My AI Writing Crisis...and Why Your Words Matter More Than Ever

Part 4 of "Musing on AI - By Dr Ruchi Sinha" (31st March 2025)

I Almost Quit Writing (And Why You Shouldn't) ✨

Last year (2024), I stared at my laptop screen, the cursor blinking accusingly on a blank document. "What's the point?" I whispered dramatically to my coffee mug. "ChatGPT could probably write this faster, better, and without spending hours of my life re-writing each line after scrolling 50 research papers for 'evidence’ and ‘inspiration’.

Sound familiar?

Lately, my inbox has been flooded with messages from fellow writers, researchers, and chronic overthinkers asking the same existential question: Why bother writing when AI can do it for us? It's like we're all violinists who just discovered synthesizers and are collectively freaking out.

I get it. I really do. When something can do in seconds what you've spent years mastering, it feels like watching your professional identity have an identity crisis. (My therapist would have a field day with this metaphor!)

But here's the plot twist: This isn't when we should retreat from writing—it's when we should double down on it. Like that one friend who started baking sourdough more intensely during the pandemic while the rest of us were just trying to remember what day it was. 🍞

The Great Data Buffet Has Closed Its Doors 🚪

Let me paint you a picture: For thousands of years, humans have been creating language from actual lived experiences. Every poem, novel, textbook, TED talk, snarky tweet, and heartfelt blog post came from someone who had, you know, a body. Feelings. That weird experience at the grocery store that becomes an extended metaphor. Then we uploaded ALL of it online. (Including those LiveJournal entries we'd rather forget. Yikes.)

And what happened? Generative LLM AI models like GPT-4 and Claude showed up to the all-you-can-eat data buffet with empty plates and bottomless appetites. They weren't trained on US. Our masterpieces, yes, but also our 2am Wikipedia edits, our passionate defenses of TV show endings, and that one perfect comeback we crafted six hours after the argument ended.

But here's the uncomfortable truth I've been avoiding: The data buffet is closing up shop.

According to research from Epoch AI, we're on track to exhaust all high-quality human-generated internet text for training AI models between 2026 and 2032. Even Elon Musk—a man not exactly known for understatement—admitted that companies have "scraped the internet dry."

It's like we've reached the bottom of the content barrel and started to hear that awful scraping sound. You know the one—like trying to get the last bit of peanut butter from the jar with a spoon that's slightly too small. 🥄

When AI Starts Dating Itself (The Bad Kind of Self-Love) ❤️‍🩹

So what happens when AI runs out of fresh human writing and starts learning from... itself?

I'll tell you what happens because I've seen this pattern before (flashback to my brief stint in academic publishing): we get intellectual inbreeding. Scientists call it "model collapse," which sounds much nicer than "creativity eating its own tail."

It's like if you photocopied a photocopy of a photocopy until all you had left was a gray blob that vaguely resembled your original document. Each generation loses something essential.

Pro Tip: The best creativity has always been messy, surprising, and deeply human. It needs the texture of lived experience—the time you fell off your bike and discovered your favorite ice cream shop, the way sunlight hits your kitchen table at 7am, the specific feeling of relief when someone finally understands your point.

AI doesn't have any of that. It simulates. And simulation of simulation of simulation gets you...bland oatmeal. (And I say that as someone who genuinely enjoys oatmeal! Just not for every meal forever.)

The Missing Secret Sauce: You + AI = Magic ✨

Confession time: I've been using AI tools in my writing process for over a year now. And you know what? It doesn't feel like cheating—it feels like having the world's most enthusiastic, slightly weird research assistant, expert editor, and perspective-pusher.

When I use GPT-4 to help me brainstorm metaphors for explaining complex psychological concepts, I'm not replacing my creativity—I'm extending it. I still choose which ones sing and which ones sink. I still refine, build o,n and shape the final message with my unique perspective and lived experience.

It's like the difference between karaoke (just singing along) and jazz improvisation (taking something familiar and making it newly yours). We're learning to jam with AI, and that collaboration produces something that didn't exist before.

The AI-Human Partnership Toolkit: Baby Steps Edition ⭐

🌟 The Idea Expander: Use AI to generate variations on your core idea. If you're anything like me, you'll find that rejecting nine mediocre suggestions helps you clarify what you really want to say. (Warning: May cause excessive clarity)

🌟 The Draft Buddy: Write a messy first draft, then ask AI to help identify gaps in your logic or places where you could add more detail. It's like having a friend read your work who won't get offended when you ignore their advice!

🌟 The Metaphor Machine: When you're stuck explaining something abstract, ask for metaphor options. Pick the one that makes you go "oooh!" and then make it your own. (I've found the truly brilliant ones usually need human tweaking to really shine.)

Your Writing Is Training the Future (No Pressure!) 📚

Here's a perspective shift that hit me like a cold shower on a Monday morning: The writing we do today isn't just for us or our readers—it's actually shaping what future LLM AI will learn from.

Let that sink in.

Every time you publish something online that's genuinely yours—that captures your unique perspective or experience or expertise—you're not just expressing yourself. You're contributing to the dataset that will train the next generation of AI.

And that next generation will be what our children, students, and future selves will use to learn, create, and solve problems. So I have to ask myself (usually while staring dramatically out windows): what kind of intelligence do I want to leave behind?

Do I want to leave behind endless variations of the same safe, predictable content? Or do I want to contribute something that could only come from my particular messy, contradictory, embodied human experience?

Don't Just Consume. Create Your Way Forward. ✍️

I've made a decision recently. I'm going to write more, not less. And I'm going to write with AI, not against it.

Because here's what I know about human creativity: it thrives under constraints. It flourishes when challenged. It grows when it has new tools to play with.

(Remember when we all thought calculators would make us worse at math? Okay, bad example—they kind of did for me. But you get my point!)

So here's my invitation to you: Don't just scroll. Don't just consume. Don't just prompt AI to regurgitate variations of what already exists.

Instead—live fully in your weird, wonderful human body. Create from that place. Collaborate with AI to push your ideas further than they could go alone. Then share what you make.

Because when the next great scrape of the internet happens (between 2032 and 2034, I suspect), I want the machines to learn from what we dared to create—not just what we dared to copy.

And I want them to learn that humans, with all our flaws and contradictions, still have something essential that no algorithm can simulate: the courage to create something new.

---

Now I'd love to hear from you! How are you using AI in your creative process? Have you been writing more or less since these tools emerged? Drop a comment below—let's figure this out together!