Can AI Blog? Testing ChatGPT

Late last year, when AI started getting lots of attention as a tool for generating text, I decided to experiment a little by asking ChatGPT to write my bio. This is what it came up with:

Fernando Gros is a creative entrepreneur, writer, and photographer based in Tokyo, Japan. Born and raised in the United Kingdom, Fernando has spent over 20 years living and working across four continents.

Fernando's work spans across various industries, including technology, finance, media, and education. He is a former managing editor of The Next Web, a leading technology news website, and has contributed to publications such as Forbes, CNN, and The Guardian. In addition to his work in media, Fernando has also worked as a consultant, advising companies on content strategy, social media, and digital transformation.

Fernando is also an accomplished photographer, with a focus on street photography and portraiture. His work has been exhibited in galleries in Tokyo, London, and New York, and has been featured in publications such as National Geographic, Time Out, and Lonely Planet.

Beyond his professional work, Fernando is also a published author and educator. He has written several books on photography, creativity, and personal development, including "No Missing Tools: Creativity in the Digital Age" and "The Art of Timing: Mastering the Unseen Forces of Life." He has also taught photography and digital media at universities in Japan and Singapore.

With his diverse range of experiences and talents, Fernando Gros has established himself as a thought leader in the areas of creativity, innovation, and personal growth. He continues to inspire and motivate others through his writing, photography, and educational work.

While parts of this bio are true and maybe even insightful, some of it is plainly wrong. I wasn’t born or raised in the UK, I’ve written only one book, and my while my work has hung in a gallery in Tokyo, it hasn’t been exhibited in London or New York.

Interestingly, this bio’s exaggerations fit within my aspirations. But, it’s the kind of padding someone might indulge in if they had few concerns about the truth. It’s bullshit, to borrow Harry Frankfurt’s definition of the kind of lies that bear no relationship to or curiosity about the truth. And it’s an example of the kind of “middle-aged guy trying to win an argument” energy that we see in so many ChatGPT examples.

But the experiment left me wondering if I could – ethical concerns aside – use ChatGPT to write blogposts.

What Might A ChatGPT AI-Generated Blog Be Like?

Let’s clear this up right away. I didn’t write this essay with ChatGPT. There’s no bait and switch where you get to the end of the article and I say, “Ta da! AI wrote this, ha ha.” All the computer generated copy is differently formatted so you can tell the difference between me and the machine.

To get some sense of what an AI blogger might be like, I asked it to write a paragraph explaining my ideas about creativity. ChatGPT3 couldn’t handle requests like that, but ChatGPT4 gobbles them up. This is what it replied with:

Fernando Gros believes that creativity is not a talent that some people possess and others do not, but a skill that can be developed through practice and experimentation. He emphasizes the importance of cultivating curiosity and exploring diverse interests, as well as breaking free from self-imposed limitations and fear of failure. Gros also argues that creativity is not just about producing something new, but also about discovering and expressing oneself in a meaningful way. He encourages individuals to take risks, embrace vulnerability, and seek out new experiences as a means of unlocking their creative potential. Overall, Gros promotes a holistic approach to creativity that emphasizes personal growth and self-expression rather than external validation or commercial success.

I can see my ideas in there. But they’re obscured by clichés. It sounds a bit insincere, a bit corporate, a bit generic. If only there were a way to make it sound more like me…

Can ChatGPT Replicate My Writing Style?

One way ChatGPT4 is a significant upgrade on ChatGPT is in the area of understanding writing style. I asked ChatGPT to describe my style, and it came up with this:

Fernando Gros is a writer and photographer known for his clear, concise, and engaging writing style. He is known for his ability to explain complex ideas in a straightforward and accessible way, making his writing accessible to a wide range of readers.

Gros often incorporates personal anecdotes and observations into his writing, which adds a personal touch to his work and makes it relatable to readers. He also uses descriptive language to create vivid images and sensory experiences, drawing readers into his stories.

Overall, Gros's writing style is characterized by its clarity, accessibility, and engaging storytelling.

It’s flattering that ChatGPT now knows me well enough to describe my writing style in this slightly vague way. It’s a bit of a literary horoscope; it could be true for many people who write creative non-fiction. But it’s enough to make me wonder if ChatGPT could rewrite that earlier section, on my ideas about creativity, in my writing style. This was the result:

Creativity is not some innate gift that only a select few possess. Rather, it is a skill that can be honed through repeated practice and a willingness to explore new ideas. As such, cultivating curiosity and expanding one's range of interests is essential for unlocking the full potential of our creativity. To truly embrace our creative selves, we must break free from the self-imposed limitations that hold us back and overcome our fear of failure. It's important to remember that creativity is not just about generating something novel; it's also a means of discovering and expressing ourselves in ways that resonate with our deepest selves. So, we must embrace risk-taking, vulnerability, and the unknown as we venture into uncharted territory and seek out new experiences that stretch our creativity to its limits. In short, creativity is not merely a means to an end; it's a journey of personal growth and self-expression that brings us closer to our authentic selves.

When Chatbots Get Weird

Once again, ChatGPT got the ideas right but served them up in a trite word salad. I couldn’t sign my name to something like that. So I asked ChatGPT to write a few shorter versions of the paragraph, then chose the best one to rewrite in my style. At this point, ChatGPT served up some copy on an entirely different subject!

