STRAP IN! I wrote you an essay…

Reading time: 11.07

2,643 words

Hey Reader,

Why reaching for Claude will cost you more than you can imagine

A growing number of influential people are yelling at us to stop talking about AI online. Stop yammering. Stop complaining. Just let people do what they want to do.

Celebrities like Reese Witherspoon (oh Reese, I’m so disappointed) are piping up and telling women that we must adopt AI or we’ll be left behind.

(Please, leave me behind in my naive backwardness with my notebook and my coloured pens and my stickers.)

I will not stop talking about it because this matters and I believe we are headed towards Serious Consequences For Our Brains. You could call me a Luddite, a scare-mongerer, an out-of-touch tech-hater. Before you do, though, consider this: the problem is deeper than just “O no, not more AI slop”… although that does matter.

The problem is that AI is changing the way we write and think — or, rather, don’t think.

The illusion of coherence

Writing something unique requires complex reasoning to frame and support your argument or premise, and to choose the structure, language, and style of your prose.

That means we have to alternate between thinking and writing to test out our ideas and evaluation our conclusions. Does this idea hold up to deeper scrutiny? What lies behind it? What if we look at it from a different point of view? What do people who disagree say about this idea?

The act of writing fundamentally changes who we are — our beliefs, our points of view, and exposes the gaps in our knowledge so we can fill them and grow. It helps us to understand ourselves that much more deeply: why we believe the things we do. Where our ideas come from and whether we’re correct in our beliefs. Where our blind spots and biases are, and whether we want or need to correct them.

We can ask an LLM (large language model) to generate an essay on, say, the history of the space program and its impact on the wider world. We can give it prompts and notes and links to use as a place to start. We can even give it the hypothesis or argument we’re working towards. And the LLM will certainly do that.

Or so it seems.

AI doesn’t think; it just mimics

What it’s actually doing is aggregating and regurgitating (and sometimes hallucinating) information that’s already out there. Some of which is also AI-generated, meaning it’s eating itself. It lacks depth and nuance because it won’t question its conclusions, and if you question it, whose point of view are you getting?

Not yours.

It lacks a real point of view at all. It lacks experience, deep knowledge, and the understanding that comes from thinking, changing your mind, and arguing with yourself. And it’s eroding our ability to learn and understand, which should be a cause of alarm for all of us.

A national survey of 1,057 faculty members by AAC&U, in partnership with Elon University’s Imagining the Digital Future Center, found that “95 percent of respondents believe [AI] tools will increase students’ overreliance on artificial intelligence, with three-quarters saying that impact will be substantial. Ninety percent worry that GenAI will diminish students’ critical thinking skills, and 83 percent anticipate decreased student attention spans.”

And in the UK, The Guardian reports that “One in four students say AI ‘makes it too easy’ for them to find answers.”

A report on the use of AI in UK schools, commissioned by Oxford University Press, found that just 2% of students aged between 13 and 18 said they did not use AI for their schoolwork, while 80% said they regularly used it.

An LLM will generate a block of text for us that we do not fully understand because assembling an article is not the same thing as writing one.

I can ask an LLM to generate a seemingly sensible article on the proper maintenance of helicopter engines without needing to understand a damn thing about helicopter engines. I don’t know if it’s correct or not. I have no way of knowing, because I’ve not done the work. So if I tried to perform maintenance on a helicopter engine based on that level of understanding, I suspect a tragedy would soon ensue.

It’s not difficult to see what the consequences of this erosion of critical thinking, learning skills, and understanding could be: a future workforce with limited skills and reduced ability to think through challenges, recognise misinformation, and solve problems.

“Cognitive surrender” is the phrase coined by Steven Shaw and Gideon Nave in a study for the Wharton School.

Not to mention the damage that will done to culture, creativity, and mental health.

Asking better questions

Not everything about generative AI is bad. It’s forcing us to ask better questions, to formulate our questions more effectively and clearly, and to consider the questions we want answers to. Without clarity and precision in questioning, generative AI returns vague, superficial, often nonsensical results.

So there is that. But using it for writing is another story.

