Mark Thompson & Steve Little chat about AI with the Genealogy Guy
Download MP3Jingle 0:04
The Genealogy Guy podcast, demystifying technology and exploring family tree research, please remember to subscribe and share the podcast with family and friends
Genealogy Guy 0:17
and welcome along to another episode of armchair genealogy with me, the genealogy guy, the growing interest in AI and what it means to our daily lives grows every day, and I thought it was very important for us to find out more about the potential impact on genealogy and research. Joining me in this episode is the guys from the Family History AI show, who are Steve Little and Mark Thompson. And what's interesting is AI is just being used all the time, artificial intelligence. People are trying to get their heads around it. And I just thought, let's just go to the source. Let's go to the guys that have sort of semi plugged into it all and trying to make sense, like the rest of us, but are a little bit in front of us. So welcome along Mark and Steve. So hello, Mark, good morning.
Mark Thompson 1:01
Mell, very nice to meet you.
Genealogy Guy 1:03
And hello Steve,
Steve Little 1:03
Great to meet you. Mell, and I've loved your introduction, because wedon't hold ourselves out as much as experts as we're just about 20 minutes further down the path than everybody else.
Mark Thompson 1:15
Yeah, you just, you just got to be one chapter ahead of the class
Genealogy Guy 1:19
To kick us off, just so that people get an idea of where you're coming from. Mark, what's your background? How did you get into genealogy? And how does that connect into the technology for you
Mark Thompson 1:27
Oh, well, actually, the technology for me came first. I was a career long technologist, and I've been involved in IT in some role or other my entire career, usually managing teams of people who are running big projects. So my whole life has been being one chapter ahead of everybody like that's I used to say, I got paid to read the manual. Andso whenever a company wanted to change something big, I usually got called in to work on the project or to run the project. And so that was, that was my whole career. And then when I was in my early 20s, I had a great uncle come visit, who was the family genealogist beforeme, and he put the bug in me. He told me a couple of stories and got me all excited, and and, and kind of as I was a young IT person. I wasin my 20s, I saw this massive data problem, like as he started to tell me the stories. And he pulled out his then paper scrolls, which I have inherited since. He pulled out these big paper scrolls. And I was like so much information, how do you keep track of it all? And he said, well, on this big sheet of banquet paper, thank you very much for the last 40 years. And I was the first thing. I thought was, well, that's not my project. Here's my project. So I started looking for software about an hour later, and that was 40 years ago, right? So, and I've been working on genealogy, you know, a little quiet, of course, in the busy part of my career, but now that I'm now that I'm retired, it's become just a full time exciting job, just tracking information and helping other people understand how to use it. And what about
Genealogy Guy 2:59
And about you, Steve?
Steve Little 3:00
I loved how Mark described himself as he was the guy who would read the manual. I was kind of that guy too. If folks know the acronym or initialism, RTFM, read the freaking manual. That was me too. And like Mark, I had 40 year interest that came together in a perfect laser like focus two years ago, two and a half years ago. So I have 40 year interest in language, technology and family history. I had an aunt who got me hooked on family history, but she was a realgenealogist. She was a serious genealogist, and I started doing data entry for her in the 1980s with an early DOS version of Family Tree Maker, and just by osmosis, typing in her research from the previous 40 years was my exposure to family history, typing in the names of my grandparents that I didn't even know their names. But over 40 years of that kind of exposure, you learn to fall in love with those ancestors you didn't even know were your own. Similarly with technology, my first computer was a Commodore 64 actually, I had three of them in a couple months. The first one blew up on Christmas morning, the replacement, this was before FedEx took about six weeks to arrive, and the second one blew up when I plugged in. And it was Valentine's day before I finally got my Commodore 64 to fire up. And I've been doing hobbyist programming since then with quick, basic and Perl and Python and other things like that. And I've always had a love of language. After flunking out of nuclear engineering school, I switched to English literature, finished that up and then study. And Applied Linguistics at the graduate level computational linguistics, which is the flip side of natural language processing, one of the pillars of today's large language models that we talk about in artificial intelligence. And so I've got a technical understanding of language and a love of language, along with my love of history and technology, have just made this the most exciting two years of my life.
