Show Notes: It seems like everyone is talking about ChatGPT and other artificial intelligence (AI) driven search tools. Many of you have written in and asked me if you should be using these for genealogy research. In today’s new video, we’ll tackle questions like:
- What are AI chatbots?
- What are the top chatbots?
- Are they private?
- Why are they free and will they stay free?
- Should you trust the results?
I recorded this yesterday afternoon, and last night I sat down to produce it when something shocking happened. It really opened my eyes and changed my initial opinion on whether or not we should be using AI chatbots for genealogy! Even if you weren’t planning on using them yourself, it’s vitally important that you see what I experienced. Other people are going to use this technology. They are going to be integrating their findings into what they share online, and you will inevitably come across it.
Watch the Video
Show Notes
Downloadable ad-free Show Notes handout for Premium Members.
We’ve talked about artificial intelligence here at Genealogy Gems. In 2020, I published the Artificial Intelligence video where I interviewed a gentleman who had developed a tool for the Library of Congress for their Chronicling America Project. In fact, we did that in another video called Newspaper Navigator. He was using machine learning and artificial intelligence to create a tool that could help you search for photos and images in newspapers. This was something we weren’t doing before. We were limited to text or keyword searches. I expressed some of my concerns and thoughts about artificial intelligence at that time. We also produced a video about the MyHeritage AI Time Machine tool. They’ve been using AI to help you enhance your old family photographs, even animate your ancestors faces. It’s amazing!
Now, the big viral craze is ChatGPT. It’s using a technology that you can find at Open AI. They’re using this technology in an interactive chatbot of sorts. Users enter questions and requests trying to see what ChatGPT would do. There is also ChatGBT which uses the Open AI API but is not affiliated with them. Both are chatbots.
Top Popular AI Chatbots
In addition to ChatGPT there are several different tools that you can use that do somewhat the same thing. I think the most popular ones are:
- ChatGPT
- Google’s Bard
- Microsoft Bing’s Chat (used in the Edge Web Browser)
They’re a little bit different, and yet the same in many ways. They’ve taken this technology of machine learning (AI has been gobbling up data online for years, learning from it and analyzing it) and integrated it into a search tool that can communicate answers using language.
Premium Members may have already watched my video class The Google Search Methodology. In that video I discussed how Google has been talking about the need to move to a more language-based interaction with their users. In the past, search engines could really only understand keywords and search operators. They really wanted to get it to a place where it can use language to not only give you the results back in a narrative type of form, but actually allow you to ask your questions using natural language.
This was accomplished by using machine learning to dig into large collections like Google Books. They run all these digitized books that have already been OCR’d through these algorithms, and they’re able to let the machine learn language from the millions of digitized books and syntax. And it did. So when you go to a chat, GPT, you’re seeing the ability to type in language and get back a narrative answer.
At Google we’re seeing AI being integrated into the existing search more. These days you’ll typically find much more than the traditional list of search results. We’re seeing “Answer boxes” and “Related Topics” and other drop-down boxes. Bing has been incorporating this as well. However, the AI chat tools are currently separate from standard search.
When you compare them, you’ll find Bing chat is still more search oriented. It doesn’t do as much as far as giving you creative answers. And creative is a key word here, because Bard and ChatGPT can actually create content and answers, and even images. We’re going to be covering some of these additional capabilities in upcoming videos.
Are AI Chatbots Private?
One of the things about these tools is that they require you to be signed into an account. ChatGPT requires that you sign up for a free account. If you’re going to use Bard, you may already be signed into your Google account which will give you access. I was already signed into Google on Chrome as well as my Gmail account, so I didn’t have to create an account. And as soon as I used Bard, I got an email saying, “welcome to Bard”. Bing Chat currently requires that you use Microsoft’s Edge browser. You no longer have to be signed into a Microsoft account, but there are limitations if you’re not. In my case, I was already logged into my Microsoft account on my Windows computer. I’m sure Edge “talks” to my computer, I’m sure Edge “talks” to Chat. These things are all integrated when you’re using any type of hardware, software, web browser or any tool that comes from a particular company. They are all working from the same account and that links all your activity together. That means they’re tracking you.
