ChatGPT is NOT for Genealogical Research

At the risk of seeming a Luddite, we dare to come out against the use of artificial intelligence tools such as ChatGPT, as they are at the moment, for genealogical, or any other research. We have read the many excited articles and listened to many of the webinars touting the tool for genealogy but we cannot agree. We have made a fair number of tests of the tool as applied to genealogical and historical research and we are underwhelmed. Worse, we are extremely annoyed.

The crucial flaws are that the "tool" lies, fabricates, gives false information, on the one hand and, on the other hand, fails to find a great deal of readily available information. Looking at the latter, the failure to find information that one can readily find with a web browser search simply makes ChatGPT an inferior product.

Looking at the former, the fabrication of information presents a new problem where there was none before. Apparently, the fashionable term for the tool's lying is "hallucination". This is a most dishonourable excuse for falsehood. (We are of a generation that knows rather a lot about hallucinations and we assure you, they did not involve flat out lies.) 

When queried, the self-promotion of the thing is also a master class in vagueness and equivocation. One could say in modern slang that artificial intelligence is gaslighting users.  In the end, it apologizes and whines that it is not perfect. Really! The whole experience is like having a conversation with an extreme narcissist. 

Where does this leave the researcher? Since one cannot trust that all that is sought has been supplied, one has to do the searches again anyway. In having an index done, since something can be missed, one must do the work one's self anyway. It is the same with every task. No saving of time here.

More damaging is the false information. As an example, a new branch of a family we are researching, purportedly in nineteenth century Paris, was suddenly supplied by ChatGPT. One cannot know that it is not true, and so, wastes hours of research time only to discover it is a lie. Not only is time not saved, a great deal of it is wasted. 

To date, ChatGPT cannot transcribe the content of a photographed document, so it is useless with that aspect of genealogical work as well. In the human despair that one would like to attribute to it, the next level of self promotion suggested its use was:

"Diverse Perspectives: ChatGPT can offer different perspectives or ideas that the user might not have considered, sparking new avenues for exploration or creativity.

Efficiency: ChatGPT can quickly generate a large number of ideas or concepts, saving time compared to traditional brainstorming methods.

Inspiration: Even though ChatGPT's responses may not always be accurate, they can still inspire the user by providing starting points for further development or refinement.

Exploration of Unconventional Ideas: ChatGPT's lack of human bias can sometimes lead to the generation of unconventional or unexpected ideas, which can be valuable for thinking outside the box."

 The arrogance of the thing's promoters is breathtaking. "Efficiency" has already been disproved. As for the rest, we have rarely come across duller, more anodyne, facile, tedious, unoriginal and often wrong content than in the responses of ChatGPT. We plan to avoid it henceforward. Unfortunately, many family historians will not; many will use it, trust it and perpetuate its nonsense in their books, articles, family websites, blogs and Wikipedia pages. Dear Readers, we have a new Gustave Anjou.

©2024 Anne Morddel

French Genealogy