RLP 318: AI Photo Analysis with Mark Thompson

Nicole (0s):
This is Research Like a Pro episode 318 AI Photo Analysis with Mark Thompson. Welcome to Research Like a Pro a Genealogy Podcast about taking your research to the next level. Hosted by Nicole Dyer and Diana Elder accredited genealogist professional. Diana and Nicole are the mother-daughter team at FamilyLocket.com and the authors of Research Like a Pro A Genealogist Guide. With Robin Wirthlin they also co-authored the companion volume, Research Like a Pro with DNA. Join Diana and Nicole as they discuss how to stay organized, make progress in their research and solve difficult cases. Let’s go.

Nicole (40s):
Today’s episode is sponsored by Newspapers.com. Well hi to everyone Welcome to Research like a Pro

Diana (48s):
Hi. Nicole. How are you doing today?

Nicole (51s):
So good. I’m just so pumped about our guest and that we get to talk about AI. How, about you?

Diana (56s):
Oh, me too. Well, I’ve been doing a project using AI, but you know when you use AI you still do traditional genealogy. And my project involves Isabella Weatherford, who’s the daughter of Clemsy Cline. We’ve talked a lot about Clemsy lately ’cause I’ve been doing her family. But there was this dark hole in their history. I had her on the 1860 census with Isabella as a 2-year-old. And then on the 1880 census, I mean that’s a 20 year gap. I could not find them in 1870 for anything. So I decided to really double down and do all of my tools and tricks for trying to find them on the 1870 census. So I went to Clemsy and I looked at where her older children were living.

Diana (1m 42s):
She had some older girls and they were married and living in Benton County, Missouri. So 1860, the family’s all out in Texas, Dallas County, Texas. I’m like, oh, the two girls moved back to Missouri, but they wouldn’t have been that old to move on their own. And they’re both in the same county. I wonder if Clemsy and Henderson Weatherford went back to Missouri. They’d been there in 1850 and Isabella was born there in 1858. So I started doing a really good search. I looked all the Weatherfords, no Weatherfords. But then I started looking for all the Clemsys. There’s the gazillion Clemsys, in case you wondered, that’s not a common name now, but it was very common in the south in 1870. Wow. And so many Isabel’s and Isabella’s.

Diana (2m 24s):
So that wasn’t getting me anywhere. So then I tried just Samuel ’cause he was only eight. I tried all the Samuels in I think Benton County where the girls were in surrounding counties. And I found this Samuel H Wellaford, which of course drew my attention. And the mother’s name was something like Classy. And I’m like, oh, I’ve gotta go look at this. But there was no husband. So I looked at the census, it was definitely Clemsy, but she was written down as Clamsy and the surname was Wellaford, you know somehow they just had the surname of Wellaford. And then here’s the kicker. We have Samuel H, which is exactly right. And the 12-year-old girl that is Isabella was not named Isabella, she was named Talitha E.

Diana (3m 9s):
And so I was really puzzled until I talked to you and you had the idea that maybe it was Delitha Talitha because in her later years, Isabel always went by Isabelle D Royston, her married name, she always had the D. So anyway, our current hypothesis is that’s Isabelle, that was her middle name. It was Talitha/Delitha. She didn’t write very well. I have some of her letters from a pension record and so I could see her spelling was not great. And so that makes sense. Plus we found on the web that there were lots of southern women that people have named them Delitha slash Talitha with both the d and t.

Diana (3m 50s):
So interesting.

Nicole (3m 51s):
Yeah, I just wonder. And the Wellaford versus Weatherford, it just seems like that Enumerator didn’t hear the name correctly or when he transcribed the copy that he would send over to the census bureau, he just couldn’t read his handwriting.

Diana (4m 9s):
Right, right. And I did do a search for Wellafords and that was not a surname that was used in that area. Didn’t see any other Wellafords. So You know it was pretty evident. And here is the other thing that really cemented this in my mind, down about five households is Clemsy’s sister Mahala Shockley, which I didn’t even know was her sister until I did this DNA project, I did this whole Cline project. So my hypothesis is that her husband died soon after the 1860 census and of course throughout in Texas we’re talking civil war era and reconstruction. And I am just wondering if, who knows when her husband died, but I have him in the 1860 tax list in Dallas County and then he’s gone from the tax list.

Diana (4m 59s):
So I have a feeling he maybe died about the time Samuel was born. So Samuel has his name Henderson as a middle name. And we know from the other side of the family that when the husband died and the woman was pregnant with the child, she named her Henderson. Well that was a female So. it kind of makes me wonder if he died while Samuel was, you know while she was still pregnant with Samuel. And so he got the name that’s that’s a current hypothesis. So,

Nicole (5m 24s):
So good.

Diana (5m 26s):
So Clemsy moves back to be near her sister. I think in the 1860s, sometime maybe after the war ended and then her girls get married there. But now I’m trying to think of, okay, now they’ll go back to Texas because they end up in Texas. You know 1874, that’s where Isabelle’s married back in Dallas County. So they were doing a little bit of moving this widow and her children.

Nicole (5m 48s):
It makes sense. The time period is fraught with all kinds of issues. And of course people are gonna move around based on what’s happening in the Civil War. So

Diana (5m 58s):
Right. And we know when we talked about Brooks Blevins book about the Ozarks and reconstruction in the Civil War that Missouri, it was guerrilla war warfare. They had both the Union and the Confederates just fighting. It was such an awful fight. It was just, you know, house to house and your friend and neighbor against each other so, so much. So perhaps they went out to Texas and then when that was all done, they moved back. I know Texas was tough in reconstruction as well. So yeah, just trying to find someplace that was a little bit better maybe. Or

Nicole (6m 31s):
Just to be by family. Well good find.

