The Unreliable Narrator: My Experience with ChatGPT

Published On Sun Apr 20 2025
The Unreliable Narrator: My Experience with ChatGPT

ChatGPT told a tale of my lamb Wonky that just wasn't accurate | The ...

I HAVE a confession to make. I use AI.

Top Pitfalls to Avoid When Implementing AI in the Enterprise ...

Before you cancel me, let me tell you how – because the how is the ­single most important part of the ­conversation around artificial ­intelligence, and it seems to be the part that is ­­­mentioned least.

The Scottish Islands News Roundup

This topic sprang to mind last ­Saturday night as I prepared the ­Scottish Islands News Roundup. I compile the main stories from the islands, the interesting ones, the fun ones and the frustrating ones, and send it out on a Sunday morning. To do that, I have a world of Google alerts set up. Every day, they dump headlines into my inbox.

Wonky lamb with wobbly walk becomes internet star

Naturally, I have one for Tiree. And ­no-one was more surprised than I was to ­discover a story about a Mairi MacLeod in Lewis with a lamb called Wonky. The story featured my TikTok videos and was ­accompanied by a set of impressive factual inaccuracies and a series of quotes which were entirely fabricated.

READ MORE: John Swinney issues statement following Nigel Farage's SNP Holyrood admission

The Inaccuracy of the Story

I’m not going to name and shame, ­because after using the story as fodder for my newsletter’s entertaining “and finally” section, I reached out and asked them to correct the story. To give them their due, they did. I sent them some actual quotes, they named me correctly, gave the correct location and as an apology, they linked to our website – which was appreciated.

The person at the end of the interaction was helpful and apologetic. I asked who had written the story, as to me, it reeked of AI used very badly. The correct story was widely syndicated by this point, the facts were all there to be had. To get a story so incredibly inaccurate was either the ­result of someone writing it with their eyes shut, or it was sheer sloth – ask AI to do your job, and then fail to do any form of fact-checking.

The Tale of Wonky the Lamb

It did, with great confidence. Here is it verbatim.

“Ah, the tale of Wonky the lamb is a heartwarming story that’s captured the hearts of many!

Role of AI in shaping modern society and advancements - IABAC

“Wonky was born with a twisted neck, giving her a unique appearance. Despite this, she thrived under the care of her crofter, Mairi MacLeod. Mairi shared Wonky’s journey on social media, and the lamb quickly became a sensation. People from all over found inspiration in Wonky’s resilience, with one commenter noting, ‘If Wonky can keep going with her head literally on sideways, I can get through today’.

“Wonky’s story serves as a reminder that even with challenges, one can lead a full and inspiring life.”

The Use of AI

If you haven’t met this concept ­before, a VA is a remote, human, ­secretary who is paid to manage your life, files, ­appointments and reminders.

They are very popular in the United States, ­particularly with employees of tech ­companies who have ideas above their station.

Avoiding AI Pitfalls: 5 Downsides You Need to Know - DMNews(Image: Rhoda Meek)

I can’t afford a VA and have no desire to add complexity to my life by having to manage someone who manages my life. I have enough problems as it is. And so, I use AI to do the tasks that take up a lot of time, but which require very little in the way of conscious thought.

The Pitfalls of AI

One example of such a task would be sucking the text out of a PDF, and formatting it neatly. Another is creating spreadsheets from information I paste in. I ask it to double-check punctuation, to remember the latest task or idea that popped into my head – then at the end of the day, I ask it to spit out a list of all the random things I asked it to remember, categorised by topic and perceived priority.

I needed to write a very formal letter last week. I could have spent ages searching for examples and templates to ­structure it correctly and ensure that it could not be misinterpreted – but I asked ChatGPT to do it instead – then I edited it.

I edited it because it doesn’t get ­everything right. I have to double-check the spreadsheets, I have to review the PDF to make sure it didn’t miss ­anything. It ­misses a lot. It ­misinterprets my ­instructions on a regular basis, and there is an argument that the time I spend ­double-checking would have been better spent doing the task myself in the first place. As a fan of efficiency, I have tracked tasks, and it does save me time, but not as much as I think it is saving me.

Challenges with AI

Humans like shortcuts. And so does AI – because it was built by humans. AI is biased towards believing the ­information it finds online, because that is what it has been told is the source of truth. ­Humans are biased towards ­accepting that ­information because we are lazy, and because we are gullible. Believing ­everything you see online requires a lot less effort than actually thinking about things.

The Mirror of AI

Then there’s just the lack of ­understanding about how this technology is built. Computers are not objective. ­Unconscious bias is present in software, because it is built by people – and we all have unconscious biases. There’s also a huge understanding gap in how to use AI tools.

These tools work best when they are given a very clear set of prompts. ­Without those clear steps, you risk ­getting ­nonsense. Rubbish in, rubbish out is a maxim used in computing, and not ­without reason.

Conclusion

After musing on this awfulness, I told ChatGPT the following: “That story is not accurate. Please research Wonky the lamb further and tell me the actual story.”

And it did. It got it right, and it noted that the story of Mairi MacLeod was ­inaccurate – but it only did that because I knew enough about the topic to know that the first version was wrong.

We’ve unleashed this powerful new technology into a world that doesn’t fully understand it, doesn’t know how to use it properly – and often can’t even tell when it’s wrong. We’re seeing rubbish out, ­because we’re putting rubbish in, and ­calling it progress.

AI isn’t thinking. AI isn’t truth. It’s a mirror – and right now, it’s reflecting just how lazy, gullible and underprepared we really are – myself included.