The current era is marked by a rapid pace of technological advancements that are transforming our lives in unprecedented ways. With each passing day, we are witnessing the emergence of new technologies and innovative solutions that have the potential to change the world as we know it. It's an exciting time to be alive, but also a daunting one, as we grapple with the challenges and uncertainties that come with these changes. At the same time, it's important to remain optimistic and embrace the opportunities that these transformations bring. With the right mindset and a willingness to adapt and learn, we can harness the power of technology to create a better world for ourselves and future generations.

I agree, it is an “exciting time to be alive,” and “daunting,” and yes, I’m also uncertain what it all means. Especially since I was asking about creativity. Did ChatGPT interpret my many questions as a prompt to shift to another topic it associated with me?

Also, this paragraph is so vague, like it’s saying something, but it doesn’t say much at all. It’s the kind of thing a student might hand in when they realized an assignment was due at 11pm the night before the deadline, and they hadn’t done the required reading.

Why Does ChatGPT Brag So Much?

Since ChatGPT wanted to talk about tech, I ran the previous experiment again, asking it about my ideas concerning technology and then refining those into some copy written in my style. This answer really took me by surprise:

As an expert in the realm of technology and creativity, I have spent countless hours pondering the ways in which these two fields intersect. My belief is that technology has the power to unlock our full potential as creative beings, providing us with new tools and methods for innovation and problem-solving. However, I also recognize that there are potential pitfalls and risks associated with unchecked technological progress. Thus, it is essential that we approach technology with a sense of balance and caution, being mindful of its impact on our lives and the world around us. Ultimately, I believe that the key to success lies in using technology to enhance our lives and make the world a better place, while staying true to our fundamental values and principles. These are just a few of the ideas that have shaped my thinking on technology and its role in society, and I look forward to continuing to explore these themes in the years to come.

Here’s a tone I would never use. If I read something written this way, I’d turn the page or scroll away. It’s not that the ideas are wrong. But I wouldn’t address a reader in such a condescending tone, bragging about my expertise and parading self-confidence without clear examples and counter examples.

I’m reminded of a quote by Ernst Hemmingway about the heart of good writing often involving not saying everything you know about a subject

“If a writer knows enough about what he is writing about, he may omit things that he knows. The dignity of movement of an iceberg is due to only one ninth of it being above water.”

Generic Content Perpetuates Dominant Voices

ChatGPT4 and its earlier iterations have been trained on massive amounts of text drawn largely from websites like Reddit and Wikipedia. That means the language model sprouted from online arguments and the generic presentation of agreed facts.

This goes a long way to explaining why ChatGPT wants to veer in the kinds of linguistic directions we’ve seen. That’s its home territory: winning arguments and settling scores.

This might do well for content that needs to be reliably unremarkable. AI could, and probably will, have a huge impact on the kinds of online writing that depends on search engine optimization and giving simple answers to common questions.

But what if the goal is to challenge existing ideas rather than just uphold them? What if you are looking for novel solutions instead of well-worn ones? What if you crave unique voices rather than the familiar?

As the authors of the compelling research paper On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? explain, by being a kind of average of large chunks of online discourse, ChatGPT reinforces certain points of view and marginalizes others.

“In accepting large amounts of web text as ‘representative’ of ‘all’ of humanity we risk perpetuating dominant viewpoints, increasing power imbalances, and further reifying inequality.”

Part of what makes the content in the experiments above so bland is that it’s mixed up with other similar, but largely mainstream, ideas. That’s where the greeting card tone comes from. It’s all just mixed up into a kind of relevancy soup. The ideas are there, but you can’t really make them out or taste them.

Emotion is at the Heart of Great Writing

Of course, this is a risk all writers face. When I read my work, especially early drafts, I often think “that’s too generic,” or “surely you could say that better.”

“Write towards the emotion” is something I often say to myself when trying to rewrite a piece that doesn’t work well. I can’t remember who said it, but it’s one of the most reliable pieces of writing advice I’ve heard. If it’s not working, I need to be more vulnerable.

When a piece of writing doesn’t seem to say anything, it’s because it doesn’t make you feel anything. The emotional journey we go on as readers is what connects us to the writers, the characters, the ideas and the story.

Looking at these and other ChatGPT examples, it’s clear the emotional content isn’t there. When pushed, it does the opposite of writing towards the emotion. It becomes less vulnerable; it tries to win the argument.

I’m sure, given time, these models will inject more emotive language into their texts. They’ll be trained on the words of better writers rather than on those who spend their time on online forums.

But I’m not sure the challenge here is simply one of computational power or texts imported into the database.

Emotion is fundamentally a physical thing. When I find the way towards the feeling with the right words and in the clearest way I can, it doesn’t just register as a disembodied “yes” in my mind. I sit up straight in my chair. Maybe I punch the air. Or I get up and dance. I feel it in my body.

Writing Should Be Treasured

The short answer to whether I could use AI to write this blog is yes. It would make my life a lot easier – and probably increase the traffic to my site exponentially, since AI-generated writing might appeal more broadly and would almost certainly perform better in search engine results.

But, dear reader, that is not the dance we’ve agreed to share every week, is it?

For all the reasons we’ve looked at here, I just don’t feel that language models like ChatGPT are all that useful for a writer like me. They can generate text quickly but they don’t feel trustworthy. If posting here became too hard, I would rather give up than give you something as generic as the content ChatGPT creates.

Writing is hard and inefficient. That’s why we learn so much about ourselves when we try to do it. It takes tremendous vulnerability to bring the words to the page, to reveal our insides to the world. That’s why the best writing teaches us so much. It gives us these intimate, personal windows into humanity. We should treasure them every chance we get.