With all the hype around ChatGPT, I wondered what would happen if I tried to get it to rewrite some of the posts on this blog.
The results were… interesting.
Let’s start out with a post which was intended to be mainly factual, and less opinionated. I wrote this in 2017 as a love letter to this markup language. In the post I’d wanted to get across what it was for someone who’d never heard of it, why it was cool and a little bit of the history behind it.
For ChatGPT, I gave it the prompt:
Write me a post about Markdown, extolling its virtues.
And here’s what it came back with:
“Markdown is a lightweight markup language that allows users to write and format text in a simple and easy-to-read way. It’s a popular choice for writing online content, such as blog posts, documentation, and README files, because it is easy to learn, quick to use, and produces clean, structured text that can be easily converted to HTML or other formats.
“One of the main virtues of Markdown is its simplicity. Unlike more complex markup languages like HTML, Markdown uses a small set of simple symbols and characters to format text, such as # for headings, * for bullet points, and > for blockquotes. This makes it easy to learn and use, even for those without any prior coding experience.
“Another virtue of Markdown is its flexibility. Because it is a plain text format, it can be used with a wide variety of tools and applications, including text editors, word processors, and even email clients. This allows users to write and format text in a way that works best for them, without being locked into a specific tool or workflow.
“Markdown also allows for easy conversion to other formats, such as HTML, PDF, or even Microsoft Word documents. This makes it a great choice for writing online content that needs to be repurposed for different platforms or audiences.
“Overall, Markdown is a powerful and versatile tool for writing and formatting text that is easy to learn and use, flexible, and easily convertible to other formats. It’s a great choice for anyone looking to write online content in a simple, structured, and maintainable way.”
So, this one is actually pretty good. It succeeds in explaining what it is and why you might want to use it. There aren’t examples of syntax and output, nor does it cover the interesting history of this language, but then again I never asked for this in the prompt.
This is a post from 2018, and I remember taking ages to get this one out the door but being quite pleased with the result.
I was becoming frustrated with overcrowding on Montreal’s Orange Line and the lack of transparent transit planning in the city. Retrospectively, it’s quite an opinionated article but one based on personal experience and I tried to do a good job researching around the topic and wanted to offer some ideas as to how things could be better.
The prompt I gave GhatGPT was:
Write a post about overcrowding on the Montreal Metro’s Orange Line and its impacts and possible mitigations in the future.
And here’s what it came back wtih:
“The Montreal Metro’s Orange Line has been experiencing severe overcrowding in recent years, causing delays and discomfort for commuters. This is primarily due to population growth and increased demand for public transportation in the city.
“The impacts of overcrowding on the Orange Line are significant. Commuters are often forced to stand during their entire journey, leading to discomfort and fatigue. Delays are common, causing frustration and making it difficult for people to plan their days. Additionally, overcrowding can also be a safety concern, as it increases the risk of accidents and injuries.
“To address the problem of overcrowding on the Orange Line, several potential solutions have been proposed. One solution is to increase the frequency of trains, which would reduce the number of people on each train and decrease waiting times. Another option is to add more cars to trains, which would increase capacity and reduce the number of people standing during their journey.
“Another possible solution is to improve the integration of the metro system with other modes of transportation, such as buses and bicycles, to encourage people to use alternative methods of commuting. Additionally, encouraging more people to work from home or at flexible hours could also help to reduce peak hour crowding.
“Finally, the long-term solution could be the extension of the Orange line to other areas in the city to increase access to the metro and reduce the pressure on the current line.
“Overall, overcrowding on the Montreal Metro’s Orange Line is a major concern that needs to be addressed. While there is no one solution that will solve the problem completely, a combination of measures, such as increasing train frequency and capacity, improving integration with other modes of transportation and extending the line, can help to mitigate the impacts of overcrowding and improve the experience for commuters.”
OK, so this one didn’t turn out so well. It’s true the Orange Line was overcrowded before the pandemic, but this isn’t the case anymore since lots of people are opting to work from home. So, it took the assumption I gave it in the prompt at face value and not the current state of affairs.
It speaks in general terms about overcrowding in a way which could apply to any subway system. Suggestions to resolve the problem such as adding more cars make little sense for this specific line, as trains already fill the platforms. The idea of integrating the transit system is on-point though, and something most transit experts would probably advocate for.
However, the suggestion that extending the line into other areas of the city to alleviate overcrowding is nonsensical. In fact, the Orange Line was extended and this was one of the things which made overcrowding worse.