Diana (6m 35s):
Yes. It was very exciting. I had to call and tell you right away, found Clemsy. You know when we have those people missing in the census, it drives us crazy. We just wanna know where they are. Exactly. So anyway, well that was fun. I’m excited to write up that research and add another piece to her story. Well let’s do the announcements. We have our Research Like a Pro Webinar Series continuing in August, August 20th, 2024. This is a Tuesday at 11:00 AM Mountain time and our presenter will be Steve Little, our friend, and we are so excited to hear him talk about his case, which is titled, “Who’s Eli’s Daddy, A Civil War Era Open Secret- A DNA case study”.

Diana (7m 19s):
And the description reads, James Eli “Bawly” Bower was born in 1863 in Ashe County, North Carolina, during the Civil War.

Diana (8m 3s):
Oral family history suggests that Bawly’s father was a Confederate soldier who, while on leave in 1862, allegedly returned home not to his wife and children, but another woman, Margaret Riley Bower. Nine months later, Bawly Bower was born, and shortly after the solider, William McMillan, was dead. This case study aims to determine if documentary evidence and DNA analysis (both autosomal and Y-DNA) can confirm or refute the family legend that William McMillan (1830-1865) is the father of James Eli “Bawly” Bower (1863-1960), born to Margaret Riley Bower (1840-1915). So the topics will be North Carolina, Civil War, non-paternity event, oral family history, YDNA testing, autosomal, DNA analysis, pedigree collapse, Endogamy, multiple relationships, Visualization Techniques in Genetic Genealogy. Wow, that’s going to be amazing And I am so looking forward to that. Our next Research Like a Pro study group is beginning soon on August 28th, 2024. And I am excited to return to my research. It’ll be fun to think of an objective for that.

Diana (8m 44s):
Our peer group leader application is on our website. So whether for this study group or a future one always consider being a peer group leader for us. It’s a great way to get to know people and to get that complimentary registration. And then as always, join our newsletter for any coupons or news about what we are doing

Nicole (9m 1s):
Well today we have a guest, Mark Thompson. Hi Mark.

Mark Thompson (9m 5s):
Hi. So happy to be here. Nicole and Diana,

Nicole (9m 8s):
We’re glad you could come. Mark has been with us for a while as one of our peer group leaders in Research Like a Pro. He was with us in Research Like a Pro with DNA number four and a peer group leader in numbers five and six. And Mark and Steve Little have started a podcast, the Family History AI Show and we highly recommend that you listen to that and get all the news and updates about AI and genealogy. And it’s a very fun podcast already. Well let me tell you a little bit more about Mark. Mark is a professional genealogist who specializes in digitizing and archiving family memorabilia and genetic genealogy. Prior to becoming a Genealogist, Mark served in leadership roles in information technology.

Nicole (9m 53s):
He worked in several industries to introduce technologies that improved the way those companies worked. Mark leverages his combined expertise in genealogy and technology to guide Genealogists through the rapidly changing field of artificial intelligence, showing them how it can help them with their research. Mark is the president of the Victoria Genealogical Society in Victoria, BC and is the chair of their DNA special interest group. He is also a member of the Association of Professional Genealogists. Mark has decades of teaching and public speaking experience on a variety of genealogy related topics and you’ll definitely notice that because he is so well spoken and we’re so excited to talk today.

Mark Thompson (10m 32s):
Well Thank you very much, it has been, it has been a pleasure to get started with with podcasting. I think the Research Like a Pro podcast I have listened to more episodes of than any podcast in history. We, it’s definitely been a lot of fun for Steve and I to learn from you two and be sort of your, your biggest cheerleaders out there in the community and to, and to get a chance to start a podcast in the genealogy world. It’s been a lot of fun and we love talking about artificial intelligence but we just love genealogy.

Nicole (10m 60s):
So true. You guys do love it and it’s evident in all that you do.

Diana (11m 3s):
It’s so fun. And I am absolutely thrilled that you started a podcast because I love listening to podcasts and You know we just don’t always have time to do all the research on the latest and greatest in AI. And I really appreciate the research that you and Steve do and then share it with us. It’s, it is a beautiful thing. So Thank you for doing that. I know it’s time consuming and it’s not an easy thing to do. So I think on behalf of the genealogy community we Thank you for that.

Mark Thompson (11m 30s):
Well thank you and I, I have learned it’s a lot more work than I thought it was going to be, but it’s a lot of fun. Just the regular tempo of getting ready for a podcast really drives you to stay ahead, to keep up to date, to try out new things and really keeps us on our toes even though we love to do this stuff every day anyway, knowing that there’s a deadline coming next week really forces us to stay on our toes.

Diana (11m 53s):
Yes, we know that feeling So

Mark Thompson (11m 55s):
It iss. I’m sure

Diana (11m 56s):
It’s good. And it’s so great to have two of you You know Nicole and I love talking together and I know you and Steve do as well so it’s fun.

Mark Thompson (12m 4s):
Yeah, it makes it a lot easier. Makes it a lot easier to do with somebody that you like hanging out with.

Diana (12m 8s):
Agreed. Well let’s dive into our topic. I am really excited to talk to you about this whole idea of family photos because we get a lot of questions from people about how do you organize your photos and I think we’re going to talk a little bit about how AI can help us with that. So I’m excited to to just see what you have to say about managing your family photo archives. So can you give us an overview of how you do that?

