AI and Pictures - Will it Enhance your Genealogical Research?

Created by ChatGPT

Happy New Year! Hard to believe this is my last blog of 2023. I've been writing about using Artificial Intelligence in family history research for the past few weeks. I hope it has been helpful and taken the fear out of the tech. Whenever I've talked about it with my personal FAN Club, I get the same reaction which is either a sign of the cross and the question, Haven't you seen the terminator movie? to Oh, I want to learn how to do that. Absolutely no reactions in between!

Which I think is quite interesting because I'm a middle of the roader. I love the time saving aspect and the assistance in handling large amounts of data to analyze but I also am very cognizant that this tech is still in it's infancy, makes mistakes (like people do), and due to lack of transparency from the companies involved, we have no idea what training information was used. So, it could have been an item that was under copywrite, it could have been biased information, or it could have been false information.

That said, I still think it can be useful. I really miss an early Google AI attempt called Picasa. I had saved all of my family photos there and it could identify family members that were infants I could not. I have two sets of twin cousins and in some of the photos I can't tell them apart. Picasa was able to separate them out individually.

Today, AI can also create pictures. I haven't tried to input photos from say, 1890, and ask it to take the person in the photo and create a new picture in 1820 period clothes but I suspect that can be done. Sounds interesting and want to get started?

Through ChatGPT I use Dall-E. It has become one in the same since November. Before that, Dall-E was Beta but now they are joined. Here's how to use AI to get a picture.

1. In the chat box, ask to "Create a picture or photo of" whatever. You can be specific or not.
2. When you get a response you might want to edit the picture. You can have the AI do that by specifically stating what you want altered. In the photo above, I asked to update the first image with the word "2024" added. I never told it where to add but perhaps you wanted it front and center. You can then tell it to add a calendar showing the month of January and the year 2024 displayed.

I will be one of seven professional genealogists giving a lecture at an upcoming genealogical conference later in 2024. I will be using AI to create some of the slide pics. For some slides, I had an idea but for others, I let AI create for me with only the slide title as the instruction.

Ethically, I believe that I need to credit AI for the pictures, just as I would source any photo that I use.

I hope you are enjoying the holidays with family and friends. I look forward to 2024 and providing you with more heartfelt and informative blogs. All the best to you and yours!

A Little Bit of Genealogical Luck

                                                                Photo by Lori Samuelson

As the year closes, I'd like to reflect on some luck I had this past year. I was able to find the school enumeration records for my husband's grandmother by getting a tip after writing a journal article. I hired a genealogist in Croatia who knew someone I could hire as a driver who just happened to have had a Ph.D. in archaeology and just happened to have done her dissertation on the area my ancestors once lived. Another lucky strike was writing to a small library in Ohio to ask if they had a list of Masons from the 1820s and getting a response they did not but they had in their vertical drawer notes from an unknown researcher that provided the source for a lost deed that hadn't been filed until years after it was made and hadn't been included in the index.

Those were all wonderful unexpected finds but I think the best luck came when I went cemetery hunting.

The photo above is just one of many that I took this past summer as I traveled across Ohio and Indiana searching for ancestor's graves, then cleaning and photographing them. The one showing in the right corner is for my husband's second great grandfather and his second wife, Maria Erickson.

Notice where the car is parked? I had no idea where in the cemetery the stone we were looking for was located. I only had a map for one cemetery, Graceland in Valparaiso, Indiana, but the sections weren't marked in the cemetery so the map was useless. Somehow, my ancestor GPS was fully on as except for Graceland, hubbie and I found every grave in record time. I just drove in and something told me to STOP!!!! So we did.

Usually cemetery hunting is a spring-summer-early fall activity but two weeks ago one of my adult kids and I went to Chicago. The weather was frightful - sleeting, windy, and bitterly cold. We had wanted to go to the Field Museum but they had closed the parking lot close by, there was no street parking left and I didn't want to pay for the parking garage down the street so we decided to go to the cemetery.