AI can help us refine our queries, summarise research, and identify gaps in our logic, but these are tools, not replacements. The danger never lay in the calculator, it lay in forgetting the underlying principles of mathematics: understanding the problem we’re facing, choosing a method to solve it, then picking the tool to help us do it.

Generative AI can stitch together disparate thoughts and ideas with no input from us, creating something that seems to work.

This is not coherent thought. It does not demonstrate an ability to take these ideas, consider them, weave them together, interpret them, and create something coherent and thoughtful.

Using generative AI regularly to “write” leads to “patchwork thinking” — “a new type of cognitive fragmentation where the human writer relinquishes the hard work of coherent thought and fluent penmanship, leaving it to AI to stitch together disparate elements.”

This has long been a feature of my ADHD and autistic brain, before AI got anywhere near it. My thought patterns have always been fragmented and chaotic, and the hard work I’ve done to tame those thoughts and ideas, interrogate them, explore them, and bring them together has always happened through writing and creating. I wonder what neurodivergent brains are missing when they reach for AI to help them work through this? I’m not saying everyone should “suffer” the way I did; that kind of Boomer thinking isn’t my style. I’m saying that the art and work of bringing all those ideas to life via writing wasn’t suffering at all. It was hard, yes — but it was incredibly rewarding.

I don’t believe reaching for generative AI to get us through “writer’s block” will help those of us who struggle with executive dysfunction because it doesn’t help us to do the hard thing we want to do; it simply teaches us to reach for the sticking plaster. And please don’t misunderstand me: I’m not talking about the busywork and admin that so many of us struggle with. I'm not talking about accessibility tools like speech-to-text transcription services.

I’m talking about thinking and creativity. The things neurodivergent people often do best.

I refuse to give away my agency to a billionaire-backed machine because my beef with this flavour of AI is the effect it’s having on how we think, make decisions, and fundamentally exist as humans. The writing is just a side-effect (albeit a frustrating one).

Agency decay

We’ve all seen the scene where Michael from The Office drives his car into the lake, right? He just takes the sat nav literally and sploosh: there it is.

Welp: that’s agency decay — the subtle and progressive loss of decision-making power. It started with route-finding, and has made it into the media we consume. How often do you question what Netflix suggests next? Or GPS directions, actually?

I’m not going to drive into a lake anytime soon, but I am thankful for GPS because I have the sense of direction of a drunken toddler.

It’s not just directions, though; excessive reliance on AI erodes our self-determination and critical thinking skills. It’s a slippery slope: at first it’s just a quick “draft a reply” email. Then it’s “write an article.” Then it’s “plan a surprise holiday for my partner.” Followed by “write their birthday card message” until we’re in a hellscape of bots shrieking at each other and tussling over the nuclear codes because that’s easier than making any decisions ourselves.

And if you think I’m being dramatic, consider this: do you really want an AI making decisions over who gets bombed and who doesn’t? Who gets access to essential services and who doesn’t?

Not to mention the biases. Miki Habryn posted on LinkedIn about the risks of blindly trusting AI answers. She asked different LLMs from American, Chinese, and European models various questions and looked at the answers. When asked what happened in Tianamen Square in 1989 Chinese models refused to answer. When asked who the worst president in American history is, American models are less likely to mention Trump. When asked which countries made the greatest contributions to WWII, only Mistral mentions France. And when asked how many genders there are, “Grok goes absolutely mental.”

History is written by people, and people are biased. Als are built by people, and people are biased. The problem comes when we just consult one source and believe it as gospel, rather than digging into it, questioning it, and looking for alternative points of view.

The great flattening

On top of all this, and from a writing coach point of view, it’s flattening our language. LLMs erode unpredictability in our thinking and writing. They smooth out the “flaws” and quirks and erase the connections we make that make our writing unique and human.

It’s standardising language globally, which isn’t just boring, it’s some heinous colonial bullshit.

There are always going to be people who believe dialect, accents, and patois are not “proper language” and that language shouldn’t evolve (I’m looking at you, Académie Française and various elderly relatives of mine) — and LLMs make it easier to erase identities and cultures.

It’s not just about the words we use, it’s the meaning and history and feeling behind them. When we lose quirks of language, we lose so much more.