Mark Thompson (12m 36s):
Sure. The one thing I’ve learned with artificial intelligence, I think like with most technology, it, it starts with what you’re going to do with it, what what you’re already doing before we ever get to the technology part. And so I, I approach all of my artificial intelligence related work with how would I do this manually and then I look for the places where AI can help me. So when I do photo, when I, I do a lot of, I am the family archivist for, for my own family of course I have painfully large number of photos in my collection. Tens of thousands of photos And it. Just organizing themselves over the last 20 years has, has helped me learn a lot of the tools and the techniques. But then of course you learn more as you go along and you You know you read some books and you get in touch with the professionals like you guys and you learn how to do this at scale and when you get into a large number of anything that you have to do, that’s where AI can be really helpful.

Mark Thompson (13m 30s):
AI is so good at looking at large numbers of anything and giving you feedback quickly. You, you won’t get perfect feedback but it can really help you work through things at scale. That’s actually one of the things that it’s really good at. But having a good understanding of how you do it manually is the best way to approach it. So I have a pretty simple process that I’ve worked on over the years with how I manage my photos the manual way, which really isn’t that different than an archivist would’ve done it a hundred years ago with photos, you know, figure out what’s important to you and set your goals. ‘Cause a lot of people have different goals with their photos or their family archives in general. You know you might wanna archive and preserve your photos and make sure that people can look at them. I might want to research who is in them.

Mark Thompson (14m 12s):
Somebody else might want to figure out everything they can about them and update their family tree. And so everybody might actually have a different purpose for why they want to do what they do with their family archive and that actually can, that changes what it is you’re gonna do with them. I’ve gone back and forth on this a few times. I’ve had a few debates with folks about whether or not it’s important to preserve the physical photos first or the physical letters first or to digitize them first. I, I’m now firmly in the digitized first camp. I find that’s the, the value of having all of my photos and all of my letters digitized and then researching them.

Mark Thompson (14m 52s):
I find that digitization step happening as early as possible really makes things easier. And it actually, it’s even made things more easy to do artificial intelligence with them because the earlier that I’ve got a digital copy in how I do my work, the easier it is to actually do additional stuff with it with AI. So I digitize very early and I kind of get myself organized and I primarily do research off of the digital copies. Like now with big monitors and you know good software, I find that I will rarely go back to the physical copy of anything, whether it’s a photo or a letter. I will primarily, I will actually research off of the photo or off of the letter because it’s so much easier to blow it up.

Mark Thompson (15m 37s):
You know to mess around with the contrast or the colors so I can see things that I couldn’t actually see. Sometimes I can even get a better copy for research purposes on the computer than the original is. You still have to go back to the original sometimes just because of the nature of the original, like maybe an old newspaper that’s all folded up is hard to copy or something that’s really big that’s difficult to scan. Like there’s a a few reasons but for the most part once I scan it I actually, I, I tend to use cameras for all of my copying rather than scanning for a whole variety of reasons. I find it much faster. And also there’s so many different sizes of things that I want to copy that don’t fit well in a scanner.

Mark Thompson (16m 19s):
Whether it’s a really small picture or a really big, you know certificate or a big legal document, something like that that doesn’t fit well on a scanner. So I tend to use my cameras and copy stands to do it kind of like a small archive would. But once I’ve got the digital copy then I just work off of the digital copy.

Diana (16m 38s):
Well I totally agree with that and your point that it’s so much easier to blow it up on your screen, especially if you’re working with like tiny handwriting or any kind of something you’ll be transcribing, you know you’ve gotta have it digitized but also with photos. And I love what you said about using some of the tools to make them more clear, really blowing ’em up, trying to see in the background You know there’s lots of neat things that you can do. So I’m curious really quickly as a follow up to that, you mentioned cameras plural. So do you have different types of cameras? Are we talking about a phone camera or do you have like another type of camera that you use?

Mark Thompson (17m 17s):
So when I’m doing my archival work I have a digital SLR that’s on a copy stand that you know points down at a table. It’s all lit and everything like that. So with that I connect it right to my computer so I can go really fast. That being said, you can get amazing quality photos with the phone that’s in your pocket. I have no reservations about recommending that people use their phone to take copies while they’re, you know they’re visiting, you know Aunt Frieda or Uncle Bob, you know at the family reunion and copying the, the pictures that are on the wall. With a little bit of practice and paying attention to the lighting you can get amazingly good photos with very little effort with just with your camera phone.

Mark Thompson (18m 0s):
But at scale though, when I’m gonna do a hundred photos and I want them all to be properly lit and everything else, I use a digital SLR to do that much faster and, and it’s really helpful like doing anything when you’re doing a lot of it, it’s all about trying to figure out how to do it consistently, repeatedly, every single time. So actually I have the camera plugged right into my computer so when I take the picture it goes directly into the software that I use for organizing and cataloging them. So it just every single step that you can do to save yourself 10 seconds per copy just pays off in the long run.

Diana (18m 33s):
I love your system. Would you come and take care of all my photos too?

Mark Thompson (18m 39s):
I’ve been, I tell you it’s, it’s school of hard knocks. I, I learned a lot from like you can learn so many things off of podcasts and YouTubes but my home system now looks more like an archive than it does a home system. You know, but it’s all, there’s very few things that I’ve actually purchased that are you know specific to it. The only thing that I did do is actually getting a copy stand. For anybody out there who’s really serious and has a large collection, a copy stand is a really, really helpful piece of equipment for their home digitizing system. It can really help you go fast. I found one secondhand that was about 35 years old from a retired newspaper photographer for 75 bucks off of Facebook. And I, I can’t tell you how happy I am that I found this retired photographer and his old piece of equipment that he had sitting around in his garage.