The conclusion that a combination of solutions is the best approach seems to make sense, but again is very wishy-washy and not specific to the topic in question. Transit writers, I think your jobs are safe!
Back in 2020 I visited London, taking the circuitous route flying from Montreal to Paris and then taking the train to London. The entire content of this post was relaying a uniquely personal experience, and any other person who’d made the same trip would have told you a different story.
So I though it would be interesting to see what ChatGPT would do if I asked it to write about an experience as if it were its own. I prompted with:
Write a post about the experience of flying from Montreal to Paris, and then taking the Eurostar from Paris to London.
The story it wove came back as:
“Flying from Montreal to Paris was an exciting experience. The flight was smooth and the views of the Atlantic Ocean were stunning. The service on the plane was also great, with friendly and attentive flight attendants.
“Upon arriving in Paris, I took the Eurostar train from Gare du Nord station to London. The train ride was comfortable and efficient, with great views of the French and English countryside. The Eurostar is a high-speed train that travels through the Channel Tunnel, connecting France and England. The journey takes around 2 hours and 15 minutes and is a great alternative to flying.
“Overall, the experience of flying from Montreal to Paris and then taking the Eurostar to London was enjoyable. The flight and train ride were both comfortable and the views were spectacular. I highly recommend taking the Eurostar if you are looking for a fast and convenient way to travel between Paris and London.”
I don’t think it was reasonable to expect an AI bot to write about a personal experience, and we can see it’s looking at the two legs of the journey as completely separate experiences without any unity in the narrative. It’s interesting as well to see it assume such a trip would be a positive experience.
Back in 2020 the pandemic had just started, offices had closed and overnight everyone was suddenly an expert in how to work from home, including me.
I asked ChatGDP:
Write me a post about working from home during the COVID 19 pandemic, with a focus on what office and desk equipment is needed as well as how to get the most out of the experience.
And here’s what I got back:
“Working from home during the COVID-19 pandemic has become the new norm for many people. To make the most out of this experience, it is important to have the right equipment and set up a comfortable and productive workspace.
“First and foremost, a reliable computer and internet connection are essential for working from home. A laptop or desktop computer with a fast processor and enough storage will allow you to run multiple programs and handle large files. A good internet connection will ensure that you can participate in video meetings and access files from the cloud without interruption.
“Next, you will want to set up a comfortable and ergonomic workspace. A comfortable chair and a desk that is the right height for you will help you to work for longer periods of time without experiencing discomfort. An adjustable lamp or natural light source will help to prevent eye strain.
“In addition, a good set of noise-cancelling headphones or earbuds will be helpful when you need to focus on a task or take a call. To help stay organized, a whiteboard or calendar can be a great way to keep track of deadlines and appointments, and a good set of pens or pencils will be necessary for taking notes.
“To get the most out of working from home, it’s also important to set boundaries and establish a routine. Create a designated workspace and make a schedule that includes breaks and time for exercise or other activities. It can also be helpful to schedule regular check-ins with colleagues to maintain a sense of connection and collaboration.
“In summary, to make the most out of working from home during the COVID-19 pandemic, it’s important to have the right equipment, set up a comfortable and ergonomic workspace, establish boundaries and establish a routine. With the right tools and mindset, you can make the most out of the experience and be productive and effective in your work.”
So, this is pretty good. It covers all the basic points on working from home and I like it also talks about the health and mental health aspects such as keeping a routine and getting exercise.
There’s so much which has been written about this topic now, I wonder if it had a good data set to draw on which resulted in higher quality than some of the other prompts produced.
Another transit post, this one from back in 2015 when a new rail line had opened in Greater Montreal. After riding it, I had some pretty strong opinions.
Here’s the prompt:
Write a post about the “Train de l-Est” in Quebec.
The first time I attempted this I left out the fact it was in Quebec, and it started writing about a rail line in France. After adding in some more context, it returned with:
“The “Train de l’Est” is a passenger train service that operates in the province of Quebec, Canada. It runs from Montreal to Quebec City and stops at various stations along the way, including Saint-Jérôme, Sainte-Thérèse, and Sainte-Foy.
“The train is operated by Exo, a subsidiary of the Agence métropolitaine de transport, and provides a convenient and comfortable way to travel between Montreal and Quebec City. It offers several classes of service, including economy, business, and first class, and features amenities such as Wi-Fi and power outlets.
“The Train de l’Est also offers a number of packages and discounts, including group rates, student discounts, and special offers for seniors and children. In addition, the train is fully accessible for passengers with disabilities.