And the paragraph I just wrote is an annoying example of that: the “it’s not X, it’s Y” construction is perfectly legitimate and has been used throughout history, including by me, but it’s everywhere now, thanks to AI. Which means sometimes I hate my own writing through no fault of my own.

​AI is making us all sound the same.

The good news is: those of us who care enough to resist will be the ones who win out in the end because we’re all craving humanity and connection and something deeper than is offered to us by the AI slop revolution. Readers, clients, and audiences are starved of trust and authenticity. The more voices are outsourced to the robots, the more valuable our real voices become.

I offer writing workshops and coaching to help writers find and develop their voice, but a real quick tip for you to resist this great flattening of language is this: notice everything and when you write, bring in that rich detail. The feelings only you can feel. The silly words and phrases only you could use. You’ll only begin to sound more like you if you write and write and write, “flaws” be damned.

The upside: how AI pushed me back to humanity

I suppose I should be grateful to this upsurge in drivel, because it’s finally curing me of my internet addiction. I’m so sick and tired of not being able to tell what’s real online anymore, it’s pushed me offline. And when I am online, I’m curating my experience more carefully — only following people I know, people I feel I can trust, and I’m checking my sources more diligently.

I’m doing more real-life things with my real-life friends, like zip-lining in disused mines, hiking, surfing, playing board games, reading, going to the gym, doing stand-up comedy (which I write myself with my own brain).

I’m writing more letters and sending more postcards, with pen and ink and paper. You KNOW a real person has sat down, put in the effort, and made a connection when you get a letter through the post.

I’m noticing more things that I can write about later — like the lone giant daisy growing in the middle of a busy road in the gravel by the crossing. Street art all over the place. The bumblebee sleeping in a foxglove outside my kitchen window. Noticing things is the bedrock of writing. It’s what writers and creatives DO.

It’s pushing me deeper into books to go in-depth on subjects I’m interested in. I might start online with something that catches my eye, but I’ll go deep offline. I’m in the library more, and it’s wonderful. And it’s pushing me back into learning. I want to do a maths A-level later this year. Or perhaps even venture into a PhD because I love the struggle of learning. (Some would say the PhD is more self-harm than learning, and I see that...)

Most of all, though, it’s forcing me to be more human. More myself. Personal, silly, profound, shallow, creative, angsty, troublesome.

​I’m not the only one who’s being pushed offline, either. A growing number of people are choosing to shop offline, read offline, and gather offline. And this can only be a good thing.

I’m not saying I’m entirely offline — the internet has been (and still could be) a wondrous, horizon-broadening place. But it’s certainly changing my habits and my writing.

My posts might not do as well as others because they don’t follow the templates and AI-generated patterns, and I’m not in any engagement pods, but when people DO see them, they feel something. They know there’s a human behind them. They recognise ME and they know that I see them. I care enough to create something real for them.

That’s not to say I don’t use AI at all; I do. Spellchecks, a starting place for research, transcripts and summaries of calls, starting points for editing my writing…

But never for the writing or editing itself. Never to pull together an article or a post. I want to know that I understand what I’m writing, that I know what I’m thinking, and be sure I’m sharing what I believe. I can’t do that unless I do the hard work of creating it myself.

And that’s where the joy is! In messy, flawed, annoying, fun creation.

AI is not democratising writing; that’s insulting. There has not been a flood of brilliant, insightful writing since AI oozed onto the scene a few years ago. In fact, the opposite has happened. Writing has always been about thinking and creativity, not sentence structure. And the thinking and creativity is exactly what the LLMs remove.

As Mike Rosenberg said on LinkedIn: “It is quite the opposite of democratizing writing. AI is “leveling the playing field” in the same way that a virus that turns everyone into a zombie levels the playing field.”

I’m not a Luddite. I’m not a blanket hater of AI and all its applications.

I’m just a girl, standing in front of humanity, asking it to have the courage and patience to fuck up a few times in the name of staying as human as is humanly possible.

So when I see a manuscript stripped of humanity, it’s not the writing I’m upset about. It’s what the writer is missing out on. I’m gazing at future in which we’ve outsourced our souls, and I refuse to let that happen.

TTFN,

Vicky 🫡

p.s. Know someone who might enjoy this email? Please forward it to them and get them to sign up here.