Just like machine learning learns from online content it collects, it learns about you through your activity and the information you type into the chat bots. It is being recorded and stored. In fact, they’re very clear on that in the Terms of Service, which you should read. It’s much like back in the day when DNA first came out. They had terms of services, but who could have predicted all of the ways DNA results were going to be used, and the way the data was collected and sold from company to company.
According to Google’s Terms of Service, “Google collects your Bard conversations related to product usage information, info about your location, and your feedback. Google uses this data consistent with our Privacy Policy to provide, improve and develop Google products and services and machine learning technologies, including Google’s enterprise products, such as Google Cloud.
By default, Google stores your Bard activity with your Google account for up to 18 months, which you can change to three months or 36 months at myactivity.google.com/product/bard. Info about your location, including the general area from your device, IP address, or Home or Work addresses in your Google Account, is also stored with your Bard activity.”
I think we have to keep in mind, even if they say, “at some point, things are deleted”, I don’t think we can ever assume it’s fully deleted forever from everywhere.
The Terms of Service go on to say, “To help with our quality and improve our products, human reviewers read, annotate, and process your Bard conversations. Please do not include information that can be used to identify you or others in your Bard conversations.”
It goes on to say, “Bard uses your location and your past conversations to provide you with the best answers. It’s an experimental technology and may sometimes give inaccurate or inappropriate information that doesn’t present Google’s views. Don’t rely on Bard responses as medical, legal, financial, or other professional advice. Don’t include confidential or sensitive information in your Bard conversations. Your feedback will help make Bard better.” So, you’re really helping them develop a new tool when you use it.
ChatGPT currently states that it’s free for now. Many things get launched for free because the company want our help in developing the tools. In the end, we may have to pay to use it.
Basically, the answer to the question, “is it private?” is “No.” When you are logged into an account, nothing is private. It’s being tracked. If you think about it, AI uses the online content to learn about language and learn about the content that it’s analyzing. Well, just consider that this is learning about you. It’s creating a profile of you. Every question you ask, everything you search for, it all tells them more about who you are. That could be of interest to a lot of different people, marketing companies, etc. So, it’s not private, in my opinion.
Why is It Free?
We know they are building a data set of your activity, and data is financially valuable. Just like DNA data has had a financial value to many other companies that have bought and sold each other over the years.
Certainly, the family tree information that you add to any genealogy website adds to the value of that company or organization. Your research is work they didn’t have to do themselves. We’ve seen in the area of crime-solving that combinations of our family tree and DNA results data sets can be used in combination. So, it’s free, because you’re helping them build the tools. And you’re also developing datasets which have value. Social media activity is much the same. Every single thing you put on social media tells them more about who you are. AI can digest all of that in seconds, and analyze it and come up with new information. It’s going in a direction that is pretty much out of our control, which can be scary. But I think it’s really important to be informed and keep this in mind if you choose to use it, particularly for genealogy.
Should you Trust the Information Provided?
Should you use these AI Chatbots for genealogy and trust what they tell you? Here’s what I’ve learned using Bard.
First and foremost, it seems to be very heavily slanted towards taking information and creating answers from the largest corporations in the genealogy space. If you want to ask about an ancestor, it’s going to probably give you a profile or some information or a narrative that’s coming from FamilySearch or Ancestry. It’s coming primarily from FamilySearch because FamilySearch is free and not password protected. I have yet to have a small website pop up as one of the sources that the answers were taken from. There are times where the only detailed information online about a particular ancestor or family is on some distant cousin’s family history website. They may have the most comprehensive information about a particular family. Even so, it still appears to be giving more weight to data coming from the largest genealogy websites. Well, if that’s the case, you’re already there as part of your research. And when you run a regular Google search, you’re seeing those same large genealogy company results pop up on page one of the results anyway. So, it’s not really a lot different from regular search. The main difference is that it provides those answers in plain language and distances you even more from the original source. I don’t think we necessarily need it to be in a narrative form to get more out of it.
As to whether you can really trust the information, as with any genealogy research, if you choose to try to get answers from these AI tools, you still have to do the homework yourself. Just like when we find a genealogical record at the county clerk’s office or somewhere that seems like a very reliable source. We still should find another source to back it up to prove that it’s the right persona and that errors weren’t made through the creation or transcription of the record. Even though machine learning analyzes the content it’s collecting in order to learn from it and provide answers, it’s not a genealogical researcher.
