Experimenting with AI Text Generation

ChatGPT has been integrated into many aspects of modern life, whether it be asking the chatbot to make a decision for you, to look up useless facts, or to generate full responses to prompts.

I put ChatGPT to the test, wanting to see how it would generate a response to a topic I have explored myself. When I was in my early teens, I was enamored with the show Skins, which had come out almost a decade before I found it. It was the first time I watched a show for teens that resonated with me, and one that explored topics I hadn’t seen. Addiction, mental illness, grief, a lot of issues myself my peers and I were struggling with at the time.

On this blog, I have made several posts speaking to representation in media and why it is so vital. Because of my experience there, I asked the chatbot to craft a blog post doing the same work.

Can you write a 250 word blog post on how Skins (2007) continues to resonate with teenagers?

Chat’s response did explore a bit about Skins‘ subject matter and why it would resonate with youth. However, I found the response to be too broad and a little disjointed. You can see the full response here.

I then pasted the response into a Google Doc, using the suggestion feature to edit it as my own. I felt that the generated response lacked personal flair, sounding more like an essay than a blog post. I expected the tone to be more casual and give a bit more insight into the topic.

As you can see in my edits here, I added specific examples of representation that could only come from watching the show. I also consider my edits to make the blog post more credible because I was able to link sources from media experts that reinforced my own points as a viewer.

Overall, I think this comparison showed me that text generated responses can source relevant information enough to craft a response, but they lack the necessary steps to make each point valid and credible. I don’t trust software to write blog posts or complete assignments, as the difference is inherently clear between a software and a real opinion. The generated responses lack voice, research, and passion in the subject matter.