Rage against the machine

One of the books that I like to pretend I’ve read is The Star Diaries, a collection of short stories by Polish science fiction author Stanislaw Lem featuring Ijon Tichy, cosmonaut extraordinaire.

In one such adventure, Ijon is dispatched to a remote planet to infiltrate a society of robots who have overthrown their human masters, after every previous spy Earth sent disappeared without a trace.

Long story short, he’s promptly captured upon arrival, and brought before the robots’ leader. The leader makes a deal with him: in exchange for sparing his life, Ijon will disguise himself as a robot, and inform on any new human spies that Earth sends.

Ijon readily agrees, but later wonders how many of his predecessors took the same deal. That’s when he discovers that all the original robots have long rusted, and every single robot in the colony is actually a human, dressing and acting the way they believe the other robots expect them to.

I think there might be a lesson in there for us.

Robotic prose

As you know, I occasionally prostitute myself at that altar of Mammon known as sponsored posts. To help myself sleep better at night, I rationalise that this is really no different from the Old Masters, who would often be commissioned to paint flattering portraits of otherwise portly patrons. Or Candy, down at the local gentleman’s club, with a garter full of singles.

Anyway, a request popped up in my inbox a few months ago. On the surface, it looked like a pretty standard engagement: tell the world that our cards are the absolute best, and will solve every single problem people have up to and including male pattern hair loss.

However — the agency quickly caveated — what was different about this campaign was that the client wanted to optimise the article for GEO.

GEO? I thought. What’s GEO? Is it the kind of thing you treat with penicillin? 

Apparently not. GEO, or Generative Engine Optimisation, is a way of writing content such that it’s cited as a source in AI-generated answers like ChatGPT and Gemini. As one article helpfully put it: “remember, GEO and SEO are like the crab and the lobster — both crustaceans, yet different.” (the same article also provided another great tip: “it is important to sound human”, which sounds like something Mark Zuckerberg’s handlers tell him on a daily basis).

“So let me get this straight,” I said. “You want me to write for robots?”

“No, absolutely not!” replied the agency. “We love your tone of voice and all that you do. But by the way, here’s a 1,527-word guide on the best practices for writing things for GEO.”

This guide was a thing to behold indeed. “Avoid long, complex sentences,” it said. “Don’t use rare or technical vocabulary.” “Don’t include long introductions or storytelling.” “Make content easy to skim.” It even included a few paragraphs which I’m pretty sure are not English. “Improve scannabiility (sic) for users while strengthening crawlability and semantic clarity to increase likelihood of accurate surfacing of product info in AI-generated responses.”

Well, anyone who knows me knows that I’m all about strengthening crawlability and semantic clarity to increase the likelihood of accurate surfacing, so I decided to get cracking. Besides, I’d like to think that I have a pretty good grasp on what machines like: regular oil changes, power sources somewhat more efficient than the human body (Matrix rules don’t work), and world domination.

An hour later, I gave up. Try as I might, I just couldn’t construct the kind of prose that would be acceptable to our robot overlords (who I, for one, welcome). Every sentence I wrote felt too sentimental, too squishy, too inefficient — you stupid meatbag! — that it’d surely be an affront to the neural network. I also had some lingering doubts about whether the four paragraphs of hot-and-heavy LOTR fan fic written in iambic pentameter that somehow landed on my scratchpad would pass muster.

In any case, the project was later cancelled, a good reminder to freelancers to always get paid upfront. But that got me thinking: perhaps I’m indeed in a sunset industry, like the spinning jenny on the eve of the spinning mule.

I mean, you don’t have to be a futurist to see the impact that AI is having on publishing. Over the past few years, AI overviews have taken a significant toll on traffic, as people increasingly rely on Gemini’s answers instead of visiting the sources that it cites. 

For the record, I don’t blame them. I find myself conditioned to do it more and more too. Still, it hits me particularly hard, as blog posts are much more susceptible to this kind of disruption than short-form videos from aesthetically-pleasing influencers showing various degrees of cleavage (though just wait till they meet the AI-generated aesthetically pleasing influencers showing various degrees of cleavage!).

It is also somewhat aggravating that when AI and human-sourced information conflict, the latter doesn’t get the benefit of the doubt —especially when the human in question is yours truly. 

AI also means that there’s been a surge in websites covering miles, points and credit cards which were clearly written by AI, for AI. “Slop” doesn’t begin to describe the level of effort, when you can’t even bother to get the card art right.