Mark Thompson (19m 27s):
It has changed my archiving life.

Diana (19m 29s):
That’s so awesome. And when you have a good system and you have it all set up, then it’s fun. So You know whether it’s big like that, a large scale or something more simple. I think the idea of a system is so important.

Mark Thompson (19m 41s):
Oh yeah. Like I said before, even if you’re doing it with your phone consistently learning how to use your camera phone is perfect. If you’ve got a scanner getting used to how your scanner works and getting your settings right, just just rubbing off those rough edges about how you do things so that you can do them consistently every time and builds, build a bit of a process for how you do it. It’s very much like, well like the Research Like a Pro process, right? You find a, you find a once you find a process that works, if you do it all the time, you’ll learn the parts that work for you and the parts that You know where there’s some rough edges and you’ll grind them off and you’ll get better at it every time. And this is family archival work is all about, unless you’ve only got five things in your collection.

Mark Thompson (20m 22s):
But when I do workshops and talks and such about family archival work, the number of people who hold up their hand for I have more than a thousand items in my collection. Like it’s half the room. And if you’ve got a thousands of anything, you’ve, you better have a process or you’re gonna give yourself a lot of frustration.

Diana (20m 41s):
I love that.

Nicole (20m 42s):
Well Mark, after you digitize the items, did we finish all the steps?

Mark Thompson (20m 47s):
So once I’ve got it digitized and then you research of course, this is where AI really comes into play after you’ve got the items digitized. ‘Cause until it’s digitized, there’s no way for artificial intelligence to really get involved other than the learning steps. But once you’ve got a copy of it digital, there’s, there’s two places that AI can really help. One is helping you with the research of each individual item. And the other way is how all of the items relate to each other. So are any of them consistent? Are any of the people in them, do they appear to be the same person or related to each other? Do any of them appear to be taken in the same place? AI can really help with that.

Mark Thompson (21m 28s):
Organizing, categorizing, building collections out of your items that is so important in archival management. A lot of people think about their archive as a hundred independent items or a thousand or 10,000 independent items, but the real power of having a lot of items in your collection is organizing and categorizing them in ways that help you understand them collectively so that it can make it easier to understand them individually. That’s really kind of what distinguishes trying to figure out what’s in a single photo from how all of those photos fit into the big collection. And if you can do both, you can learn things you can’t learn any other way. It’s a lot of fun. I love laying out all of my photos.

Mark Thompson (22m 10s):
Like in the old days I would lay out all of my photos on the biggest horizontal surface I had in the house, my dining room table. And you do this bucketing on the table, you, oh this is the same person. These look like they were taken in the same place. These ones look like they were taken at the same time. These ones are in this a different family group. That exercise that people who are do a lot of of photography get used to that. Building the stacks and understanding your collections. AI can help with that so much at the collection level. But also once you learn about the collections, sometimes it can answer questions about some of your individual photos.

Mark Thompson (22m 50s):
Like for example, if you can find all of the photos that were taken at a funeral, which is a really common place where where photos get taken, I’ve got a bunch of pictures at a funeral or I’ve got a bunch of pictures at a wedding or I’ve got a bunch of pictures at a reunion. Any one of them can help you figure out who was in the photo when the photo was taken, where the photo was taken. And then it answers the questions for all of the other ones at the same time. It’s a really fast way to get answers about every individual photo by understanding how all they relate to each other.

Nicole (23m 20s):
I love that. I think it’s really important to think about how they relate to each other. And I think we often do skip over that idea.

Mark Thompson (23m 27s):
I have learned so many things about photos that I, I have long had in my collection by looking at other photos that I have on the collection. Th this is kind of the superpower of looking at all of your photos and it’s so hard when you’ve got hundreds or thousands of photos to look at all of them. And this is where artificial intelligence c can come in very, very handy. And I like there’s one particular tool that’s, it’s actually, it’s been around, this isn’t something new in the artificial intelligence world but it’s an artificial intelligence tool that’s been available to us for probably I’d say six or seven years, maybe even touching on 10. But it’s become quite popular I would say in the last five is artificial intelligence supported tagging tools.

Mark Thompson (24m 9s):
Tagging has been around for many years, over 20, and tagging for those that aren’t familiar with it is the act of going in and looking at a photo and then documenting who’s in the photo or where the photo was taken or when the photo was taken right within your software over 20 years ago. People flipped over the back of the photo if they were very nice and wrote on the back of it or wrote on the front of it or did something and, and we all love when somebody has taken the time to document a photo particular as a family archivist and researcher, but where do we record it now? And the tagging tools that are available to us now, probably the granddaddy of them all is Adobe Lightroom.

Mark Thompson (24m 50s):
But there’s also other tools that are built right into probably every single piece of photo editing or photo organizing software that anybody out there owns. I, I happen to use Lightroom but there’s other ones. There’s a really good one out there now called Mylio that’s another that’s sort of trying to be a cheaper version of Lightroom for doing some of this stuff That’s really targets the the family history community. But you can even do this stuff with the tools that you already own like Apple photos or Google photos or Amazon photos. All of them have some kind of tagging tool and and have for years and years. But what’s gotten really cool in the last few years is artificial intelligence capabilities that have been built into these tools.

