With all the swirling emotions, it’s good to know that Humans still – and may always – have the edge when it comes to communicating with other Humans (which, thankfully, we are still doing). Why the absence of Human-generated content will actually make Large Language Models “dumber”. The “poisoned pool” and seven ways we still outperform.

AI Leadership, Fiona Passantino, early July 2023

AI Content Explosion

By some estimates, just over 19% of the internet today consists of AI-generated content[i]. That’s an extraordinary claim given that widespread use of generative AI portals have only been around since autumn 2022.

Truth is, words have become cheap when we have free generative tooling widely available for nearly language on earth.

Today, we are producing 7.5 million new blog posts every day; far more than any of us could ever read[ii]. Imagine if this were an exponentially exploding number of words on the internet, based on auto-generative tooling. By 2026, given current trends, 90%-99% of online content could be AI-generated[iii].

But as we discovered early, AI generative language is bland, vanilla, and always opts for the most commonly-found turn of phrase, the most widely-used terms, the “safest” bet for the most likely outcome to a partially-complete sentence. It’s the surprising, distinctly Human voice that AI can’t produce; a colourful phrase, old-fashioned expression, a surprising analogy, bit of sarcasm.

---

Workflow: Where the Human Steps In

When engaging in AI-Human creative co-creation, the Human is obliged to step in and re-write the AI’s material. In the end, it all has to go. It all has to be re-written, word for word, just like this book.

Thankfully, our Human contribution is (still) needed to ensure readability, to make the final work more precise, engaging, relevant and enjoyable to other Humans.

1. Precision

The bot tends to be long-winded, and it’s our job to collapse those rambling sentences into finer points. We still outperform on accuracy and clarity.

When asked to be “eloquent”, ChatGPT becomes a Drama Queen on her third Bloody Mary, excessive, effusive to the point of ridiculous. With no stylistic instruction, the bot will slip into its default comfort zone of blended-banana-bureaucratese that, while doing no harm, still uses 50% more words than necessary.

AI rough cut: Now, with the laughter still echoing in our hearts, let us embark on this serious journey together, ready to tackle the great mysteries and complexities that lie ahead.

Human adaptation: Remember our laughter. It fuels our journey as we face the challenges ahead.

2. Humor

AI humor is exactly what you might expect from an algorithmic amalgamation of the full collection of Human literary history reduced to a single phrase. Humans are masters in the house of hilarity, can generate the unexpected, the dark wit.

There will come a day when AI will surprise and delight us with our own language, but for now, we are still in our element.

AI rough cut: Life is short. Smile while you still have teeth!

Human adaptation: Smile; life is short and you still have your teeth.

3. Analogies, metaphors and similes

Metaphors, analogies, similes and other tools we Humans use to simplify complex concepts are utterly unnecessary to the bot. They don’t need illustrative crutches to enhance their understanding (which is instantaneous) nor do they need the motivation of endorphins to make it through an overly technical manual.

Humans need help to spur engagement and activate our emotions, so it’s only natural that we are better positioned to provide it for other Humans. AI, out of its element, falls back on tropes and canned nonsense it scrapes from the very Human internet.

AI rough cut: This book is like a wise old owl trying to keep up with AI’s breakneck speed.

Human adaptation: No earthly book can keep pace with the light-speed of AI evolution.

4. Cliché-Purging

A cliché is a phrase or expression that has been overused to the point where it has lost its originality and effectiveness. They may be phrases like “time will tell” or “the calm before the storm”.

While clichés are a cheap way to convey a common idea, they inevitably weaken the impact of the message.

AI rough cut: Buckle up and embrace the ever-changing AI adventure!

Human adaptation: Prepare for a massive change towards an AI-driven future.

5. Empathy and emotion

Because machines can only regurgitate words reflecting empathy, they are not able to adequately weave it into their messaging in a natural or authentic way.

A Human writer will instinctively find emotional pathways into the mind of her reader by considering his perspective and context. Our job is to strip the strained, AI version of empathy and replace it with our own.

AI rough cut: Remember, you’re not alone in this journey. We all face our own battles and setbacks. But it’s in those moments of vulnerability that we find our strength and resilience.

Human adaptation: In your battles and setbacks, you will feel alone and afraid. These battles reveal your core and make you strong. Strong enough to help someone else feel less alone and less afraid.

6. Context, Relevance and Getting Personal

Personalisation is one of the key benefits of AI. With the right data, messaging can be granular down to the individual. Humans will need to feed the beast with the right data to make this happen. Similarly, context and the specifics of the business case you are trying to make.

This makes all messaging relevant to the reader’s experiences or interests. If you are writing speeches, it’s up to you to provide personal anecdotes, examples or insights to make the content relevant to the audience.

AI rough cut: On a serious note, I wanted to take a moment to say how incredibly proud I am of each and every one of you.

Human adaptation: I want to express my pride in the Supply Chain team, who was able to source the paper and the ink just in time for the autumn run. Not only did they prevent a crisis, they provided us with a new supplier that’s faster and more affordable.

7. OMG-Cringe-Removal

Here’s an actual social media post ChatGPT wrote to humanise a CEO. We can classify this one under #HRviolation #nightmare #didhejustpostthat. It’s a situation we Community Managers lose sleep over now and then.

AI rough cut: 📢 Attention Team! 🎉 Guess who just discovered a hidden talent? 🕺 It turns out I’m a world-class dancer in my spare time! 💃🕺 Who knew, right? 😄 But fear not, I promise not to start a dance-off during our next board meeting! 🕺🚫 I’ll stick to leading our amazing team to new heights instead. 💪✨ Your hard work, dedication, and the occasional office dance party make this company what it is. Together, we’re unstoppable! 💪🌟 And if you need a quick dance break to get those creative juices flowing, don’t hesitate to bring out your best moves. Just warn me in advance, so I can clear some space! 😄 #CEOInTheHouse #DanceLikeNoOneIsWatching #TeamAwesome #WorkHardPlayHard

Human adaptation: Delete, start over

AI Poisoning

The more AI tools flood the clear, mountain lakes with a sudden abundance salt water, the more difficult it will be to find unsalted water. AI trains on the same well, and the more AI-like it becomes, the blander, wordier, the new models become. As the amount of AI content spirals ever upward, it poisons itself, dulling us all down in the process, threatening to destroy the internet for everyone, Human and AI alike. AI ceases to perform, becomes dumber, more repetitive, quashing the full range of the art form of Human language.

AI needs us, in the end, to stay sharp, high-performing and intelligent.

---

---

---

---

---

[ii] Nead (2023) “How GPT-3 and Artificial Intelligence Will Destroy the Internet” ReadWrite. Accessed July 11, 2023. https://readwrite.com/how-gpt-3-and-artificial-intelligence-will-destroy-the-internet/

[i] Davis (2023) “AI-generated content is detectable, new study claims” Martech. Accessed July 11, 2023. https://martech.org/ai-generated-content-is-detectable-new-study-claims/#:~:text=19.2%25%20of%20the%20articles%20tested,75%25%2B%20AI%2Dgenerated%20content.

[iii] Veda (2022) “Internet in 2026 is Non-Reliable! Could be Filled with AI-Generated Content” Analytics Insight. Accessed July 3, 2023. https://www.analyticsinsight.net/internet-in-2026-is-non-reliable-could-be-filled-with-ai-generated-content/

---