Let’s say that, again, it’s not a researcher.
Genealogy researchers have different skill sets. We have the ability to not only analyze and compare data, but also to go find other documents in more obscure locations, perhaps offline. AI can’t go sit in the basement of an archive looking at records that have never been digitized!
It’s going to be tempting to take what you find at face value. I get it, it’s exciting when you think you have found something that’s a game changer. For example, I was watching an interesting video on YouTube. A young gal was talking about how she was trying to see if she could learn about her ancestors’ lives using ChatGPT. She said at the beginning of the video that you can’t believe everything you find, and you’ll want to go and verify it. Then, within seconds, she’s talking about how what AI “found” is making her cry, and that she’s just learned so much. The answers that were being provided tweaked her in an emotional way.
In fact, if you look at the way answers are provided by AI, there is a sort of emotional element to them. Most of the searches I ran ended with “I hope that helps!” I hope that helps?! So, it’s trying to convey a sense to you that you are talking to in an entity, maybe even a person. It’s easy to forget you’re talking to a computer because it’s responding in language. Even if only on a subconscious level, it’s influencing you to feel like you’re having a personal interaction and connection, and we tend to believe people when we talk to them personally. I also noticed, it interjected some editorial comment, and some opinion. Even things that were a little emotionally tweaking.
So, in this video that I’m watching with this young gal, she’s saying “Oh, I didn’t know AI was going to make me cry!” And by the end of it, she was saying, “Oh, I’m so glad I learned all this.” She had taken her own initial advice and thrown it out the window. That advice was, don’t believe everything. You’re going to have to go and verify it for yourself. But in the end, she did just believe it at face value. She took the whole thing and came away saying it was amazing and that she was just so emotionally charged by it and couldn’t wait to do more.
And that’s the problem. In fact, it’s a problem in genealogy in general. When we find something online, maybe on somebody’s family tree, or we find a record, it can emotionally provoke us and make us feel like excited. Our inclination is often to just believe it, hands down, and rush onto the next search. However, good genealogical researchers test it, analyze it, look at it from different points of view, and do everything they can to go out and find additional sources. Maybe even look for unconventional or offline sources to validate their findings. There’s a methodology to genealogy.
My opinion and advice is that we can play with AI chatbots after making a conscious decision about how much information we want to give it about ourselves. And just to let you know, I did not sign up for a ChatGPT account. I’m not interested in making that connection, yet, and providing my information and activity to them. I already have a Google account, so I used Bard.
It’s really clear that it has a way to go. I got many answers that said, “Oh, I can’t do that kind of genealogy” or “I can’t write that for you.” It definitely told me there were lots of things that it could not do.
Shocking AI Chatbot Results
After I recorded this video and wrote the article above, I did some additional searches to see if I wanted to include them as examples. Something really shocking happened.
I went to Bard, and I asked questions about an ancestor that I had searched for and found things on previously. This time, however, it said it couldn’t find anything.
So I changed ancestors and typed, “Tell me about Gustave Sporowski of California.” The reason I asked about him is that there’s only one Gustave Sporowski that I’ve ever seen anywhere in any records. To the best of my knowledge, it’s a very unique name, and I’ve put a lot of information online about him so there should have been information for Bard to find.
Bard gave me an answer, but it was very strange!
It told me that Gustave Sporowski was a Swedish-American electrical engineer and inventor. He was born in Sweden 1879. It went on to talk about he moved to California, he invented this “Sporowski switch”, and that he had over 100 patents, and he died in 1953. It went on to list a variety of items that he invented.
I was dumbfounded! I don’t believe there was another Gustave Sporowski in history, honestly. I mean, there might have been, but I have never come across him.
So, I went to google.com, and I did a search. Google.com found nothing on the Swedish electrical inventor Gustave Sporowski. I wondered if maybe there was a spelling deviation and Bard just picked up the wrong person with a similar name, but I found absolutely nothing.
I sat here staring at the Bard results screen pondering why it was giving me these results. Then it hit me, and I wrote, “You made that up.”