Mark Thompson (25m 35s):
It’ll actually organize all of your photos and say it looks like this same person across these 50 photos is the same person. Did I get that right? And it’ll give you an option to identify them and confirm whether or not it guessed correctly so that you can quickly tag all of your photos. It’s like a superpower as a family archivist to be able to get a tool that can help you quickly identify all of the people in your collection. ’cause it’s so hard, like I said, to look at 50 or a hundred or a thousand photos I’ve tried and you get really good at building a photo memory when you do this all the time to organize your pictures. But when you can have a tool step in, particularly if you’re doing client work, step in and go, Hey You know what across all these thousand photos, these 50 photos look like they’re of the same person.

Mark Thompson (26m 23s):
Did I get it right? And you click yes or no. That’s a superpower.

Nicole (26m 26s):
Oh that is a superpower. I love that so much. I use Google Photos And, it recognizes all of my children and categorizes them into collections for me automatically, which I love. However my, my first and my fourth look so much alike that they get grouped into one person.

Mark Thompson (26m 45s):
You know I get some joy when my dad and myself get tagged as the same person or my grandfather and me get tagged as the same person or my brothers and I get tagged as the same person. Like some people might look at that and think that it’s an error on the software’s part. I look at that and I go, there’s genetics in action where we actually, the software thinks we’re the same person. So I actually kind of enjoy it when it gets it wrong in a fun way.

Nicole (27m 11s):
That’s true. It is. I mean I do love talking about that. I think it’s cool

Mark Thompson (27m 15s):
Although like all artificial intelligence tools, we can never rely on them to give us the answer. So when it does get it wrong, the other thing that’s really nice about these tools is it also makes it really easy to click whether they got it right or they got it wrong and if they got it wrong to edit it. So particularly when you’re looking with old photos in a family archive, it regularly makes errors but the fact that it makes it so fast to fix the errors, even that itself is a bonus. Right? Doing it manually was, I couldn’t tell you how many hundreds and hundreds it might even be in the low thousands of hours I’ve spent tagging photos. But before the artificial intelligence tools came along, like I just got back from a family reunion, our 10th annual family reunion a few weeks ago in Ontario and what would’ve taken me a day to tag all of my family photos and who was in them and all that kind of stuff.

Mark Thompson (28m 9s):
So I remember them for the next family reunion probably took me 30 minutes ’cause I had 500 photos to go through and I went through 500 photos in 30 minutes and tagged everybody with AI support. And what’s worse is I probably wouldn’t have taken the time to do it the right way manually because it just would’ve taken too long. But I got essentially everybody tagged that I could reliably tag within 30 minutes. That is again, it’s a superpower

Diana (28m 35s):
That is amazing And I think it’s really fun that we have been using this ability to tag photos for years without really thinking about it being artificial intelligence. Right? Some of these tools have been around and we just didn’t really, we just started taking it for granted, oh there’s this fun new thing we can do. And we’ve all had fun doing that and I love the point that the genetics do show up in in our photos. It’s so awesome. Well we’ve talked a lot about tagging but I think you’ve probably in your experience found some really specific benefits of why would we would wanna take even those 30 minutes.

Diana (29m 15s):
Why do we wanna worry about this?

Mark Thompson (29m 17s):
Findability is the word that I use, but the ability to go into your archive and to quickly find things in your digital haystack or in your physical haystack of photos or letters, the importance of it. I don’t think it can be overstated. We all joke sometimes about how we found a a document after looking for it again the second time or the third time or the fourth time because we lost track of it in our own archive. This happens all the time. So this documenting as you go in your family archive is the difference between having a bunch of stuff on your desk and having it nicely in file folders that are all labeled.

Mark Thompson (29m 58s):
You can pull it out when you need it and tagging is essentially the way that you do that with photos. It’s kind of funny, I feel like the family librarian sometimes people call me and say, I’m looking for that picture of, you know Uncle Joe. And I say, oh okay, I’ve got 50 here in my 20,000 photos. And so the fact that they’re all tagged, it just makes it so easy to search for them and find them and all of the softwares that do this, you cannot just find that picture of Uncle Joe. But you can find the picture of Uncle Joe in the 1950s or you can find the picture of Uncle Joe at the family reunion in 1945. Or you can find all of the pictures of Uncle Joe with Aunt Helen like that. The ability to be able to search for them and query them becomes possible because they’re all tagged.

Mark Thompson (30m 42s):
You know a lot of it depends upon the software that you’re using, but it’s not just documentation like it’s written on the back of the photo, but it’s documentation that allows you to do searching. And that is, that’s like a researcher’s dream. The ability to quickly to be able to find the things that you’ve got based upon something that you can hold in your head. Otherwise you’re just going back to the great big filing cabinet in the sky and flipping through every single folder and hoping that you can find the thing, which is the way that most people store their photos and letters that are in their family archive, right?

Diana (31m 13s):
And it does speak to memory You know how many times have you looked through your archive and you find one that you’d forgotten all about that was such a great photo that you forgot about. So You know we can’t hold it all in our heads obviously

Mark Thompson (31m 26s):
It’s just too hard. Organizing your, your photos in such a way that your memory is all that you need is really important. Like for example, I, I think we’ve, we’ve all decided long time ago in the genealogy community that, and you guys teach this as well, when you store information about a person, you always use sort of their genealogical name, their birth name as their profile, right? You, you don’t refer to them by the third name change that they might have had the third time they got married. You refer to them their name at birth. I do exactly the same thing with all of my photos and letters. Everything about my documentation ties back to my family tree. So when I learn something new, I never have to do that translation in my head.

Mark Thompson (32m 8s):
It’s all about making it make sense with respect to the genealogy. ’cause as much as I love photos and as much as I love letters and other things that show up in my family archive, it’s all about taking it back to the genealogy. For me, I’m a Genealogist first who happens to really like all of the things about photos and letters. And so it really makes it easy if I organize everything so that it supports my family tree.

