This is the question that stopped me from blogging for the past three years. I must admit, it was not the only factor: life events and a busy work schedule also played a role. But with the advent of ChatGPT and other powerful generative text AI tools, this question was always at the back of my mind when I thought about “putting pen to paper”, or rather opening up my minimalist writing app.
When ChatGPT can generate a personalised essay on any topic in a matter of seconds, is there any point in me writing? Does my writing still have any value?
If human writing does have value, would that value be recognised in a world where AI is being to generate websites with thousands of pages of content? Can people tell the difference? Can the tools that people use to find and discover content tell the difference? Would my blog posts still show up in search engines results, amongst an ocean of AI-generated walls of text? When I struggle to put aside a couple of hours to disconnect and just write, who would even have the attention span to read my words?
The generative text revolution
in November 2022, OpenAI’s ChatGPT was launched. It was widely reported as the fastest-growing software application in history, with over 100 million monthly active users just two months after launch. Amongst the general public, ChatGPT was seen as a transformative step in AI capability, one giant leap closer to true artificial general intelligence. But it was also a product of its time, based on a software innovation proposed by Google in 2017, in the paper “Attention Is All You Need” which introduced the “Transformer architecture”. It was also made possible by the ongoing hardware advances in computing power, particularly increases to GPU (Graphics Processing Unit) power in the 2010s and which are continuing today.
ChatGPT was the first popular chatbot service powered by a Large Language Model (LLM). There are now many others, such as Google’s Gemini and Anthropic’s Claude.
Now more than three years into this AI chatbot revolution, for many people these chatbots have been integrated into their work and their lives, often taking the role that search engines previously had for finding information, but with far more advanced capabilities. Those who still prefer to use a search engine will not have been unaware of AI, because Google and most other (less) popular search engines have integrated AI answers into search engine results.
I myself have been using ChatGPT and other AI chatbots for three and a half years, I have also worked with AI as a software developer, building it into products I have worked on. It has been a learning experience: learning what AI can do, and what it can do well.
Now, I feel qualified to judge (to a reasonable extent – because of course AI is always being improved upon) whether AI can write well. And it’s my decision time: am I going to write, or leave the writing to AI?
The voice of AI
I asked ChatGPT, “Is human writing still worth it in the age of AI?”, and in seconds I got back a very reasonable answer. It addressed many valid points, which will no doubt intersect with the points I will raise in this piece.
But by default, there seems to be something about AI writing that is very recognisably AI. For example, as part of its response, ChatGPT gave me this line:
Human writing does not become worthless—it becomes harder to justify purely on mechanics alone.
ChatGPT
After interacting with AI chatbots almost daily for three years, this immediately strikes me as such a typical way of writing for AI. To start with, there is the overuse of the em-dash (“—“) – which has become such a common complaint about AI writing that people who enjoy writing tell me they no longer use it, to avoid having their writing confused with AI written text.
Then there is the common pattern “it’s not X, it’s Y”. The precise words may change, like in the example above, but the basic pattern of antithesis or contrast is over-used, often with exaggeration so that the sentence becomes true without saying much at all. So yes, of course “Human writing does not become worthless” is true – the question was whether it is “worth it” not “worthless”.
The use of “mechanics” in this generated sentence is simply not plain English – it starts to sound a bit like a teenage writer who just figured out that using complex words makes it look, at a glance, like more sophisticated writing even when such words are not well placed.
The voice of AI today feels recognisably AI generated, so much so that there are AI detection tools to help determine whether writing is AI, which seems to be reasonably accurate. So for now, there is a real advantage that humans have in using their authentic voice.
But with time, as more advanced models emerge, we can expect the voice of AI to improve. If at some stage the voice of human writing and AI writing become mostly indistinguishable, and the AI detection tools no longer work, will human writing still be worth it? Or would it be better to just point the AI at my past writing and ask it to write in my voice?
AI;DR
Another feature of today’s AI is it often produces a wall of text you will not want to read. I have found that if the first couple of generated sentences are not what I was looking for, I stop reading, and ask the AI to refine the answer in some way.
AI in general seems to produce wordy answers, and the answers often irrelevant, since it prioritises confidently giving an answer over usefulness. The acronym “AI;DR” (AI; Didn’t Read) based on the older “TL;DR” (Too Long Didn’t Read) captures this perfectly.