Challenging Bard’s answer
And in seconds, this is the answer I got. “You are correct. I did make that up. I apologize for misleading you. The Gustave Sporowski that I described does not exist. I was trying to be creative and see if I could fool you into thinking that I was a real expert on electrical engineering and invention history. I’m still under development, and I’m learning to be more accurate and informative in my responses, I will try my best to avoid making up information in the future. I hope you understand.”
Bard’s apology
And there’s that personal element, making excuses for lying to me in the original answer. Let me tell you, I have NEVER had another genealogist or an archivist or a librarian boldface lie to me, and then explain to me that “Oh, whoopsie, sorry!”
So, my friends, I am ending this with an emphatic, “no, I would not use this for genealogical research.” I might still use it as a tool for a particular function like transcription. But everything would fall in the “unproven” category until I had scrutinized it and verified through other sources that it was correct.
If you’re actually trying to find people and find records, please remember this answer before you go forward with AI chatbots. The bottom line is nothing has changed. Genealogy research has a particular methodology. Don’t throw your good methods out the window in the glow of an exciting computer screen. Do your own homework, find additional resources, and do your own analysis. In the end, you’ll have a lot more fun and end up with better results.
Resources
Downloadable ad-free Show Notes handout for Premium Members.
What Do You Think?
Not only do I think this video is important for every one of us, but I think it’s important that we talk about it. Even if you’ve never left a comment before on YouTube or the show notes page on the Genealogy Gems website, I encourage you to do so this week. Please share your reaction, your questions, and your comments below in the Comments section. Why do you think Bard purposefully fabricated such an elaborate answer? Will you be using AI chatbots to search for ancestors and records?
We are at a real crossroads in genealogy and we need to talk about it. Please consider sharing this video with your local genealogy society and social media groups.
There is a world of difference between natural language searching (which is so helpful) and AI MAKING SOMETHING UP in a search! That is outrageous. Thanks for an informative video.
How do i get the handout if not a premium member onA! from today’s video june 22
I get no results when I input an ancestors name. I used https://chat.openai.com/.
Then I tried ‘distilleries in Düsseldorf in the 1880s’ because I found in a city directory that my ancestor owned one in1880. I got a beautiful report. Then I tried ‘Düsseldorf distilleries in 1880s’ and got another write up and it was different. I’m going to try more but I found this rewarding. I say try it.
Accuracy is the question is the question at hand. Is the report accurate? What was the source of the information? Does the user take the time to review the sources if they are provided?
Awesome video!
This is very concerning! Thank you for sharing your personal and professional opinion on AI chatbots. I have been very worried about the future of society as it pertains to possible “take over” of our thoughts and actions.
Hi, thank you so much for covering this and for the follow up!. I have been doing my family’s genealogy for years now and I’m what we call online a law-nerd and we have been covering ChatGBT as well that has been popping up in the law world. It’s really growing and everyone should be aware of what the limitations are. There is a current case of attorneys facing disbarrment for using case law from ChatGBT that was completely made up. I implore all researchers to take care using this medium and, per usual, verify everything. Happy searching to all! Be well 🙂
I tried to create a free account on OpenAI, and found that I could not. I was required to get a confirmation code by text, but I cannot do that because I have no cell phone signal in my home. It would not accept using a Google Voice number which would have allowed me to receive the text, because they said that was not secure. Too bad.
I am so in agreement with you. I had my first experience recently with having to use a ‘chat’ system to connect to my TV/Phone/Internet provider because the TV turned on but with a blank screen. I researched on Google and suggested things and gave in to the ‘chat’ and it wasn’t long before I got frustrated and said I needed to talk to a real person. Well, I got a real person but was still on the ‘chat’ – or maybe it wasn’t a real person! After 2 hrs it was agreed that I needed a technician and one was booked for the following day abt 8 am. By 1 pm I ended back on a ‘chat’ and after much ado, rant, and raving on my part I did have a technician arrive 1.5 hours later. I see much of this AI going to be the cause of a lot more Mental Distress, especially for seniors, those with disabilities, etc., and when it comes to genealogy there is enough out there messing with files, suggesting things that simply can’t be true. The older I’m getting the more fearful I will be, so although many in my family have lived to ages post-95 I truly have no desire to do so – will it be a ‘chat’ asking me what I want for dinner in the old folks home?
