AI cannibalization and the importance of our writing styles
I once saw someone compare the use of calculators in math class to the use of AI in writing. Claiming that, “if we don’t have to work so hard in creating essays for classes, then what’s the big deal?” When I initially read this take, I wanted to punch my hands through the screen to grab this person and verbally assault them for a bit.
Being taught the process of writing an essay is the point of a class that makes you write them. Unless explicitly told otherwise, AI tools are always focused on creating the output, not explaining the steps.
All language models – like ChatGPT – train off of large bodies of text. These samples of text often come from articles on the internet. And if AI is accepted into the classroom the same way the calculator has been, then we will have created an infinite positive feedback loop, causing the language model to cannibalize itself.
The problem that language models currently face is the inability to properly train off their own data. AI generated text is almost poison to the model, causing a “model collapse,” where the model produces less accurate text and increased repetition. Human-made text is essential to language models, as it contains differences and subtleties that are lost in AI-generated text.
Over time, we will become more and more reliant on AI without even realizing it. Google Overview is a great example of this, people often take what it says at face value without considering the sources it grabs.
A while back, around 139 papers in Google Scholar were found to be AI generated. This is a problem, and sets a scary precedent for the future. When AI doesn’t know the answer to a question, it will just make it up. For reliable information to be accessible in the future, we cannot rely on AI to do our research. However, AI can be a good tool for writing, just not in the way a lot of people are using it. AI can work as a foundation for research, but not as the source of information.
At the moment, AI still needs human writing to improve. So there will always be a demand for real people who can contribute in meaningful ways without the help of AI.
There is an importance in our unique writing styles, the minor differences and unique traits that someone writes with is what makes us human.

Love your article Elvis Presley, love your songs more 😉
I really like how clearly you explained the differences between AI and human writing. Your points about the importance of process, originality, and human nuance were super well thought out, and I love how passionate and detailed your argument is. It really shows your understanding and care for authentic learning.