Diana (32m 34s):
I like that. I like what you said about being consistent with your naming because it would be really, really easy to start naming, oh this was Joseph, or this is Uncle Joe You know and it’s all the same person. But I think, you know that idea of naming them correctly, consistently so we don’t have, you know some something crazy going on out there and we’re trying to find someone is so important. That’s a great tip. Yeah.

Mark Thompson (32m 57s):
Well who amongst us hasn’t pulled up the photo, flipped it on the back and the have it say Mom.

Diana (33m 6s):
Yeah, our grandma. Which

Mark Thompson (33m 7s):
Grandma? Yeah grandma. Yeah. It’s not very helpful. Not very helpful at all.

Diana (33m 12s):
Well

Nicole (33m 12s):
We are going to have a word from our sponsors now. Today’s episode is sponsored by Newspapers.com, your go-to resource for unlocking the stories of your ancestors. Dive into the newspapers where your family’s history unfolds as you search nearly a billion pages in seconds. Newspapers.com offers an unparalleled treasure trove of historical newspapers providing a window into the past with papers from the 17th century to today. Newspapers.com is the largest online newspaper archive. It’s a gold mine for anyone seeking to uncover stories from the past. Whether you’re a seasoned genealogist or just starting your journey, Newspapers.com makes it easy to search for obituaries, birth announcements and the everyday stories that shaped your family.

Nicole (33m 57s):
It’s like having a time machine at your fingertips. And here’s the best part, our listeners get an exclusive offer. Use promo code FamilyLocket for a 20% discount on your subscription. That’s FamilyLocket at Newspapers.com. Sign up today at Newspapers.com and embark on a journey of discovery.

Diana (34m 16s):
Well I wanna dive in a little bit now into how you actually do some research with your photos. As we’ve been learning about how to use artificial intelligence, we have learned that we have to talk to it correctly. It’s like a child, we have to, we have to tell who it is and what we want it to do and we call those prompts. So I know you have gotten very good at your prompt engineering, so can you talk to us a little bit about You know how do you do research and when do you do research on your photos? So you research every photo, You know kind of what’s your system, your tools for that?

Mark Thompson (34m 54s):
That’s a really good question. You know prompting is sort of the heart of the new type of artificial intelligence tool that we haven’t talked about much yet, which is the chat bots from say ChatGPT or Gemini, Claude, these are sort of the the big ones out there. And learning how to speak with them and give them directions is the heart of using those tools. And there’s some really, really simple but incredibly powerful techniques for making the most of your time with those tools. Particularly when it comes to photo research there, there’s a few that stand out from the crowd because they’re actually able to see photos. Not all, not all of the large language models are actually able to look at a photo, you know to have what they’re called AI vision.

Mark Thompson (35m 40s):
And the two that are very good at AI vision are ChatGPT and another one that a lot of people don’t put in the same sort of crowd as as the other ones, but Perplexity. Both of them I find are very good at looking at a photo and answering questions about what it can see. The thing is is just like the large language models have been trained on 6,000 years of, of words to understand how words work well together. They’ve also been shown literally billions of photos. And so they, they’ve been trained on all of these photos and the way they’ve been trained is the large language model is shown the picture and it’s shown a description of the picture that goes along with it.

Mark Thompson (36m 22s):
And so after looking at billions and billions of pictures, it’s actually able to see, I’ll use the air quotes, it’s able to see what’s in the image. And so if it sees a thousand pictures of an apple in various stages of apple-ness, it can look at it, it can look at a, the thousandth and first picture and go, holy cow, that’s an apple And. It can do this with amazing abilities ’cause it’s not just able to see things like apples and oranges and barns and cars. It’s also able to see other things like colors because all of those photos were described as colors. It’s able to see even things like emotions in a picture. Somebody’s happy, somebody’s sad. It’s able, for a family history researcher, it’s able to do things like take a guess at when a photo was taken or where the photo might have been taken or why the photo might have been taken.

Mark Thompson (37m 12s):
And then your job as the person who shows the photo is to create a prompt that asks a question that the large language model can turn into a, a response that it can give back to, to help you with whatever your research goal might be. And and with photos, there’s a few that people tend to care the most about and it’s usually the five Ws. Although the one that people tend to focus on is the Who as a photo researcher. Sometimes starting with the Who is the hard one, unless you can immediately identify who’s in it because of, you know your own knowledge or the help of the AI facial tagging tools that we talked about a minute ago. Sometimes you gotta come at it from Where was it taken and how does it help me figure out who was in it or why was it taken?

Mark Thompson (37m 53s):
And use that to figure out where or who was in it. This, you know as a, as a researcher, you sometimes you have to come at it from different angles to get to the question that you are, that you care the most about. But AI is so good at looking at a photo and helping me see things that I might have missed. Like so often I will actually stare at a photo for literally hours going through and trying to see the story that it’s trying to tell me. What are all of those buildings in the background? What river is that? What kind of tree is in the background, what paper is it on? All of the little things that can give you a clue about where, when, who, why, and and help you answer the questions you can’t answer with your own eyes.

Mark Thompson (38m 37s):
And with AI, because it’s so methodical, when I give it a prompt, what I try to do is I try to ask it to comment on everything that it can see. You know, I ask it to act as if it’s an expert in photo analysis from a particular kind of period. I ask it to look for all of those who, where, when, why clues and then use it to tell me everything that it can see that might help me figure out the other things that that it didn’t see. And that because because computers are so methodical, oftentimes I find that even though I’ve looked at it for hours and hours, when it comes back with everything that it can see, it gives me a clue that I might’ve missed.

