AI and ChatGPT
The news has been awash lately with talk about ChatGPT, the artificial intelligence chatbot developed by OpenAI and launched at the end November 2022. A post recently loaded on a local Residents’ Association site, immediately made me wonder if this was my first exposure to ChatGPT’s authorship. It was littered with ornate language and flowery done-to-death phrases. Shortly afterwards 2Summers posted “What ChatGPT has to say about 2Summers”. She had asked ChatGPT to write a piece in her style of writing.
Using LizatLancaster’s voice
I had just written a post on Siemon Allen’s installation Stamp V 2010 which forms part of the exhibition Otherscapes at JCAF. Out of interest, I asked ChatGPT (very politely as it is important to feed into AI’s algorithms of courtesy and positive language), to write an article on Siemon Allen’s work in the language and style of Lizatlancaster.co.za/blog. Instantly the bot began typing a reply – lots of vague, fluffy “stuff” and trite phrases emptied of any substance. I sent 2 further refining requests, again using my “pleases” and “thank-you’s”. And on receiving the third attempt, like 2Summers, I was reassured that there is still a space for real live bloggers/authors/ journos, even amateur writers like me.
Factual Errors
Each of the three ChatGPT attempts had factual errors: JCAF [Johannesburg Contemporary Art Gallery] was identified as the Johannesburg Art Gallery; the exhibition was identified as Siemon Allen’s exhibition rather than a group show; ChatGPT writes of 1000 stamps in the installation rather than 23,000; and that each stamp is “carefully chosen to tell a unique story” (which if you read my blog you will see is not the case at all).
Regressing to cliches
The over-the-top florid language bears, I hope, little resemblance to mine … “Get ready to unravel the fascinating facts and profound messages behind this captivating installation”! And the installation as a “mesmerizing visual tapestry”.
Cliched phrases abound: each stamp is “like a small piece of this intricate puzzle, carefully chosen to paint a broader picture of a nation on its path towards progress”. And “it’s not just a retelling of historical events; it’s a celebration of resilience, unity, and the unyielding spirit of the South African people” – lots of feel-good Rainbow Nation stuff suggesting a collective vision and effort, rather than a top-down governmental imaging of an official national identity.
Missing the point: the difference between “Show and Tell” and “Hide and Seek.”
In my 3rd “refining” request, I asked if ChatGPT could take more cognizance of the political and social contexts in South Africa at the time the stamps were issued. AI did take my request into account with:
As you explore the installation, you’ll notice that the subject matter of the stamps themselves tells a compelling story about the political agendas of each era. The choice of images and motifs on the stamps can be seen as a reflection of the prevailing ideologies and narratives propagated by those in power.
But ChatGPT took the heavy-handed and erroneous approach that the stamps show the inequalities and injustices of colonialism and apartheid rather than mask them: The stamps from the apartheid era are particularly poignant. They bear witness to the systematic discrimination and oppression faced by various communities in South Africa. If this was the case it would be the equivalent of a Facebook page full of fighting families, unhappy brides, disastrous holidays and birthday dinners for one.
To suggest that: “Through these stamps, Siemon Allen prompts us to confront the painful legacy of apartheid and the struggle for justice and equality”, is to completely miss the point! As is the contention that Siemon Allen’s “message is clear: South Africa’s history is one of resilience, transformation, and a continuous quest for justice”. The first 90 years of the historical sweep in Stamps V points to the complete opposite!
Dear Reader
Not only would I never write “Dear Reader” unless ironically and I certainly would not urge you Dear Reader “as you walk through the gallery, [to] immerse yourself in the colors, shapes, and stories of “Stamp V 2010. Let it be a reminder that art can be a vehicle for understanding, healing, and coming to terms with the complexities of history.”
For the moment I am secure than I can offer something that is more accurate, more nuanced and more critically informed than what ChatGPT is able to gather together.
Plagiarism in the 21st century
As a postscript… in the many years ago when I was in academic life, dealing with plagiarism was an ongoing battle. To now have to deal with entire essays written by ChatGPT and filled with inaccuracies and purple passages emptied of any meaningful content, would be too much to bear. But maybe the flip side is even worse – as AI becomes more sophisticated and can be steered towards honed language-use and critical thinking, what does this mean for the training of these skill in future graduates?