Testing AI with Jane Austen

Once upon a time, I used to attempt Kaggle problems with the help of some techie friends. Let me be honest—we often ended up learning from GitHub, where some clever souls had uploaded their solutions for the Kaggle problems. We tried to understand their code, tweak it a bit, and test it out. It’s our version of “learning by doing.”

The other day, after a long time, I decided it was high time to give my left-brain some attention. Lately, it has been whining that I spend too much time with my right-brain. I resolved not to let the right-brain wander off… though slipping into wild scenarios had already become a daily habit.

I decided to test a large language model (LLM), a good left-brain exercise. An LLM is like a super-curious bookworm on steroids: it has read mountains of text—books, articles, web pages—and learned how words usually go together. Now it can write stories, answer questions, or chat almost like a human, though sometimes it still sounds a bit too polite and formal.

The moment I opened my laptop to prepare the input prompt, I realized something was already missing—focus, or maybe just a touch of left-brain logic. But before I could figure it out, my mind, along with my mischievous right-brain, had taken flight. Next thing I knew, Jane Austen had materialized as my coding buddy.

I imagined Jane Austen sitting beside me as I tested the model, watching my laptop with sharp curiosity, as if it were the most intriguing thing she had ever seen.

“What is this thing?” she asked, raising an eyebrow. I stumbled through a messy explanation of AI and how it works. She took notes as if I had just shown her a completely new kind of social blunder.

“Could it attempt one of my novels?” she asked after a pause.

Panic hit me instantly. I mean, who tests their LLM in front of Jane Austen? I was in a real dilemma, mostly because she’d soon realize I had no idea what I was doing.

But I had no other choice, so I fed Pride and Prejudice into the model and held my breath. And then it happened: Mr. Darcy came out sounding less like a moody romantic hero and more like a polite tech support agent with a script.

“Dear Elizabeth, sorry for any inconvenience my pride caused. Please consider activating the lifelong companionship plan.”

Jane’s response was a sigh so deep it could have rattled the piano. “Where’s the wit? The tension? The delicious awkwardness of two people who clearly fancy each other but would rather eat a whole lemon than admit it?” she grumbled. “My Darcy’s first proposal was magnificently arrogant. This…is just utterly ridiculous!”

I worried she might demand a project change, since, as her project lead, I’m not exactly an expert at training models. Trying to sound smart, I babbled about “better prompting.” She didn’t look convinced—or at least, that’s how it felt to me.

So we tried again. I explained “parameters,” while she countered with “awkward silences,” and she insisted on “more pride, more prejudice.” Somewhere between my frantic typing and her dry commentary, I realized a simple truth: AI can get the grammar, the structure, even the feelings, but it can’t catch the music between the words.

The more we messed with it, the more obvious it became: AI can copy the words, but it can’t capture the magic when Elizabeth and Darcy finally admit their feelings. That hesitant, sweet, heart-stopping moment—forever human, something no machine could ever touch!

“Besides,” Jane added with a mischievous smile, “your machine would never survive a proper drawing room. How would it understand the little moments that make social life so wonderfully human?” I had to admit, she had a point. No chatbot could ever replace the joy of watching Mr. Darcy fluster and Elizabeth blush, both pretending nothing was happening.

So there it was—our verdict: intelligent machine, sure. Charm, wit, and the flair for saying exactly the wrong thing at just the right moment? That’s still ours, all delightfully imperfect.

After all, no algorithm could ever match a perfectly timed eyebrow raise, a misstep that sparks laughter, or those little moments where she knows that he knows, and he knows that she knows he knows—purely human, delightfully unpredictable. And in the end, that’s exactly how it should feel.

P.S. As I closed my laptop, I could almost hear her chuckling. ‘My dear,’ she might have said, ‘you’ve spent all this time trying to teach a machine to write like a human. Perhaps the real question is: are we still writing like humans ourselves?’