Diana (39m 21s):
Well I think you gave us a lot of food for thought and I think all of our listeners like me are thinking of those photos that we’ve just looked at and said, oh there’s grandma. And look at that. That’s so fun. She’s in that cute outfit and wondering what we’ve missed because we didn’t have this capability to do the research with artificial intelligence. And so I wanted to ask you if you would like to share the name of your custom GPT on ChatGPT for photo analysis because I’ve used that and it’s really fun for our listeners that don’t know much about the large language models and the tools out there, ChatGPT has something called custom GPTs where anyone if you have a Pro account, can go in and create a custom GPT and you would do this if there’s something that you do all the time.

Diana (40m 10s):
Like for Mark, it’s analyze photos and you can put in all of your prompts, all of your parameters, what you want it to do just once. And then you can just ask the questions because you’ve already set the stage for what you want the chat bot to do. So Mark has created one that you can literally just drag and drop your photo in and it will start analyzing it for you And it basically gets you started on the analysis and on the chat and then you can continue to ask it questions and you know, have the conversation to learn more about it. So Mark, do you wanna share the name of that?

Mark Thompson (40m 51s):
Sure. I’ve, I’ve built a custom GPT called Mark’s Historical Photo Analyzer that anybody can go into the GPT store at ChatGPT and look up and use. And I’ve, I’ve worked on it over the course of a few months to try to put all of the things into it that will help it act on my behalf to look at a photo and see anything that I might have missed or to confirm things that I mi might have had a bias about that I might have thought it was A, when it turned out that it was B. It’s, it’s a little bit long-ish, but that’s what happens when you actually build a custom GBT. You can take the time to actually really tune it to what you need. So I’ve, I’ve asked it to you know, be an expert in historical photo analysis ’cause I tend to use it for historical photos, not modern photos.

Mark Thompson (41m 38s):
And then I’ve, I’ve loaded it up with all kinds of things that it should look for and to try to have an opinion on to see if it can see it inside the image. And I ask it for everything that I can think of that I as a photo analysis expert might try to do that I want it to do on my behalf to make it easier for me to do my job. I ask it to look at the lighting, I ask it to look at all the things that it can see. I ask it to look at the people it can see. I ask it to look at the actual container. It might be sitting in a frame, it might be hanging on a wall. Anything that it can see I want it to have an opinion on and then to give me information back. So rather than have to type that in every single time I built a custom GPT that I can load up every time that I want to.

Mark Thompson (42m 24s):
That’s available to anybody, any of the listeners out there who want to use it to, to run. And all you have to do is find it in the GPT store, load it up with a photo and then just click go. You don’t have to do anything because all of those instructions are already written into the custom GPT and they get run as soon as you click submit. You don’t have to do anything except load the image itself.

Diana (42m 46s):
Now I wish I had 10 hours in the day to go look at this with all of my photos because I do have some that I’ve got some questions about. So that’s so fun and Thank you for being so generous to share that with the public because you can make custom GPTs just for yourself and keep them private if you have information you don’t want to be out there. But this one is shared and it’s public and so all of our listeners can go try it out and just give yourself a couple hours to play with it because you’ll want to continue playing with it once you get started.

Mark Thompson (43m 19s):
Oh it’s a lot of fun. Like like I said, I’ve looked at some photos for hours and I’m always amazed there’s that one thing that I missed and sometimes it was an important thing but sometimes it’s just a little thing that my eye wasn’t attracted to. You know there’s, particularly in photos when there’s lots of people or it’s a big event, we’re all human. We tend to You know, gravitate toward the thing that we wanna see that we most wanna look at. You know computers are methodical, they look at everything, they’ll do exactly what you ask them to do. So by using building up this, this custom GPT, I’ve kind of trained it to try to see all the things that I might miss. So I use it all the time with my photo analysis just to make sure that I didn’t miss anything or if I don’t have much time and it’s not that important a photo, I’ll start with it just to kind of kickstart my thinking about what I wanna apply my eye to.

Mark Thompson (44m 7s):
Like all things with artificial intelligence though I never trust it. I always, you know, confirm it with my own eyes. I never copy and paste everything with artificial intelligence. I just use it as a starting point and it’s a great starting point. Can save you a ton of time.

Diana (44m 21s):
It is great. And I, while you were talking, I uploaded one of our mystery photos that has a woman with gray hair that is supposed to be the daughter and she looks as old as the supposed father. We have puzzled over that photo for so long, but I just popped in the photo and clicked go into your photo analyzer and oh my goodness, it comes back with a table with observation about the subject matter, composition, clothing, lighting, background, facial expressions, condition and mounting, photographic technique and age. Okay, so here’s the key one. Let’s see what it says about the age. The man appears to be older possibility maybe his fifties or sixties suggesting he might be the patriarch.

Diana (45m 3s):
The women’s ages disappeared to range from middle age to older adults. The women seated next to them could be his wife and the women standing behind them might be their daughters or other close relatives. Oh that is so interesting. Well this is a fun thing. And, it gives you just other ideas I think is the important thing.

Mark Thompson (45m 21s):
Yeah, isn’t it interesting how it goes from the seeing it to actually starting to draw conclusions about what it’s seeing, like going from just looking to analyzing, that’s the place, the researcher, you can look at it and go, oh, I never thought about that possible conclusion. That’s actually where it gets a lot of fun is ’cause it just, it kickstarts your idea machine in your own brain about that might change your, change your mind about what it is that you’re looking at.