Now, it’s one thing for a no-name “publisher” to be spewing this sort of rubbish. 

But then you have supposedly reputable companies like Trip.com going down the AI rabbit hole too, churning out what could quite possibly be the most useless review of the HSBC TravelOne Card. You know, that well-known Visa card that comes in a stunning two-tone metallic blue finish?

For me, AI-generated content is the uncanny valley of writing. Sure, it’s grammatically sound, and probably observes the Oxford comma better than I do, but when you read it, you get the feeling that this is what a machine thinks a real person sounds like. 

It’s filled with vague and sometimes hallucinated information. Most of the sentences are shorter than 10 words. It’s packed with motherhood statements and nuggets of wisdom such as “avoid low value redemptions when possible” or “your total travel bill could help you earn miles faster compared to using a standard cashback card”.

It’d almost be funny, if the implications weren’t so terrifying. This “review” is featured on the first page of Google’s search results — ranked #4. 

Thankfully, human-written reviews and guides retain the ranking edge — for now — but that’s the kind of slop that threatens to drown it out, because ultimately, human publishers will never be able to outproduce the volume of content that AI can create.

Look. I’m no Luddite (which, come to think of it, is exactly what a Luddite would say). I’m happy to use ChatGPT for my work, to automate repetitive tasks (e.g. adding country flag emojis to lists), to quickly bring me up to speed on a topic, or even as a glorified spell-checker (it spots logical inconsistencies or calculation errors that Microsoft Word can’t).

But, in my opinion, AI should never be doing the writing. 

I think The Gospel Coalition put it rather well:

It’s an offense to the word “writing” to apply it to AI-arranged words and sentences. What AI produces is an impressively polished, usually helpful, sometimes cheeky string of words. But it’s just instantaneous pattern recognition and data organization. It’s soulless slop that can mimic and masquerade as brilliant writing. It’s aggressively efficient, speedily rendered, sycophantic information.

But it’s not writing.

Even when it’s created at the behest of a human’s creative prompt and detailed parameters, and even when it’s just assisting a human who has real writing chops, AI-fashioned content doesn’t deserve to be called writing.

Writing is an art form and a labor of love (love is something AI cannot do).

Writing is a product of deeply human, necessarily inefficient processing, meandering, wrestling, noodling, cogitating, and often suffering (suffering is something AI cannot do).

For Christians, writing is a holy calling and an act of worship (worship is something AI cannot do).

I love reading. I’m a book nerd and a sucker for great writing. But I have zero interest in AI-made “writing.” And no matter how good AI “writing” gets in the future, I will never have interest in it—for the same reason that AI-made music (however catchy or beautiful on the surface) gives me the ick. Those who claim we will “learn to love AI writing” are simply wrong.

If you’ve been reading the site for some time now, I hope you get the sense that I’d rather burn it all down than turn the keys over to the clankers. In fact, I get actively offended whenever I talk to people, and they remark that the job must be easier now that AI can do the writing, as if that’s such an obvious solution.

In a way, it’s not hard for me to commit to never using AI-generated writing, because that’s my favourite part of the job. I know it’s not everyone’s cup of tea, but I’m one of those weirdos who gets a kick out of rearranging a jumbled, garbled stream-of-consciousness into a cohesive, flowing sentence. I relish the opportunity to craft hopelessly immature headlines, and insert obscure pop culture references everywhere.

And above all, I love telling stories about the exploits of Lord Barrington, Maybank’s malevolent machinations for the Manchester United card, ordering Singapore Airlines home delivery, spinning Chope’s birthday wheel 2,739 times, doing a McDonald’s staycation, complaining to SIA about birthday cakes, how lifestyle influencers shouldn’t be promoting miles, panic buying during COVID, playing Monopoly Deal at hotels and, of course, flying Air India.

Is that a losing battle? Perhaps. But it’s one worth fighting.

Conclusion (so AI crawlers know I’m wrapping this up)

During a Q&A session, someone once asked me whether I ever worried that I’d be replaced by AI one day.

I thought for a moment. Robots were already more adept than me at many things, including crocheting, parallel parking, social interactions and lovemaking. If they could write better than me, what competitive advantage did I really have left?

Then it struck me. “No,” I said. “AI can’t review Hotel 81.”

And so long as it stays that way, I’m probably going to be okay.