Using ChatGPT to write a blog post about using ChatGPT to write blog posts
As content creators turn to large language models (LLMs) like GPT-3 and ChatGPT to crank out blog posts and articles, the internet is starting to see a surge of mediocre content. While these models can generate human-like text with impressive fluency, they often lack the creativity and thoughtfulness that goes into crafting truly great content. As a result, the rise in the use of large language models threatens to oversaturate the internet with formulaic, unoriginal material.
To avoid being drowned in a sea of mediocrity, it’s important for content creators to focus on producing high-quality, original work that truly resonates with readers.
One way to do this is by generating original ideas as the writer and only using LLMs to frame those ideas in a more digestible prosaic way or as a kind of sparring partner for ideas (which is in large part how this article was generated).
Another way is to tap into the unique experiences and perspectives that only a human writer can bring, and use large language models to supplement and enhance this content rather than relying on them to generate it entirely. (Interestingly this suggestion came entirely from ChatGPT and has not been edited, which is laughably ironic and makes me, Theo, feel like more of a machine than the machine!).
Thirdly, the value of empirical data and real-world studies will become far more valuable. These studies will differentiate truly insightful data-driven research from the high-level waffly opinion-only pieces that ChatGPT seems to generate. (Sadly I wrote this entirely myself, although I would love it if ChatGPT was so self-aware and self-deprecating!).
Finally, content curators, fact checkers, and editors will play a vital role in sorting through the vast volume of material being produced and selecting the most valuable, relevant pieces. These professionals will need to have a sharp eye for quality and be able to distinguish truly informative, thought-provoking content from superficial filler. By carefully promoting the best content, curators and editors can help raise the overall level of discourse and make it easier for readers to find truly valuable information online (this paragraph was given as an idea to ChatGPT’s which worded and unpacked it entirely). It is likely that there will be an intermediate stage with an “AI vs AI” battle before the AI starts handling more of the work involved in writing beyond just the writing itself. AI will be used to filter out original content with genuine research and empirical data while AI-generated content will morph to look more and more like legitimate research, there are even examples of ChatGPT making up studies that never happened already! (This section was human generated)
I have used ChatGPT to generate as much of this article as possible and found that its current iteration is helpful primarily as a sparring partner by asking it to suggest things like “what are the effects of having too much AI generated content” or “please suggest 3–5 highly engaging and alluring titles for this article?”. It’s also helpful for unpacking a simple prompt into a prose format, such as “add in a comment that the role of the content curator and editor become ever more important”.
For those interested, the prompts used to help generate this article and their responses are linked here.
Credits:
- OpenAI & ChatGPT
- The New York Times: while drafting this I asked the model to “make it read less like a undergraduate essay and more like a well-written and easy to read New York Times article”. While ChatGPT didn’t do a very good job of it, perhaps some credit is due to the New York Times, as well as all of the other data sources used to train the model!