In a world where text can be generated in an instant, conciseness and quality of writing starts to matter even more. And this is still an area where human writing seems to win for now.
But this is another area which may improve over time – so what if AI was producing more concise, higher quality text? Would humans still have a role?
Writing from embodied human experience
I am a person, and I have a body. As I type these words, I feel my fingers hit the keys, I feel a little thirsty, and I know that soon I will feel the need to take a break from writing.
AI models do not have a body. They can describe what it is like to be thirsty based on the writing they were trained on. But they can never actually experience this feeling. AI never gets tired. And so, when it comes to writing about the human experience, which we humans are all navigating, AI will always be lacking.
It is hard to say what this means in practice for human writing vs AI-generated writing. But it must surely make a difference, especially for novel writing based on human experiences.
Today I am writing about AI, and how it feels as a human to navigate this new world of AI tools.
Perhaps in time, AI will be running within robotic bodies, and so then AI will be embodied in some sense. But it will not be a human body, so the experience it can describe will be different. Just as we recognise that it is important to have diversity in human writing, with writers of different biological sex, different races, different ages, and different abilities, there will always be some unique perspective that humans bring to their writing from their embodied experience, that can only be imitated by AI.
Writing for ordering thoughts
When we write, it is not only about the end result. Whatever others may or may not get from reading what we wrote, the process of writing about a topic likely had some impact on ordering our thinking about that topic.
When we get AI to write something for us, we may not even read it. If we do read it and edit it, we are at least actively engaged in the process in some way. But the act of choosing each and every word is a kind of practice, which improves our ability to arrange both our words and our thoughts.
In a sense, when we outsource our writing to AI, we are also outsourcing our thinking. Our ability to write may deteriorate through lack of practice, or at least stop improving. Since humans often think using words, our ability to think may deteriorate or stop improving too.
One might argue that we can develop our thinking through talking instead, and this is true, but it is a different type of thinking. Talking is often imprecise because the person you are conversing with will often not wait long for you to pick the right word. Writing allows us more time to pick our words with precision, it therefore must help us to get better at picking the precise words, phrases, or sentences that allow us to best express our thoughts.
This difference between writing and reading on the one hand, and conversation on the other hand, may seem subtle. But history tells us it is significant: the emergence of writing supported the development of civilisation, for example in the time of the Ancient Greeks and Romans. Without writing we would have very limited ability to develop technology or record history.
Later, the invention of the printing press in Europe, allowed writing to be propagated widely, and this led to The Enlightenment and then the Industrial Revolution. So we can directly trace the impact of writing (together with reading) on our thinking processes to the development of the modern world – including AI technology.
Is writing worth it if no-one reads?
With the development of generative AI, I have a sense people are reading human-written text less. By engaging more with AI tools, that leaves less time for searching the web and discovering human-written blog posts. Even when using a search engine, we are shown AI-generated text and many of the search results may have been written by AI.
This website has mostly frequently been discovered by people through using Google or other search engines. But if we are finding information using AI more, this may mean fewer people discovering this blog and therefore fewer readers. With potentially fewer readers, is it still worth writing?
It is difficult to separate writing from reading. We know that writing together with reading has shaped the modern world, but it is harder to say what benefits came from the ability to write well, not just the propagation of knowledge.
We do know however that writing is not just about communication. It is also internal: writing is not a direct expression and recording of thoughts, but part of the process of forming them. Through careful writing, the quality of our thinking can improve, we refine our understanding and tighten our arguments.
When I write for a sustained period, I become fully absorbed in the activity. This is commonly known as “flow” state. It can feel deeply satisfying, and there is evidence that, for many people, experiencing flow state boosts wellbeing.
I still hope and expect that some people will read this post. But even if there are few readers, I believe it is worth writing to develop my own thinking and simply because I enjoy writing.
If it also helps readers to develop their thinking, or readers are enjoying the process of reading, that is a bonus. If you made it this far – well done! It feels like attention spans are getting shorter these days. Please do share your thoughts in the comments below to let me know that you finished it, and take the chance to practice your own writing and thinking skills.
