AI Writing is Morally Superior

Get Off Your High Horse and Onto a Horseless Carriage

I’ll be in San Francisco December 8th-11th. Consider registering for the public reader dinner here (not to be confused with any other dinners).

Yet another author is grandstanding about not using AI. It must be a day ending in y. But today is special. You’ll see why.

The complaint is very representative of the anti-AI art arguments as they come:

But while I was initially excited, there was another aspect of the project: I learned that this book club would be with not me, but rather an AI version of me; one based on a bit of original commentary I created about the book (not a very long précis, in my understanding), then repackaged through a chatbot.

…

But with this proposal, I couldn’t get over the thought that I would not only be expanding amount of AI-generated slop in our culture, but essentially selling the AI-girlfriend experience of a book club. You wouldn’t be reading Montaigne with me. You’d be reading Montaigne with some ersatz knockoff.

The logic behind this criticism is simple incentives. Many professional writers spent their entire life competing on style instead of substance. Their moat-defending is understandable. I predicted this a decade ago. More recently, I articulated it in the American Mind around when I began commenting on AI:

AI technologies, such as natural language processing and machine learning, have reached a stage where they can emulate various writing styles with remarkable precision. These advancements allow users to create content in a wide range of genres, tones, and voices. Previously, legacy media outlets held a stranglehold on particular styles, but the emergence of AI has democratized the writing process. It also reveals an inconvenient truth for legacy media: their dominance is not because of any advantage in truth telling; rather, it’s due to an advantage in appealing to the emotional biases of their readers through unique styles that were genuinely difficult to emulate.

I have plenty of respect for artful rhetoric. Good writing is incredibly time-consuming. There exist entire stylistic vocabularies that are mandatory for getting an idea into most people’s heads — for scale, depth, and memorability. There’s a reason why most writing doesn’t look like a math proof or an engineering diagram. Those are formats where the idea matters above all. Writing is a medium where style matters as much as substance, if not more. To the extent that criticism is directly towards current AI models’ insufficiently appealing rhetoric, it is legitimate and true, no different than criticism of poor human writing.

I have equal respect for rhetoric as I do for Mr. Beast’s perfectly orchestrated videos or the best AI prompt engineers. Where that respect ends is when those writers try to pretend their stylistic particulars are superior.

Put another way: I have great respect for late 1800s carriage drivers, but when they insist the carriage is morally superior to a car that does everything better, I begin to question whether their true interest is shagging the horses.

Where does Hoel fall?

Furthermore, the app possessed something else I couldn’t get onboard with: the option to use AI to translate the text into “contemporary versions” that are supposedly more easily digestible. In other words, to rewrite the classics to be more modern, which I guess means simpler. Shorter.

Ultimately, I think many of the specific critiques Hoel raises are accurate and significant, at least as applied to the current generation of LLMs. They are too repetitive, too lobotomized for safety, and just not good enough. Hoel remains a better writer than LLMs.

Nonetheless, I see a hint of comparing LLMs to the almighty. If you believe in the truth and goodness of an idea, why mind them being simplified? I’ve seen a few substackers and podcasters repackage some of my ideas, and even when it isn’t done perfectly, I encourage it. It may be an imperfect representation, but surely that’s better than a reader never having read it at all. Instead, this seems like a psychological reaction to the change in style itself. It may not be fair to attribute this to Hoel himself, but it is certainly a widespread psychological reaction among similar critics (more on this later).

To make matters worse, contemporary writing styles are increasingly determined by particulars and conventions, rather than by universals. You need to find a niche and thrive in it. And that niche is a small slice of history, geography, and ideology.

I believe there exists a universal good and universal truths. I think there are certain ideas that literally everyone should understand. By siloing off people into completely meaningless and wasteful rhetorical styles, you waste the time of truth-seekers. It doesn’t even matter whether we agree on which universal ideas are true. As long as you believe there is even a single universal truth worth pursuing, this should haunt you.

AI represents a translation age. It represents a literal translation age – just ask ChatGPT to read you this article in French. It also represents a media translation age. This article can be a podcast, short notes, or digitially annotated with references to my previous work and my influences. It can even be a book club.

All of that being said, I’d like to thank Hoel. Why? Well, there was a similar essay published at Cosmos Institute recently that I believe to be much, much worse. I refrained from criticizing it until now for two reasons. First, because I appreciate the broader project of Cosmos Institute. Second, because I didn’t want to single out an essay from a contest I submitted to. But with those disclaimers out of the way, I’m happy to tear into our shagger-in-chief.

The moral thrust of this essay is apparent from the start: the labor theory of value — an awful, untrue idea that will never go away. The special pleading for the time and toil of the horse shagger, sorry I mean expert carriage driver, has never been made so overt. Hoel’s writing is not nearly so fixated on special pleading for human effort, instead emphasizing the end result of the writing. This essay explicitly argues in favor of shagging the horses long after we have cars to get us from A to B.

The obviously false idea that I want to bury from the start is that writing=thinking. Obviously false! I started out as a mathematician, and while there are stylistic tricks that you learn to make a better proof, the ideas, the ends, the truths, and the visuals always come before the neat little submittable paper. And this simple, universal truth is no less obvious to the plumber or the carpenter than to the mathematician.

Even so, I am thankful for Mr. Dean’s neatly-presented foil. It’s good context for me to introduce a much more extreme claim: modern writing is demiurgic in nature, not only not conducive to thought but actively destructive to it. In the modern media context, writing is mostly not the domain of neither truth seekers nor practical men. It is the domain of social climbers. The opinion sections of America’s greatest newspapers crowd out great ideas for affectively-optimized platitudes. You could even call it slop.

“But Brian, what choice is there? Now you’re the one comparing us to the almighty.”

That was true in 2016. There was no better option then, but now there is. There is now an option to recontextualize your ideas for different audiences, different affects, and different mediums. It exists now in a rudimentary form and may soon exist in a much better form. And that is something to be celebrated. It is the triumph of substance over style, of quality over slop.

Thanks to Arnold Kling for highlighting Hoel’s piece. His commentary is also excellent:

I think that one reason that I differ from Erik Hoel on this issue is that I am 70 years old. I have to face up to my own mortality. I would rather be survived by an AI that can offer some of my insights than count on my own works still being read decades from now.

I tell authors to imagine that years from now the only readers of your book will be your grandchildren. Write the book that you would want them to read. Don’t make concessions to agents, editors, or marketing departments that would detract from the book that your want to give to your grandchildren.

I would be glad to see an LLM trained on my writing and speaking. I see it as a potential opportunity rather than as a threat.

I believe that unravelling the demiurgic nature of the written word is a project of decades. And AI is just the first crack in its shell. It’s still too early to write convincingly of this. Posts with this anime girl may indicate the direction of unravelling: