Encounters with AI

June 2, 2024

Encounters with AI


R

ecently, someone reached out to me about writing. This person (let’s call them Z) wanted to know how to write for a paper. The profile looked shady but upon further investigation seemed unsuspicious enough to continue the conversation or at least investigate further.

I noticed that Z could not pick up simple phrases in the texts I was sending. I asked them to share some work that they had done so far. They shared a whole portfolio of recent writing. This included pieces Z had written for national papers. All very recent. A lot of them, considering the thin time duration within which they had all been published. The language in those was flawless, although of the search engine optimised variety.

Unsurprisingly, the portfolio also included SEO ‘click-bait’ articles. The odd thing was the national paper. That stood out.

I noticed that Z had no picture of themselves on any of the writing profiles online. I tried to look for signs of AI-processed writing, there were no obvious ones. Now, as someone who got excited about large language models early on and tested the capabilities of ChatGPT aplenty, I think I am rather good at spotting generated text. The text in Z’s articles didn’t look like it. Yet, the experience of texting Z was extremely awkward. It was a combination of bad English language fluency and a lack of pop culture familiarity.

What does Z write about? CULTURE. How? I have no answers. But I have theories. One theory is that they are really good at using large language models like ChatGPT. However, you will quickly realise the limitations of GPT models if you send them out to the editor without first reviewing it.

So, another theory is, maybe there’s a network of people. Maybe behind the screen is not one but many people working together to chunk out work with minimal effort and make maximum bucks.

Much of the press online is already click-bait. Are we now entering the phase where being a writer who gets published takes nothing more than a badly written prompt in a chat box? 

But maybe I don’t know. I am not invested enough to pay for the paid GPT 4 that I am told is markedly better than the free Chat GPT 3.5. Maybe you can use this premium version to chunk out ready-to-publish articles. As artificial intelligence gets better and large language models get more sophisticated, it will become more and more difficult to spot it. I really hope not but this seems inevitable.

What bothers me is that it can be used to create work by a person with no talent or even appreciation of what it means to write or practice any other art form. People who think people like me are stupid to waste so much time and that they are done now that AI is here.

Z told me that they felt that skill was the only thing that mattered. Were they talking about writing prompts and being a crafty networker? Or were they trying to say what they thought I would want to hear?

I think writing has little to do with skill and a lot to do with being vulnerable. It’s about putting out what you think. It’s like a private conversation with strangers who are listening to you and taking out the time to do so. It is about respecting that time. I am one to think that when reading a piece, you don’t just read what is being said but also who is saying it. It’s vulnerable and, as regular paper readers know, you start to form a bond with columnists and their columns. I remember being excited for Cyril Almeida’s Sunday op-ed. I used to look forward to the inside jokes, the satire. So much of the press online is already click-bait. Are we now entering the phase where being a writer who gets published takes nothing more than a badly written prompt in a chat box? What do you think?


Uneeb Nasir writes on culture and identity in Pakistan. He can be reached at uneeb.nas@gmail.com

Encounters with AI