AI Can’t Steal Our Words (Yet)

So, the robots are coming for our jobs, eh? First the truck drivers, then the artists, and now… creative writers? Hold your horses. While the rise of Large Language Models (LLMs) like ChatGPT has definitely stirred the pot, let’s not start drafting our AI overlord acceptance speeches just yet.

The Stochastic Parrot Problem

LLMs, at their core, are sophisticated prediction engines. They Hoover up vast quantities of text from the internet and learn to predict the next word in a sequence. Think auto-complete on steroids. Claude Shannon proposed this in 1948, Noam Chomsky hated it. Now it is real!

This “stochastic parrot” approach, as some linguists call it, has limitations. LLMs regurgitate patterns they’ve already seen. They are excellent mimics, but originality? Not so much.

Imagine tasking an AI with writing a moving eulogy, or a gut-wrenching break up letter. It might generate something grammatically sound and structurally coherent. It might even sprinkle in a few well-placed metaphors lifted from Hemingway. But will it have the emotional depth, the lived experience, the soul of a human writer? Probably not.

Creativity: More Than Just Word Prediction

True creativity involves more than just stringing words together in a statistically plausible way. It’s about having something to say. It’s about perspective, insight, and the ability to connect with an audience on an emotional level.

An LLM can generate text based on a prompt. But the creativity lies in the prompt itself. If you, as a human writer, already have a clear vision, an LLM might be a tool to help you flesh it out, to overcome writer’s block, or to explore different stylistic options. But it can’t replace the initial spark of inspiration. It cannot generate the novel idea.

Essentially, LLMs are sophisticated cut-and-paste machines, albeit ones that can cut and paste in ways that are surprisingly convincing. But they are still, at their heart, rearranging pre-existing material.

LLMs: Useful Tools, Not Replacements

That’s not to say LLMs are useless for creative writing. Think of them as advanced brainstorming partners. Need help crafting a compelling opening paragraph? Stuck on a plot point? An LLM might offer some unexpected suggestions.

In fact, software developers have already seen the benefits of AI assistance for coding tasks, like quickly generating boilerplate code or database queries. The trick is to recognize the tool’s limitations and use it to augment, not replace, your own skills.

Prompt Engineering: The Emperor’s New Clothes?

There’s a lot of buzz around “prompt engineering” – the art of crafting prompts that elicit the desired output from an LLM. This often involves techniques like asking for an outline first, or requesting the AI to “show its reasoning.”

But let’s be honest: if a particular prompt-engineering technique consistently yields better results, it’s only a matter of time before it’s baked directly into the model itself. The need for explicit “prompt engineering” will likely diminish over time. The AI will learn your prompting strategies by watching you prompt.

The Human Element

Ultimately, creative writing is a deeply human endeavor. It’s about exploring the complexities of the human condition, sharing our stories, and making sense of the world around us.

While AI can undoubtedly generate text that is technically proficient, it lacks the lived experience, the emotional intelligence, and the sheer weirdness that makes human writing so compelling. So, rest easy, wordsmiths. Your jobs are safe for now. The robots may be able to mimic our style, but they can’t replicate our souls. Yet.