I realize that is a tad weird to those who aren't interested in family history but this worked for us. I've written about Drusilla Williams DeWolf Thompson before and I've shared the also lucky find of the picture below in an attic in Dayton, Ohio where Dru never once set foot:

Dru is the woman with her head on her hand under the tree. Husband Thomas is impersonating Abe Lincoln. Daughter Mary is to his right, that's my husband's great grandmother.

You would think this stone on a bleak December day would be easy to find but it wasn't. We had a map, too. We could see it was close to the cemetery office so we decided to just park there and inquire where to find it. The sweet office clerk donned his jacket and said it wasn't far and he was right. It was just a few yards from the office. I wouldn't have discovered it, though, without his help as the limestone is now barely readable. The trees are gone, as is the fence. The stone is off kilter as the base has eroded. It's missing the top. It's also filthy:

                                                        Photo by Lori Samuelson Dec 2023

You can see that other stones surrounding it are also gone.

The cemetery does not allow families to clean stones so I've consulted with a company who will go out next spring and take a look.

I had always wanted to visit these folks as their story fascinates me - arriving in Chicago by wagon in the 1840s, surviving the fire, and watching Chicago grow into a metropolis. If only they had left their memoirs!

As a new year peaks around the corner I can't wait for more exciting finds. Hope your holidays are delightful and that you continue to follow my fabulous genealogical adventures next year. I'm planning to take another AI course through the National Genealogical Society and for two more heritage trips. Lady Luck is who I'm hoping to accompany me. Keep your fingers crossed for me!

AI and Your Genealogy Notes

 

AI Image

Previously I blogged about using Artificial Intelligence (AI) to create narratives of your ancestor's lives. Today, I'm going to show you how you can take those online classes you are attending and make concise notes from the chat box.

Why the chat box you may think? Because the chat box often contains important information that participants have thought about the subject to add - new questions to ask the speaker - and resources and links that someone attending found useful.

Here's the how to turn your chat box chatter into useful info:

1. At the end of the session, either highlight the enter chat box conversation and copy (ctr c) to a Word document (ctr v). Save as a pdf.
2. Upload to ChatGPT, Claude, or Bard. (Try more than one to see the results and which you prefer).
3. Provide a prompt - "You are a professional genealogist and expert writer. Please create an executive summary from the information provided. List attendees, key topics discussed, key questions raised, and notable moment.

You can also do this from YouTube videos for the class itself if you don't have time to view it. Here's a how to get a copy of the transcript. Then, follow the three steps noted above. You can tell your AI helper to make a title of the video you watched, add the date, and add the speaker's name, too. Personally, I'd save the results to my Dropbox as I keep syllabi there to refer to in

Can AI Establish Relationship from Indirect Sources Only?

Picture Courtesy of ChatGPT

Like last week, my answer is a strong maybe! I've been researching my John Duer, Patriot, and his relationship to his purported son, Thomas for YEARS! I have so much indirect evidence and one piece that does state Thomas was the son of John but it is in a collection made beginning in 1865 that doesn't provide the source. Of course!

So, I turned to AI to get its take on the problem.

Before I tell you the result, I want to share with you how to take data and create an Excel table using AI. It's so super simple and saves a lot of time. WARNING: Like with all new tech, you much check for accuracy. Don't assume it is 100% correct. Even with the time taken for that, it will be a time saver in the long run.

1. Create a prompt to "extract structured data from..." whatever you uploaded. From the class I took, it was suggested we add "prioritize fidelity to the information below; dry recitation, avoid editorialization. Prioritize sentence-by-sentence from beginning to end. Extract from the (uploaded item) this structured data... (list whatever info you want). Display your response in the form of a CSV file. The response should have the header row: (titles you want - could be) "Person's Names", "Relationships Noted", "Evidence of the Relationship" (or whatever you deem important). Enclose all cells or fields in quotation marks, omit, ignore, and avoid commas and other punctuation in the original text in your response in an attempt to keep the CSV error-free. Display your response in a code window for easy copy-and-paste.
2. Upload your item as a .pdf file.
3. A black box will display with the text in white letters. You can ctr c and ctr v into Notepad (you can find notepad by typing Notepad or Text in your computer search bar). Or, you can click on the upper right screen on ChatGPT and share, then copy code.
4. Open the file in Excel and clean it up, if necessary.
5. Then I opened a new ChatGPT message and gave the following prompt: Today you and I are an expert genealogist and writer. Your goal is to generate true and accurate reports. Find below the contents of a CSV file created from my research on the relationship of John and Thomas Duer. Transform the structured data into a narrative report; dry recitation, no editorialization, please.
6. This is what I received:

John Duer was born on January 19, 1748, in Cranbury, Middlesex, New Jersey. Various sources, including family data collections and genealogical websites, document this event. The death of John's mother, Mary Ann Hollingshead Duer, occurred in 1751 in Newton, Sussex, New Jersey, although specific source documentation is absent beyond public family trees.

In 1773, John married Susannah Miller in Sussex, New Jersey. Their marriage is recorded in the Family Data Collection and referenced in genealogical records from Youngstown and Mahoning County, Ohio. The couple had several children, the first of which was Esther Duer Hopkins, born in Monmouth, New Jersey, in 1774. However, there is a discrepancy in her birth location, as other sources suggest Sussex County, New Jersey, as her birthplace.

John's military service is confused with that of a cousin who served in the Pennsylvania militia. Despite this error, records from Ohio acknowledge John Duer's service, including listings in the official roster of American Revolution soldiers buried in Ohio.

Thomas Duer, John's son, was born in 1775 in Sussex, New Jersey, with multiple sources confirming this, though his gravestone lacks specific dates. Jonathan Duer followed on November 11, 1777, also in Sussex, with his life documented through family data collections and census records.

Joseph Duer's birth in 1779 is only supported by public family trees, lacking concrete source documentation. In 1780, Mary Duer Hazen, another of John's daughters, was born in Sussex. Her lineage is established through the DAR lineage book and a family history published in 1947.

John Duer was a church member at the First Baptist Church of Wantage, Sussex, New Jersey, in 1783, with historical records from the Sussex County Historical Society likely supporting this fact. His military service is further documented in DAR and New Jersey Revolutionary War records, as well as in family genealogical files.

In the following years, John resided in Newton, Sussex, New Jersey, where he is listed as a vestryman in a history of Christ Church, and in Hanover Township, Morris County, New Jersey, as indicated by tax records. He also engaged in land transactions, including purchasing land that would later become part of West Virginia, and in 1797, he was a resident of Mansfield Township, Sussex, New Jersey, acting as a bondsman.

John's life from 1798 to 1829 included various land dealings, military service records, family events such as the deaths of his sister Anna Duer Hull and his son Thomas Duer, and his own death on April 3, 1831, in Jackson Township, Trumbull, Ohio. His will was probated in March 1832, and his estate included guardianship arrangements for his grandchildren.

The narrative concludes with references to Thomas Duer's estate papers, tax records, and land transactions that involved John Duer and his son Thomas. The document also notes a DNA connection to the family through Elizabeth Suriner and references to historical maps and newspaper articles related to the Duer family in Hubbard, Trumbull, Ohio.

Interesting, huh?! Nowhere in the csv data did it say that Thomas was John's son. My headings were Date, Event, Place, Source, Name. Would this be accepted by a lineage society? Definitely not but it was an unbiased analysis of the indirect evidence. I did provide in the prompt to look for a relationship of John and Thomas Duer but did not state it was father-son.

I'd love to hear what you find in using AI to establish relationship.

AI Prompt Chain and My Creepy Genealogy Result

 

Prompt Chaining in Artificial Intelligence (AI) to me just means taking baby steps to get to your desired end product. It's like wanting your three year old to follow four directions but that's too much so you give them three and then add the fourth.