“The train service is a great option for those looking to explore the beautiful region of Quebec, as it offers scenic views of the countryside and easy access to popular tourist destinations. It also provides an alternative to driving or flying, with the added benefit of being more environmentally friendly.
“Overall, the Train de l’Est is a valuable transportation option for those living in or visiting Quebec, offering a comfortable and convenient way to travel between Montreal and Quebec City, while also providing access to the beauty of the Quebec region.”
This one is, to put it lightly, misinformation. It’s pretty much totally factually incorrect. It claims the line runs to Quebec City and the intermediate stations, none of which are on this line and some of which don’t even connect. It doesn’t have multiple classes and doesn’t offer discounts. The last two paragraphs are full of fluff which you could say about nearly any rail line.
My guess is it wasn’t able to tie “Train de l-Est” to the actual route in question (Montreal to Mascouche) and just made assumptions based on the fact it knew the line was in Quebec.
From 2015, I wrote this post because I had a connection with this imposing building from my childhood, and was also fascinated by how Christopher Nolan decided to CGI it into his original vision of Gotham City. It blew my mind when I saw the film for the first time.
I asked ChatGPT:
Write a post about the National Institute for Medical Research in the UK and how the building was represented as Arkham Asylum in Christopher Nolan’s film: Batman Begins.
And it came back with:
“The National Institute for Medical Research (NIMR) in London, UK, is a renowned medical research institute that has made significant contributions to the field of biomedical research. The institute, which was established in 1911, is known for its cutting-edge research in areas such as cancer, genetics, and infectious diseases.
“In Christopher Nolan’s 2005 film “Batman Begins,” the NIMR building was used to represent the fictional Arkham Asylum, a psychiatric hospital where many of Batman’s enemies are incarcerated. The building’s imposing architecture and historic features lent itself perfectly to the dark, brooding atmosphere of the asylum. The building’s unique design provides a perfect backdrop for the film’s portrayal of the asylum as a place of confinement, isolation, and madness.
“The use of the NIMR building as Arkham Asylum in the film showcases the versatility of the building and its ability to provide a perfect setting for different types of productions. It also serves as a reminder of the important work that is carried out at the institute and the contributions that it has made to the field of medicine.
“Overall, the National Institute for Medical Research played an important role in Christopher Nolan’s Batman Begins, adding to the film’s atmosphere and realism. It’s a representation of how a real-world location can be used in the film industry to make a story more authentic and believable.”
This isn’t bad, but it misses one crucial bit of new information – which to be fair is also missing from my own post – and that is this building no longer exists, sadly having been demolished from skylines of the valleys around Totteridge and Mill Hill.
Again, I notice it treats the prompt as two separate topics, as if the paragraphs from two posts have been interlaced and a flimsy conclusion added to the end.
I am impressed with ChatCPT. If given the right prompt it can give you some coherent prose. But, stray into more niche topics and it starts to bulls***. It’s also naïve in performing no fact checking on the prompts it receives, although I understand its knowledge base is purposely locked-in from a few years back, so it can have no concept of recent events. But still, garbage in; garbage out, and it’s easy to see how it could be abused by bad actors.
In terms of the writing style, it’s pretty consistent and I think a practised human eye would probably be able to spot texts which have been generated by this AI. There’s just something about the way it writes which is slightly odd, although I can’t quite put my finger on why I find it a bit weird.
I think a year or two from now, we’ll see massive improvements, and I’m not sure if that’s a good thing or not.
As a last experiment, I prompted:
Write a post about how I used ChatGPT to rewrite older posts on my blog.
And back came:
“I recently used ChatGPT, a powerful language model developed by OpenAI, to rewrite some of the older posts on my blog. The process was surprisingly simple and efficient.
“First, I gathered a list of all the posts that I wanted to update. Then, I used the “prompt” function in the OpenAI API to input the text from each post, one at a time. I asked ChatGPT to “rewrite this post in a more modern and engaging style,” and the model generated a new version of the text in just a few seconds.
“I was impressed with the quality of the rewrite. ChatGPT was able to maintain the main points and message of the original post, while adding new insights and making the language more polished. I made a few small edits to the generated text, but overall, it required very little manual work on my part.
“Using ChatGPT to update my old blog posts has been a great way to refresh the content and make it more relevant for today’s readers. I highly recommend this approach to anyone looking to update their own blog or website.”
Who knew a chatbot could have such a self-inflated opinion of itself?
The version of ChatGPT used was January 9th (2023). If you want to try it for yourself, visit the ChatGPT Website and register for an OpenAI account.