Genealogy Guy 5:28
So that leads us in, nicely into everyone thinks that AI is a new thing, and what people don't realise is it's slowly coming to into its own being. So take us back to where's it come from? What was the first time that we got exposed to all this stuff?
Mark Thompson 5:42
I was going to say, you want to start with Alan Turing or something a little bit more recent than that?
Genealogy Guy 5:47
Bit more something where people will actually useong AI.
Mark Thompson 5:51
I think as far as Joe and Jane genealogists are concerned, the AI generation started about two years ago. You know, Steve. Steve and I met four or five years ago, in a actually, in a genetic genealogy class, and we made fast friends and but when the we both knew that we both had a love of of artificial intelligence and IT stuff in general, through our interaction talking about genetic genealogy and how we approach those kinds of problems. When chat GPT got launched, I was like, I gotta give Steve a call. And I think we had a zoom call about an hour later. We were just, we could just tell that we were both really, really excited. And we thought that this, there was definitely going to be something there worth diving into. It literally started just as an interest, and then it got kind of exciting about a month later, and we bought, there's actually a thing here. So we decided to, you know, sure, you know, roll the calendar forward aboutsix months. We were, you know, we were doing classes on AI, and we were and Steve got really big, really fast with the National GenealogySociety here in North America. And, you know, things just took off because there was such a, there was such a need to, there's a lot of people saw an interest in what it could do, sort of like the sort, I would say, the technology centric folks, but the genealogy crowd just saw so much potential. But it was, it's a whole new way of thinking, a whole new way of interacting with computers. And so there was a lot to understand, particularly initially, where the limitations were, because if you look back 18 months ago, not very much worked well, but those things that worked really well worked amazingly well. And the curve has shifted, of course, over the last two years, but there's still a lot of aspects of confusion as people are trying to figure out what the right place is to use these tools.
Genealogy Guy 7:39
For me, I mean, I became a way when you had things like predictive text on your phone, and that was like, you know, that's the origination for me from my perception is, that's when AI started to just sneak in through the back door, and people were going, Oh, it's just a handy tool. It's like, a whole of AI is a handy tool. It's just getting better. So where are we at now? What sort of things are we able to access through like genealogical research and using AI?
Steve Little 8:05
Like the way you described it as a predictive text, that's a very solid grounding that folks can have, except you might qualify it to say it's anot wholly reliable predictive text, that it can sometimes be a creative predictive text. Old MacDonald not only had a farm, but he had all kinds of stuff on his farm, cows and chickens and goats. And sometimes it'll guess a cow, a chicken or goat, but there's a non zero chance it'll also predict Old McDonald had a rattlesnake on his farm. So you have to use these tools with a little bit of respect and caution and healthy dose of understanding that they are not like a calculatorthat returns the same result every time you punch in the cube root of 27 and get the answer 3, these texts have some flexibility, some squishiness built into them. They're much more suited for language than for numbers, and words are slippery fish, and so these tools can can be really helpful, used with the right expectation and in the right ways. And what's challenging and exciting is that that changes so fast and dramatically now, even if we still had the technology we had two years ago, chat GPT 4 came out in March of 2023 so it's almost two years old. My website, AI, genealogy insights. I started it at the same time, and even if nothing had changed or progressed from two years ago, we would still be learning how to do the stuff that becameavailable to us two years ago today. But the advances keep coming every month or two or three, there's significant, measurable jumps and abilities, little tasks that these tools can do with language that become make them more useful to us as researchers, but discovering what those uses are and how to safely use them, and how to understand that it can make mistakes, and check for those mistakes. All of that is falling on the shoulders of everyday users, the creators of these tools. They don't they don't know genealogists exist or is even a field. They're not making this for us, and they're not making it for anybody in particular except themselves. And they do not publish user guides or manuals. There's barely any instructions at all on how to use these tools. So it's an exciting time for citizen scientist, citizen linguist, amateur genealogist and historians and researchers to makemistakes, big mistakes, and learn what the boundaries and limits andthe failure points of this technology is, and at the same time, you can also map out their areas of usefulness. And there are many.
