Photo: AI Image Generation from text prompt: “Lasana Sekou writer fist raised portrait.” (deepai.org, 11.21.23)
By Lasana M. Sekou
In early September 2023, ChatGpt August 3 Version at chat.openai.com was asked by Offshore Editing Services (OES) to identify writers published at House of Nehesi Publishers (HNP), an indie press in St. Martin, Caribbean.
When asked about itself, the artificial intelligence (AI)-powered/generative chatbot said, “You can use it to ask questions, get information, seek advice, or engage in natural language conversation on a wide range of topics.”
For the information search, 22 writers published at HNP were selected: world famous authors; first-time authors; and three upcoming authors, two of which had also been published by other houses. Only the writer’s first and last names were inputted to generate the information from ChatGpt.
Results
ChatGpt had no information for 13 authors and the first sentence of its response for each was: “I’m sorry, but I don’t have specific information about an individual named (Name of writer) in my knowledge base, which goes up until September 2021.”
The information provided by ChatGpt about four HNP authors was generally correct. The answers for five writers ranged from correct, incorrect, to clueless: including wrong birthplace or date; attributing names of books by the author that were not written by the author; identifying fictitious names of books by the writer, and unable to name any book by the writer.
Whether “hailed as the first glimmers on the horizon of artificial ‘general’ intelligence” (Noam Chomsky et al.) or railed against as the certain threat to content creators and researchers, even thinkers, ChatGpt, and like “advanced” machine learning “models” (e.g. Google’s Bard, Microsoft’s Sydney), is “pre-trained” and draws chiefly from swaths of online or digital data to answer questions put to it by anyone searching and asking for information online.
Since the September search by OES, the Default (GPT-3.5) was upgraded to the ChatGPT September 25 Version. On October 4, the day a version of this article was posted at an HNP Facebook page, a sampling of five of the 13 authors with no information in early September was “re”-searched before the social media posting.
Only one of the previously non-identified authors in the sampling was found to have information available in October. According to ChatGPT, she had “passed away on June 6, 1996.” As of this writing, our dear writer is alive and well.
The openai.com maintains the same disclaimer of sorts: “Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts.”
OpenAI arguably warns that, “ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers,” according to a BBC article of December 7, 2022.
Libraries closed and closing? Imagine primary school children and high schoolers navigating AI-powered/generative chatbots from their computer and other digital devices. Searching for information for their homework assignments about writers, artists, and other aspect of culture … especially if they belong to cultures historically subject to erasure attempts, marginalization, and what author Toni Morrison has called “Oppressive language.”
A Bit Beyond Looking Up Writers
Generated of late from the multibillion-dollar AI industry, “TV scripts, school essays and resumes are written by bots that sound a lot like a human.
“Artificial intelligence is changing our lives – from education and politics to art and healthcare. The AI industry continues to develop at rapid pace,” said an article at NPR.org on May 25, 2023.
But “today our supposedly revolutionary advancements in artificial intelligence are indeed cause for both concern and optimism,” opined Noam Chomsky et al. in The New York Times of March 8, 2023.
No need to fear AI bots and models. It will take human intelligence to rise to the “ites” and to raise from the depths what will best [be]come of artificial intelligence.
By the way, ChatGPT got “guavaberry” pretty good. Its “Great Salt Pond” information did not mention Great Bay, Philipsburg, or St. Martin [regardless of the spelling of the island’s name]. As for “Reparations,” the AI gets an “A” for the information generated at ChatGPT (on the last date checked).