What I wanted to do was use AI to extract (pull out) key parts of newspaper articles about my grandfather Edwin Anton Leininger to do the following:

Create an executive summary of the articles

List the FAN club noted in the articles

Determine the FAN Club relationship to my grandfather

List locations from the articles

Note dates from the articles

Create a narrative based on the relationships and key details from the articles

Think about how long this would take you to do. It took ChatGPT less than 5 minutes.

Was it entirely accurate? Nope, and here's where it got creepy.

My first step was to select six newspaper articles over the time period my grandfather lived and from various locations. Chat GPT can only handle about four newspaper articles at a time so I uploaded those first sets (one from 1918 where he was collecting money for the war effort for YMCA in Ohio, a 1970 article about the historic dairy farm he once ran in Indiana, a 1950 story about a calf who had escaped from the farm, a 1940 article about him buying the farm). I didn't upload them in chronological order. I then asked for a summary of the four articles. What I received was all correct; each article had the heading taken from the article and key points bulleted.

Next I uploaded his obituary and a photo from 1944 with a short heading of his job with the railroads. I asked AI to add the two latest articles to the summary.

But AI has a short memory so what I received was the summary for just the two last articles. I wrote a new prompt asking for a creation of an executive summary for the articles combined.

AI did as I asked so I then requested to extract the names and provide relationships. I also asked to list locations from the articles. Here's where it got weird!

First, there were errors with names. I quickly figured out Mrs. Troncale was misread and should have been Tronolone. AI didn't figure out that my grandfather was listed by different names - Edwin A., E.A., and Ed. AI thought these were all separate individuals. It also placed locations with names; technically this is correct as the name of a dairy is a name and I hadn't given a clear direction for geographic names only.

I informed AI of the mistakes/corrections and received a new list. I then asked AI to substitute first names for women who had "Mrs." only. That was done correctly.

In reviewing the last list three names appeared that were related to my grandfather but I didn't recall those being in the articles. I went back and read all six articles. Nope, none of these names appeared. Where did this come from?

I asked Chat GPT to let me know where the names were found. I received an apology and a name removal.

The creepy part here is that the names were correctly related to my grandfather. One was his sister, another was his mother-in-law, and the third was his grandfather. How in the world did AI happen to select the right first and last names that were related to him?

We know AI is not 100% accurate but to select three people, all with different last names, and tie them to the extraction when they weren't listed in simply weird. So I asked (response above). The response doesn't provide the answer. I don't believe AI is psychic and the information came from somewhere. I suspect that information is in fact being pulled externally and perhaps not now, but had been in training.

The end product, a summary of the six newspaper articles taken from throughout my grandfather's life, was correct. Here it is:

A Legacy Through Time
Edwin Leininger, often referred to by his acquaintances as Ed or E.A. Leininger, led a life characterized by commitment and dedication. Born in Mercer County, Ohio, he later made Fort Wayne his home for nine years. Professionally, Edwin was recognized for his unwavering commitment to the Pennsylvania Railroad, where he diligently worked as a car inspector. Later in life, he became an
indispensable figure at the Hobart post office, where he was affectionately referred to as "the man with the muscles." For years, he efficiently managed both outgoing and incoming mail, handling approximately 9,500 pieces of outgoing mail and 2,000 pieces of incoming mail daily.
On the personal front, Edwin's life was richly woven with familial bonds. In his earlier years, he was married to Lola Leininger. Together, they embraced the joys of parenthood with the birth of their children: a son, Guy Leininger; a daughter, Mrs. Bonita Cook; and another daughter who later became Mrs. Ted Tronolone. Later in life, Edwin's heart found companionship again with Gladys
Leininger. Through this union, Edwin warmly welcomed Gladys' three sons, Duane, Donald, and Capt. Robert Landfair, as his step-sons.
Edwin's family circle extended further with his brothers, Sam and Lew Leininger, and a sister, Bertha, who later became Bertha Pifer. The bond of the Leiningers remained strong, binding them together through the tapestry of time.
In 1976, Edwin's journey on Earth reached its sunset in Fort Wayne. Yet, his legacy—reflected in his professional dedication, familial bonds, and community impact—lives on.