Diana (45m 45s):
I agree And, it does tell you about the clothing. You know, it says the women were wearing high neck dresses with lace collars, typical of the late 19th to early 20th century So. it does give you clues about when as well as it does have a possible location. You know it says likely in the western country, the United States or Europe. So that’s not very specific, but they were right because I think it was Oklahoma, so it was was the Western country.

Mark Thompson (46m 13s):
Yeah, it’s just methodical. Computers are wonderfully methodical. They will just grind away at doing stuff that humans get bored with doing time and time again. So that’s where AI can come in really handy.

Nicole (46m 24s):
Well let’s wrap up this fun episode with some suggestions from Mark. Can you tell us how our listeners can decide where to get started with using AI in genealogy?

Mark Thompson (46m 36s):
Oh sure. Another really good question. You know, I think a lot of people struggle sometimes with where to get started with AI. Like what are the things that it’s you know good at? And and I think it’s helpful to go the other way. What are you already good at? What is it that you already do with your genealogy? ’cause that’s actually the best place to start is doing the things that you’d love to do and then try to figure out how to apply it to AI or how to apply AI to it is probably a better way to say it. So for me, I like photos and I like letters and there’s things that I do in there that make it so valuable to apply artificial intelligence to. I mean, another thing that I do a lot of is I analyze a lot of obituaries, like a lot of Genealogists do.

Mark Thompson (47m 18s):
There’s another thing because I do it time and time again, figuring out a way how to use the chat bots to help me look at an obituary is really useful. Summarizing documents is something I do a lot. So I look at what I do regularly and then I look for how AI can help me do that. There’s actually two reasons why this is a good place to start. One is, is it something that actually matters to you? The other one is you actually have the words in your own vocabulary that you can use to describe how to do that thing or why it matters to you or what you want to get out of it. This is actually a really, really important part of prompting is you actually have to be able to give guidance to a computer on how to do a thing.

Mark Thompson (48m 2s):
And so if you don’t have the words to describe that, it’s really hard to get a chat bot to do it for you. But if it’s something that you do all the time, you already have the lingo. I can describe in my custom GPT what to look for in a photo. ’cause I look at a lot of photos or I can describe how I want it to summarize a letter. ’cause I have summarized hundreds of letters. But if I’m asking it to do something that I’m not familiar with, I don’t even have the words to say it. And large language models are all about language. They love words. So I would just encourage everybody to find the thing that you already do that you’re already good at. And that’s a great place to get into AI because it’ll be much less frustrating or painful because it’s already something that you love to do.

Mark Thompson (48m 43s):
So you’ll work at it because it matters to you and you can use the words that you already have to give guidance to your chat bot.

Nicole (48m 50s):
That is such a great tip. And that’s funny because that’s kind of how I started using it too. Like, okay, I have to do this thing that I do all the time and now I want to tell the chat bot to do it for me. And I’ve also heard the analogy of talking to your assistant, like if you were training somebody to do something for you, how would you tell them to do it? And I think about that a lot because I have had assistants and giving them instructions. I usually write them down in a Google Doc, so it’s something that I can easily relate to. If I was gonna write this down for my assistant, what would I tell them to do and how would I tell them to do it? I usually give them an example and I usually give step-by-step instructions.

Nicole (49m 30s):
So I love that you said that and I think that’s such a great tip for people getting started.

Mark Thompson (49m 35s):
Oh, I love looking at the stuff that you’re doing with Airtable and creating, creating research reports directly from Airtable. And I look at that and I’m like, all so kinds of jealous because you’re so good at it. And then I back it up to what I just said a minute ago and I’m like, well, it’s a thing that you already do. It’s a thing that you’re already really good at and you already understand how to take your Airtable information and put it into research reports. So you actually have this process in your head already. So going from doing that, that’s a perfect example of doing something that you already know how to do and making it easier and faster using artificial intelligence. That’s a classic example of do what you’re already good at with AI.

Nicole (50m 17s):
You’re right.

Diana (50m 18s):
Well And, it also speaks a little to learning and education in genealogy. So if you really don’t know what you’re doing with genealogy, AI is not going to help you much. You’ve got to know what you’re doing. You’ve gotta learn the basics and then it can really speed things up and help you. But there is no substitute for experience with the records and analyzing them, learning from them, writing about them. So I think it’s all about the limitations and experimentation. So we’re all just beginning to learn about how to use it. So it’s been fun to be on the journey and to have you on the podcast talking about what you are really good at with the photos. So Thank you so much Mark for taking the time to be with us.

Diana (51m 1s):
I hope everyone will go check out your podcast and start learning like the rest of us what’s the newest and the greatest things out there?

Mark Thompson (51m 10s):
Thank you very much. It has been my absolute pleasure to be here. Thank you very much for inviting me on the show.

Nicole (51m 14s):
Well, thanks everyone. Have a great week and we’ll talk to you again next week. Bye-Bye.

Diana (51m 18s):
Bye-Bye.

Nicole (51m 57s):
Thank you for listening. We hope that something you heard today will help you make progress in your research. If you want to learn more, purchase our books, Research Like a Pro and Research Like a Pro with DNA on Amazon.com and other booksellers. You can also register for our online courses or study groups of the same names. Learn more at FamilyLocket.com/services. To share your progress and ask questions, join our private Facebook group by sending us your book receipt or joining our courses to get updates in your email inbox each Monday, subscribe to our newsletter at FamilyLocket.com/newsletter. Please subscribe, rate and review our podcast. We read each review and are so thankful for them. We hope you’ll start now to Research Like a Pro.