Mark Thompson 11:16
One of the things that we found is that there's a few things that people can do when they're getting started that can really make a difference. First and foremost, you know, Steve said experiment. There's about a 20 hour window where people are a little bit lost and a little bit confused and they're not quite sure what these are, what these tools do, because they really work differently, that that squishiness that Steve referred to, that it seems a little bit random, that the answers I get back seem to make no sense. You start to get a feeling for how these things are different than Excel or a calculator or PowerPoint, and they don't do the same thing every time. It's a feature, not a bug, as the as the old product manager in me used to say, and as we figure those things out, there's, there's a magic thing that happens around 20 hours of experimenting with these tools that makes them start to make sense, and how they're so good at language, because for the most part, we've never seen a tool as consumers that is good at language. We've found tools that are goodat numbers. We found tools that are good at like laying out pictures, but we've never actually had a tool that could process words. And this takes some getting used to and you get to that 20 hour window, and magic stuff starts to happen, and things and things start to click in your head. And the one other thing that genealogists, I think are really well suited to do when they use AI tools like large language models, like chat, G, P, T and Claude, we are naturally pessimistic about the accuracy of an of an answer. That's our standing start, is you're lying to me, and that's such a good place for a genealogist to work from when you're learning how to use a large language model, because they will lie to you regularly, and genealogists assume everybody's lying to them, so they check everything. So we're actually better positioned than a lot of people in different backgrounds to use these tools effectively.
Steve Little 13:01
Love how Mark put that one of my favorite other podcasters. I hope Ican. If this is not acceptable language Mell just bleep me out. But here it comes. She says other people's trees are crap. That's the skepticism that genealogists bring toward other people's research, and that's a healthy skepticism to have, especially with large language models, you need to be very, very skeptical of the results of these tools. And depending on how you use some tools, you have to be more skeptical than others some uses, but but a healthy dose of skepticism with genealogist serious genealogists have we all know, we've got friends and cousins who just shovel everything. They've got family trees with 30 or 50,000 people in them, and they'll just add any random person to their genealogy in quotation marks, but at a serious level, folks are making sure that what they're collecting and presenting is real and true and accurate to the best of their abilities. And serious dose of skepticism when using these tools is essential.
Genealogy Guy 14:18
Douglas Adams, when he wrote the answer to the universe and everything is 42 and the question was, but you don't understand the question you've asked. That's why I'm giving you the answer of 42 and it's our job to be that person that figures out, well, what does 42 actually mean? And you can only do that by using the right prompt. And that's that's a whole new ball game, because you have to learn how to communicate with something that's thinking differently to how we expect it to communicate. And I find that always fascinating.And the bit about lying ah all the time, and you can call, you can call an AI out, and eventually they'll go, Yeah, I made that up.
Speaker 1 14:56
I love ferreting away at broken computer systems like it's one of the it's one of. Little joys in life. It's one of the things that kept me sane over the years. Oftentimes, there's a reason why that hallucination happened, like there's some aspect of the information inside the large language model that wasn't shaped the way that you thought it was that resulted in the answer that it gave. You know, it's kind of like when you do a search and replace in a big document, and you think, I'll just search and replace this one word and and everything will be fine. And then you realise that you actually that little bit of text existed in 400 places that you never thought it did. Same thing happens in a large language model, except instead of having a few 1000 words in your document, there's a few trillion words in its document that it's looking through. And sometimes unexpected things happen.
Genealogy Guy 15:37
Let's take a very short break chatting with Steve Little and Mark Thompson from the family history AI show, to say a very big thank you. And again, to all the new listeners that I keep getting in each episode I publish it really is fantastic, and I really appreciate it. And thank you for the emails and comments and positive feedback. It's lovely to hear from each listener. As a side note, I take a lot of time writing social media posts to promote each episode, and something that would be really helpful for me and you please subscribe, as it means you get notified every time a new episode is uploaded, and you don't have to rely on finding the post somewhere on social media. Talking of social media, I will stop using X, formerly Twitter, asa social media platformer. I just cannot put up with the ads that are being pushed at me and political comments that I'm just not interested in any of them. I haven't subscribed to them, but I'm just getting bombarded with them, and I've got better things to do with my time. And added to that, they're now starting to charge the companies that help publish podcasts to make them pay so that it automatically gets posted on their platforms, which I think is not good. Instead, I'm going to concentrate mainly on Facebook, LinkedIn, and my newest platform of choice, Blue Sky, (https://bsky.app/) which, if you haven't tried it, I'll recommend it as an alternative. You can find me there under the name at symbol, @genealogy guy.co.uk
Jingle 17:13
ArmchairGenealogy.com, for generations of all ages.
