Context warning: suicide
“One machine can do the work of fifty ordinary men. No machine can do the work of one extraordinary man.” — Elbert Hubbard
“Oh yeah? Hold my beer.” — Open AI (probably)
The wildest SEO project I ever worked on was for a mattress company. Like Ed Norton, I can’t say which, but it was a major one. We’ll call it MattressCo.
Our job was to write articles about how great MattressCo was to get on the front page of Google for search terms like “best soft mattress,” “best mattress for tall people,” or “MattressCo reviews.”
Now, normally, you try to create that content on your site. They were doing that too, but your site can only be one result on the Google search results page. So they bought a bunch of other blogs that didn’t look like they were affiliated with MattressCo and had us publish there, too, so they could also control what was said in the other top nine spots on Google. It was an effective, expensive, and more common than you’d think strategy.
Writing the content to fill these astroturf blogs was incredibly dull. We had a canonical article on a given topic, and then we had to rewrite it nine different ways to spread it across the other sites. So we had a team of writers who we would give the canonical article and ask them to rewrite it in a different voice for a different site while still following certain criteria around length and keywords and such and such.
I’ve thought about that project a lot in the last week. They were spending nearly $50k a month on the project. We were employing a dozen or so freelance writers to work on it. Pretty soon, the cost of doing a similar strategy will be a small fraction of that, and it will be doable by one or two writers working more as editorial directors of these new AI writing tools. At least 90% of writers, or at least writing, will be replaceable by technology in the next decade. Maybe in the next few years.
So assuming you want to stay employed as long as possible before getting turned into a human battery for the Machines, it’s worth asking:
What can’t AI writers do?
II
There’s a lovely thought experiment called The Chinese Room posed by John Searle:
Imagine a room with a slit in the door. You don’t speak Chinese, and someone on the other side of the slit speaks Chinese and wants to have a conversation with you. You have a big book of rules that tell you what to say next in response to what the person on the other side is saying. You don’t understand the conversation, but you can follow the rules and keep the conversation going.
From the perspective of the person on the other side, it seems like they’re having a perfectly normal conversation with someone who speaks Chinese. But you don’t understand what they’re saying. You’re just following the rules.
Perhaps you’ve had the experience of talking to someone, and they bring up a topic with which you are intimately familiar. They’re animated and excited because they just listened to a podcast interview about it and want to seem super smart, but after you ask a few follow-up questions, you realize they don’t know what they’re talking about. They are babbling.
A surprising amount of seemingly creative work is just babbling. To a non-expert, the hacked-together SEO-optimized blog post about “best ways to improve sleep” sounds smart and educated. But to the expert, it’s obvious that it isn’t. It’s like a high school book report slapped together from some quick Googling and repackaging of whatever material you can find online. Because, well, it is.
Repackaging content into other mediums and the formulaic style of SEO writing will be easy for an AI to take over. It’s the checkers of creative work. And as it gets solved and flooded with content, people will stop being interested in it, just as they’re starting to revolt against the Twitter threadbois.
What’s chess, then? Well, there’s another style of non-fiction writing that’s very popular. It goes something like this:
Here’s a short story about a historical event or person with a helpful lesson.
Here’s the explicit laying out of that lesson.
Here’s a tie-in of the story from the beginning to wrap it up. Don’t forget to subscribe!
It feels very creative, but it will be a fairly easy formula for an AI to start churning out. You’ll be able to jump into Lex and say, “write an article about compounding interest using a lesson from history,” and it’ll be done. You might edit it for word choice and clarity to make it feel yours, but it’ll only take a few minutes. It’s not that creative.
But surely fiction is safe, right?
Well, there are surprisingly few story formats out there. I wrote about this more in “Eternal Problems,” but there are something like six emotional arcs for a story and twenty master plotlines. Do we need humans to churn out the next fantasy series about a young person with special abilities who goes off to some school or gauntlet to test themselves and eventually rise to challenge an authoritarian evil? Probably not. That’ll be quite easy for an AI to do, and it can even toss in all your Facebook friends as characters with their dialogue tuned to their writing style on social media. Then read it to you in Churchill’s voice.
It’s tempting to say that we will still need to be the ones to come up with the truly original storylines. AI can only piece together bits of information from what’s already out there. So someone has to make the stuff for it to copy, right? It can’t imagine Middle Earth or Arrakis.
But that argument should make any honest, creative person squirm in their chair. We know that none of us make anything completely original. There is no first-principle thinking or first-principle creativity. It’s creative plagiarism turtles all the way down.
This concept is captured lovingly in the Helsinki Bus Station Theory. When you start in any creative field, you must copy what exists. You don’t have the repertoire to create newer, more creative works because you haven’t mastered the fundamentals.
But as you keep working and copying what works for others, you slowly develop your voice and style. Your bus starts to get further out from all the other routes, and you start straying into new and unknown territory. Looking at where AI writing tools are now and saying they’ll never create anything original is no different from looking at a novice who just started their writing career and saying they’ll never make something original. It’s still early.
Most people never leave the central bus routes. They never stop writing in the formulaic SEO style, or they never leave the Story, Lesson, Tie-In style. They’ll keep riding those buses in circles down the drain as they get devoured by our new AI tools.
But can you ride your own creative bus hard and long enough, so your style is impossible to be copied or improved on by a computer? Probably not. AI can ride the bus much faster. It can master your style and make it better by bringing in the tones and flairs of the writers whose prose fills you with envious rage when you can’t sleep at night. And it can do it effortlessly. Instantaneously. You have no chance.
And you might say, “well, humans and AIs will just work together to make better stories!” and yeah, sure, for a bit. I had Lex write that Chinese Room description above since I didn’t feel like it. Could you tell? But it will be like Chess and Go. The human-AI teams were only a flash in the pan. Once the AI got better, it figured out how to destroy the collaborations. An AI erotica writer will put all the Amazon e-book smutrepreneurs out of business just as fast.
I desperately want to believe I’m a special snowflake and that a computer could never be so creative as to make up a word like “smutpreneur” but I know I’m lying to myself. That’s not what will separate man from machine.
IV
There is one realm of writing that perhaps AI can never fully take over. If you can focus there, you might just be able to afford real meat in 10 years instead of getting stuck eating bug goop.
No matter how painful or beautiful AI writing becomes, there will always be something hollow about knowing there isn’t a real person on the other end. That won’t matter when you want to learn about a topic, find a simple recipe, or evaluate mattresses. But while a computer could string all the words together to create Infinite Jest in equally beautiful and insane language, it cannot create the experience of knowing someone was tortured by the awful demons in its pages.
We often read to learn. But we also read for reassurance that we’re not alone. That we’re not broken, not bad people, not hopeless. We want to know other people have felt those dark, terrifying feelings we hesitate to name because it might stop anyone from ever loving us again. We want to feel what others have felt. Can a computer reassure you you’re not alone?
You’ll need to experience a life worth reading. Either through the wildness of your adventures or the depths of your introspection. You don’t have to delve into such dark topics that you end up hanging yourself. But if you’re not willing to reach that level of honesty in your work, you should probably give up now.
Anything short of the rawest, most honest human connection will be gobbled up by robots.
Steak or bug goop. Good luck.
Loved this. I am definitely curious how much the average reader will be able to tell or care whether the author has bled for the story they are telling. Guess we'll see in coming years.
Bet the AI won't miss the subheaders. And that too makes this piece uniquely human. 😂
Even humans don't get true vulnerability right. It's gonna take the AI even longer. And while it'll feed the content machine, the good stuff will inevitably rise, because we share the things that make us less alone.