The Death of Quality

When AI can generate the guise of quality, how can we know real quality work when we see it?

One of my favorite novels has been on my mind recently. Perhaps you’ve read it, too. In Zen and the Art of Motorcycle Maintenance, the protagonist goes insane. We never know his name; he’s the narrator of the story, and as the reader, you get a first-hand account of his descent from rationality to disarray.

What causes his insanity?

In a word, quality. The world uses “quality” as a ready-made lens for judging pretty much everything. “The artwork was high-quality; he gave it a low-quality effort; we want a quality meal.” People make such everyday statements with incredible certainty, he observes. Yet no one can really tell him what quality is. What’s the quality of quality? Grappling with this shifts his brain into overdrive. The harder he tries to understand quality, the more uncertain he becomes about his ability to understand anything. Insanity follows.

Today I wonder if we are doing the opposite.

When AI can generate anything that carries the guise of quality, how can we tell what counts as true “quality” work anymore?

In the past, one proxy we had for evaluating quality was length. Michael Porter’s Competitive Strategy is about 110,000 words. The Brothers Karamazov is more than twice that. Even if you never read those works, you could at least make a safe assumption that a lot of effort and thought went into producing them. There was only one way to generate such an epic, and short of plagarism, that was to sit down and write them out. But today length is no longer correlated with effort. I could ask Claude to generate a 4-volume tome this morning, and it would be done by this afternoon.

What about reasoning as a proxy for quality? Writers like Aristotle, Kant, and St. Augustine had superpowers here. They could hold complex ideas in their head, string together tight arguments, and articulate their thinking in writing, all without editing software (or probably, an editor). Yet AI excels at reasoning. You and I can hand AI a poorly structured argument and ask it to weave it back together with sound logic. For a reader unfamiliar with the subject, the material will seem high quality.

Let’s try a third angle of attack: the style and quality of the prose itself. In the past, one could go to writing school and learn how to be a “good” writer. And not long ago, I thought this was still an advantage; AI had a certain style that was hard to define, but easy to spot. But even this is starting to erode. AI can adopt the style of nearly any well-known writer, which you can then easily pass off as your own “voice.”

This should be the part of this essay where I lay out a new approach for evaluating quality. But I don’t have one. I sat down to write this morning, in a short window of time I had before leaving on a trip to Europe, hoping an answer would emerge.

It didn’t.

High-quality work still exists. But it’s harder to spot.

All I know is that if we forfeit our ability to discern the difference between a contribution that’s truly high quality and something that merely has the guise of it, we’ll start to develop our own version of insanity.

---

John Rougeux is the founder of Flag & Frontier, a strategy consultancy that helps executive teams align around their strategy and narrative when the future is up for grabs.