Genealogy Guy 17:19
Anyway, let's go back to our chat about artificial intelligence with SteveLittle and Mark Thompson from the Family History AI Show. So the big question most people are asking is, can we trust AI and for a bit of fun, putting my late father's details or just his name and the area he lived, and asked chatgpt to tell me something about his life, and he came back with a load of nonsense and made up stuff that he was a football player and played in the FA Cup. It was just a complete made up story. And what's worrying is I know that some people would take that information and think brilliant. It's written and done some research for me. I'll put this information in my tree, which can be very dangerous,
Mark Thompson 18:04
and some people have,
Steve Little 18:06
yeah, those are one of those lessons. We can tell folks that the stove is hot, but many folks are as hard headed as I am, and just have to touch the hot stove themselves to learn. But that lesson that you described is that or my rule of thumb, and this probably will not always be true, but during these first two years and for the foreseeable future, accuracy is in proportion to publicly available information. There's very little information, no information available about my surname, fourth grade grandfather. So of course, chat GPT has nothing to return. But that doesn't mean that it won't just make something up, especially if I ask it to tell me something. If I ask it an open ended question, it's going to give me an answer. It does not know what is real or true. All it knows is the statistical relationships between words. And when I give it a prompt that says, Tell me about my fourth great grandfather, well, it just tries to figure out in the sphere of what syllables and morphemes and words might follow. Tell me about my fourth grade grandfather, and it just goes to town. It's not doing fact checking or error correction. At least some. We're starting to see a little bit of that. And if we talk about where we're going, that's something that will probably get better. That's what the whole world wants, and evidently is willing to pay for their services. Now that cost about $200 a month that claim to be able to answer those kind of questions better. But it's. Still very early days.
Genealogy Guy 20:03
Where are we heading? What's the next big thing that's going to happen with AI that's going to make people suddenly wake up to andsmell the coffee, as they say?
Mark Thompson 20:10
You know part of it depends upon who you are, you know. But if we think about the you know again, going back to you know, Joe and Jane genealogy, the people who do genealogy as hobbyists and or even serious family historians, but, but not the folks that are, you know, they've got a budget for tools, right? I mean, just the people who the people who go out and they search the internet and they visit the local archive and the, you know, the local county library to try to figure stuff out. Those people, AI's coming. AI is already here. It's now built into Microsoft Word, it's now built into Microsoft Excel, it's built into Google Docs. It's built into it's getting built into your iPhone, it's getting built into Microsoft Windows, it's getting built into your Mac. AI is not only coming, it's already here, and we're going to see these little features that take advantage of the spectacular capabilities of artificial intelligence show up in every little app that we use. We're not going to have to go and if we're not going to see everybody on the planet go out and get a subscription to chat GPT, we're going to see people just they're going to click the little button that does AI stuff directly in the tools that they already use. That's actually, when we look back in two years, it's just gonna have appeared inside every consumer piece of software and hardware that we have.
Genealogy Guy 21:28
The fact that it can sift through so much data, it's the speed at which it can do that, what would take you days, maybe going through lots of forms. It can pull out all the names of one person and then the dates displayed, and you can find the person that you're looking for.
