In 1950 the mathematician Alan Turing proposed what came to be called the Turing Test: if a human observer screened from other clues judged its responses as indistinguishable from those of a human then it was reasonable to describe a machine as thinking. During the 1950s development in Artificial Intelligence (AI) continued, with scientists attempting to create machines with human-like intelligence. Roughly speaking, or anyway enough to be plausible, language meant thought, and thought meant intelligence.
In recent years large language AI models have been developed. These are essentially computer routines designed to ‘train’ a computer program with a vast amount of text to recognise regularities. Large-language models are employed in natural language processing tasks, especially language generation. In the 2020s a large language model, Generative Pre-trained Transformer 3 (GPT-3), was released by OpenAI, an artificial intelligence organisation. In November 2022 OpenAI released ChatGPT, a generative artificial intelligence chatbot, ‘generative’ in the sense that it could generate sentences and in other ways respond to natural language inputs. It was a robot that could chat to humans.
ChatGPT and similar tools have been widely adopted by genealogists to assist in writing family stories, illustrate those stories, asking and answering questions about people’s lives.. There have been problems with these tools, which sometimes generate false, biased, or incorrect content, called ‘hallucinating‘. The tools are becoming more accurate, however, though users of course are encouraged to check the facts and review sources.
Despite the improvements I myself am not interested in having artificial intelligence tools simply go over the facts and generate biographies or other posts for me, for a machine re-hash of this is no substitute for a comprehensive checking of sources followed by a thoughtful hand-written biography or story. When compiling a post I find the research process interesting and I enjoy pulling together the result.
There are other applications of AI including web search software, for example Google Search; recommendation systems such as those used to generate playlists; virtual assistants, for example Siri and Alexa. However, many of these applications are no longer thought of as AI. It has been suggested, with a touch of cynicism, that once an application of this kind becomes sufficiently useful and common it is no longer labeled AI.
Databases such as ancestry.com and FamilySearch.org seem to be growing more sophisticated. Whether this is the use of “Artificial Intelligence” or just the use of increased computing power and improved techniques and algorithms is not clear to me. I suspect I am guilty of no longer thinking of this increasing power as a use of artificial intelligence.
For the last year or so I have found the full text search at FamilySearch very useful for searching records such as those held by the Registry of Deeds in Dublin.
I have also found recent improvements in machine translation and document transcription very helpful. The routines are never perfectly accurate but the results are always a good starting point and save time.
I have tried transcribing handwriting using Google Docs, Transkribus and ChatGPT. ChatGPT has a tendency to hallucinate, that is, to produced false and misleading results, even though the tool was asked to transcribe the document word for word. I have also used Transkribus and Google Docs to transcribe handwriting. Neither of these tools appear to fabricate spurious results.

I use Google Translate often. It is not perfect, and from time to time I have asked native speakers to help me make sense of the results. A speaker of Welsh helped me in this way with a Welsh obituary, and a German speaker helped me with the death certificate of my great great German grandmother Wilhelmine Henrietta Bertz nee Ritter.
All these tools help make records more accessible to family historians, easier to find, and more readily useful. Has this improved the way family history is researched and written? Here I can say without hesitation, ‘Maybe’.
Related posts
Don’t trust chatbots : my first experiments with ChatGPT in 2023
Using Transkribus to decipher the death certificate of Gustav Grust 1839-1901
An obituary about John Hughes (1788–1844) : looking at a Welsh obituary
W is for Wilhelmine : includes a German death certificate
Snell intel. : exploring full text search on FamilySearch
This post first published at https://anneyoungau.wordpress.com/2025/06/29/progress-using-artificial-intelligence/
I similarly tend to take a similarly conservative approach to AI.