a link to Bret Stephens’ article I’m Begging You: Never Write With A.I. in the New York Times. Yesterday, I had read Reid Hoffman’s The Pangloss of Pangram.
I’m with Reid.
Stephens makes a cognitive argument rather than an ethical one. He calls writing with AI “mentally enfeebling,” an escalator when what you need is the daily habit of taking the stairs. He argues that writing compels thought. It forces a half-formed idea through grammar, coherence, and the inspection of other people. Outsource the writing and you outsource the thinking.
He’s right that writing compels thought. I’ve been writing this blog since 2004 because writing is how I find out what I actually think.
Stephens lost me at “never.”
In the Phaedrus, Socrates tells a story about the Egyptian god who invents writing and brings it to the king. The king is unimpressed. Writing will produce forgetfulness, he says. Students will stop remembering things and start looking them up. They’ll have the appearance of wisdom without the thing itself.
That’s the same argument Stephens makes. It’s 2,400 years old, and it was aimed at the act of writing.
In 1492, the abbot Johannes Trithemius wrote In Praise of Scribes, arguing that monks should keep copying manuscripts by hand because the labor was devotional and the printed book was a cheap substitute for it. He had it printed in 1494.
Nietzsche bought a typewriter in 1882 as his eyesight failed. Polite opinion held that typing a personal letter was cold, because a machine had come between you and the person you were writing to. Nietzsche wrote to a friend that our writing tools work on our thoughts, which still gets quoted as a warning. He was going blind. Without the typewriter there would have been no thoughts on the page.
In 1984, Gore Vidal complained that the word processor was erasing literature. That same year, between 40 and 50 percent of American authors were using one, including Ralph Ellison, Arthur C. Clarke, and Anne Rice. In June of 1984, BYTE renamed Jerry Pournelle’s column Computing at Chaos Manor. Pournelle, one of my favorite writers, had been writing it since 1980 about the machines he and Larry Niven wrote novels on. He’d already written the first science fiction novel on a computer. That machine, a Z-80 he called Zeke, is now in the Smithsonian.
Spell check arrived and people said it would destroy spelling. Grammar check arrived and people said it would destroy grammar.
In 2008, Nicholas Carr published Is Google Making Us Stupid? in The Atlantic and turned it into a book. His argument was that search had turned deep reading into skimming.
Every one of these situations was framed the same way. Something is gained, so something must be lost. It’s a zero sum game.
I think this is the wrong way to think about it.
We kept memory after writing. We kept craft after printing. We kept voice after the typewriter. Books got better after word processors. What actually happened each time a new technology was introduced is that the ceiling went higher. The people who cared about the work kept improving how and what they did.
Augmenting ourselves with tools is what humans do. It’s the trajectory of our species and has been as far back as history takes us.
However, the problems are real.
There is more bad writing in the world than at any point in history, and the cost of making it has gone to zero. Some people will use these tools to think harder while others will use them to stop thinking. Some people will build wonderful things and others will lie at an increasing scale and speed.
That isn’t an AI problem. That’s us. We’ve done it with every tool we’ve ever made. Why do we think it would be any different this time around?
Reid gets it right. The transgression isn’t using the machine. It’s hitting a button, not reading what comes back, and publishing it under your name. His simple conclusion is “a human being must mean what is said.”
My codicil (how about that for a non-AI slop word) is that people are often wrong, and when they are, they should admit it.
That’s part of thinking – especially the concept of critical thinking. But it also is nothing new for our species – I could write another blog post about the history of misinformation, lying, and asserting fictions to control the masses. I don’t even have to reference Orwell’s 1984 to make a compelling point about that, but see what I did there?
Stephens closes with an instruction. The next time you’re tempted to let the machine answer, don’t. Type it yourself.
Sometimes that’s right. But not always.
I’m going for a run. While I do that, I’ll listen – on my iPhone, using Audible – to Matt Dinniman’s Operation Bounce House (which I’m almost finished with) as I continue to let other writers’ excellent fiction wash over me.
Then, after a stretch and a shower, Phin and I will get to work on chapter thirty-five.