Mark Thompson 21:43
You know, any other day I was, you know, my my wife, asked me to to make a recipe while she was doing something else. And it's a recipe that I've never made, but I've seen her make 49 times, right? So it's the classic. I can't get there because I've never driven the car myself, but I'm pretty familiar with the destination when I see it kind of problem, right? And so I wanted to get her to give me the directions to how to do this quick recipe that I kind of knew but didn't really know, and I would definitely screw up if I just tried to wing it, which is always a bad recipe for marriage happiness. And so I pressed record on my voice recording tool on my iPhone, and I just recorded it all. And so, and her description of how to build this recipe took about 90 seconds, or how to build this, uh, dessert took about 90 seconds. And I could see her thinking, Well, what's he going to do with this? Because now I got to listen to this 90 second thing, right? And I pressed the transcribe button, and it did an automatic transcription. And now I've actually got step by step directions written in text that I can use. And she was like, that's actually pretty cool. The fact that I know that there's a large language model behindit is relevant, right? The fact that it's pretty cool, and it was a tool that used that met my need, that's where I think we're gonna, you know, so many people are gonna see so much incredible benefit we can get into the tech. And there's lots of cool tech to understand, but it's working so well now for really, really small, easy to understand usecases, like little, little things that we want to do, that it's just going to show up all over the place in ways that we never imagined. And every day, some developer or some product manager or some corporate person is going to come up with a need, or some genealogist is going to come up with a really neat idea to build that one little widget that helps them do some task.
Genealogy Guy 23:23
I think that's ultimately it's that killer widget, that killer app that people are waiting for with AI that they all know it's coming and they all know, but no one can predict what it's going to be that's going to make everyone go, this is what it was built for. This is perfect, but it's an exciting like you said earlier, it's so exciting because I've come in at the peripherals. I remember the birth of the Internet in the late 90sand thinking it was like the Wild West, but it was so exciting. And AI is exactly the same for me. It feels the same. You've got the good stuff, you've got the bad stuff, and you just think, I'm not interested in bad stuff. So it's just an exciting period.
Mark Thompson 24:01
Yeah, focus on the good stuff, ignore the bad stuff. Everything will be fine. What could go wrong?
Steve Little 24:08
Exercise my frustration by trying to keep people from wandering into the tar pits, holding up the good uses, the responsible uses, is probably the better way to spend time.
Genealogy Guy 24:20
It's a tool. It's not the answer to everything. It's just a fantastic tool. It's going to be so, so great. So I'm listening to your podcast, guys, and that's that's how I discovered you guys, and that's why I came to you to to have a chat with if people want to find out more, what places, what blogs, podcasts, shows, places, where can they get moreinformation?
Steve Little 24:38
Mark, and I also have a podcast called The Family History AI Show podcast (https://podcasts.apple.com/gb/podcast/the-family-history-ai-show/id1749873836),and we also teach and will actually be in San Diego, California at theend of March. We were very glad to have been invited by CeCe Moore to participate in. In their event called the Institute for Genetic Genealogy. I believe this is the 11th year of their convention, and the they added an event to the beginning of that weekend long conference on genetic genealogy. They added an AI day, which is a practice we're seeing happen more and more at regional genealogical conferences, and so Mark and I will be in San Diego at the end of March, and about every week or 2, 10, days, we release an episode of The Family History AI Show podcast.
Genealogy Guy 25:34
I can only recommend that my listeners also listen to your podcast, because it's you just talk about it so enthusiastically, and also point people in the right direction. And just say, Go, play, go. Look at these things. Don't just sit back and wait for the perfect answer, because you have to learn yourself.
Mark Thompson 25:52
Yeah, that's the thing that we see again and again. The places where people get the most use out of AI is doing the things that they already know how to do, and that's because they have the words for it already, and it's a language processing tool. And if you know how to if you love history of your county, or if you love archiving, or if you love genetic genealogy, you've already got the words to describe it. You've already you already know how to talk to the chat bot, because you say those words all the time already. Reach in and just type the words that you already know in and see how it goes. Add 20hours, you'll be in a different you know, you look back and go, Wow, was that ever awesome?
Genealogy Guy 26:26
As byline I'll explain, have you come across a one called character.ai so you can talk to it. You can it's like having a telephone conversation with somebody. But it's an, it's an agent, it's an it's an AIsystem. But you get all these random voices. You can select famous actors and people. It's open source. People are just making things, and you have to keep reminding yourself, sometimes you're not talking to the actor. You're actually talking to just an AI, but it's so convincing, because it starts chatting to you, and you can just say, I hope you're having a nice day, and what are you up to? What have you been reading? And it's scary, but it's so fun.
Mark Thompson 27:00
They are really good at words. They are really, really good at words.
Genealogy Guy 27:04
Mark what are you up to? Have you got a blog? Have you got something separate?
Mark Thompson 27:07
Yeah so my my blog website, is making family history.com (https://makingfamilyhistory.com/). I've been writing thatblog for a couple of years, but most of my time now is spent in doingwork with Steve. We do a lot of instruction on artificial intelligence. We also have, you know, we we tend to kind of like the the technical aspects of genealogy, whether it's genetic genealogy, artificial intelligence, archiving, the things that involve a lot of data. But we spend most of our time doing AI, and we've got, by the time your show could by the time this episode comes out will probably be at Rootstech. We've both we're both doing several workshops and a panel on the responsible use of artificial intelligence at Rootstech thisyear in the first week of March, and then off to San Diego to run this conference for the Institute for genetic genealogy on March the 28th so that that particular conference is available both in person in San Diego as well as if anybody wants to attend, we're doing a full day event that will take everybody from the absolute basics of artificial intelligence through to some really interesting use cases that are really specific to genealogy. It's going to be a really fun day. We find that that one day format you when you can get in four or five hours of conversation, you can really take somebody from zero or almost zero to, holy cow, I can actually use this to do some great genealogy.
Genealogy Guy 28:27
That's ultimately what people need to be. All they need to be done is shown the direction in which to go. And as we've said earlier, it's it's the playing and having fun with the technology that you learn the fastest. You can't just be told that you need to experiment and and try it out.
Mark Thompson 28:43
Some things you can teach, but some things you have to learn
Genealogy Guy 28:48
Absolutely it's been an absolute pleasure. I've been so looking forward to catching up with you guys. And please, please, let me have a chat to you later in the year and catch up and we can talk about what's happened since the last time we had this conversation. Because I think it would be a nice sort of crossover of episodes, whereI can reflect back on the first one, and people can then go, so what's happened now? Where are we? Because the speed at which this is being delivered is just phenomenal.
Steve Little 29:15
I would love to do that and and I feel certain Mark, and I will do that. Let's lay down a marker. Here's here's some stuff that is new in February of 2025, that if we look back in six or 12 months, we'll ask ourselves, wow, remember back when in February of 2025, a new buzz word is agents. You're going to hear the word agents a lot over the next several seasons, and also deep research, three different companies have come out with products that are agentic research agents, tools that try to give you comprehensive answers to hard questions, and we're talking about the generation of 5, 10, 15, 20, page reports, 80 page reports, some with source citations, some better than others. And so it's the very earliest days of that. It'll be interesting to see where that is next year this time,
Genealogy Guy 30:17
Have you got anything to add on to that one Mark? In the last last couple of minutes?
Speaker 1 30:20
I'm resisting the urge to think more than about a week out, because every time I every time I try to pin a tail on their prediction donkey, more than about a month out, I get surprised. But I would love to come back at the end of the year and go so where, how far did we get?
Genealogy Guy 30:37
And you've mentioned "Agents". I'll put in the show notes, but I'll send you a direct link. It was a documentary that was made by Douglas Adams, of all people, and it was the history of the internet. And it wasmade in the late like, about 25, years ago now ? It's an old program, so much that's in there. And in there, he talks about agents, and you're going, this guy was just so far ahead of the rest of the population. So I'll send you the link. It's on YouTube, and it's just a fantastic documentary. It's very dated, but the content you just go it's happening. It's happening now. Link to documentary https://www.youtube.com/watch?v=1iAJPoc23-M
Mark Thompson 31:11
Well the answer is still 42
Genealogy Guy 31:13
On that note. Thank you very much. Mark Thompson and Steve Little.It's been an absolute pleasure, and I look forward to speaking to you again. That brings us to an end to another episode of Armchair Genealogy with me, the Genealogy Guy and the guests in this episode with Steve Little and Mark Thompson from the "family history AI Show" podcast, I do recommend their podcast. It's very entertaining. It's informative. It keeps you updated on all things AI and the potential and what it could mean to genealogists everywhere, until the next episode, happy and productive research.
Jingle 31:51
And remember to subscribe to the podcast to be informed when new additions are published.
Creators and Guests