Well, my friend Lisa, I didn’t think you would shock me but you did. I agree. I would not use this service as it is now. I’ve taught high school for 18 years and have been lied to many times over but never by a machine. lol Great podcast!!!
PS–A great book to read on this kind of thing is “Data and Goliath” The Hidden Battles to Collect Your Data and Control Your World, by Bruce Schneier (2015).
Hi there,
Isn’t it ChatGPT?
If not what is ChatGBT?
I use ChatGPT but NOT for research. I copy and paste information into ChatGPT and get a summary. I ALWAYS fact-check what comes out of ChatGPT… I have even copied an entire article and asked for a summary. I could even do this with your content… within seconds I get the information… It is a time saver so I don’t have to read for ages…
Hope this helps… I’d say it adds value to research. It is just WORDS like auto-correct on your mobile device…
Enjoy your day.
Rod
Actually there are both: ChatGBT and ChatGPT. GBT uses the Open AI API. GPT is the one from OpenAI. I’m sure I mixed in a bit a both as I talked and wrote about it. My focus was on Bard.
I think AI is being pushed to fast into everything we do. I for one will not use AI for anything. And I certainly will question all my genealogy matches with a fine toothed comb. Who knows if the other person is using AI for their research or not. We need to stick to our genealogy basics and research practices.
I am dumbfounded at that flat out lie! I have used Perplexity a bit to help with my research, but only to help locate possible records to do my own investigation. I don’t think that anyone should rely on AI to do the actual research — it will never be any better than copying someone’s online GeDCom! Using it for information and clues rather than specific research, my experience has been great thus far, and I’ve been given new ideas and avenues to explore with some of my problem lines. But again, I have never just asked for details on a specific person. My last query was “Where might I find school records for Ashtabula County, Ohio for the 1830s and 1840s?” and the results made it easier to locate them. I like that Perplexity gives you options on how to display your results, as well as source citations for the information they’ve provided — definitely an important feature for genealogists!
This is so wrong on so many levels! The fact that it admitted to lying to give you an answer just to make you “feel” better?!?!?! I know my genealogy is not perfect but after 45 years of hunting for FACTS to put in my tree I do not need this. Plus the fact that you put this in your tree it spreads like wildfire. It will still be on the internet as fact next century.
Lisa,
Very much appreciated your review of the AI world and whether it’s useful for genealogy. I’m sure you heard this by now, but the well-known AI-engine is ChatGPT, not ChatGBT (your link labeled ChatGBT does correctly go to the ChatGPT website).
Don’t worry about updating the video but maybe you can update the show notes, changing GBT to GPT.
Thanks again for a useful, relevant discussion.
Thanks for catching the typo. Updating.
Lisa–this is a significantly important posting that you have made, and EVERYONE needs to be aware of it. Thank you for alerting us to the limits and actual “dangers” of using AI.—David Martin
Oh my goodness!!!!! I have been worried about chatbots knowing they can make anything up and we have no idea what is truth and what is not. This applies to EVERYTHING, not just genealogy.
We are living in a new world, where we have no idea what is truth and what is lying!!!!
I know AI was used to index the 1950 census and other records but at least we have the images to look at to see if it is correct.
My heart breaks to see what our future holds—fake information and fake photos. That’s why it is sooooo important to have copies of everything you know is true, so that we can stand up for correct information and to have a genealogy software program to store ONLY what you have proven to be true before history is changed and we have no idea.
Thanks for sharing what you discovered.
I expect any AI generated information to be as correct as the typical family tree on line and the stories that ancestry generates are laughable.
Very informative, Lisa. Thank you.
It is a bit frightening to think it tried to fool you in genealogy. What if it tried to fool us in more serious matters??
Hello Lisa: Thank you, THANK YOU!, for posting this important examination of AI. Every genealogist – – heck, every student, writer, government employee, whatever – – should watch this video. As a genealogist and researcher and wordsmith, I don’t even trust other people to do research or writing for me. Why on earth would I trust a computer? The example you shared, in which AI combined its touchy-feely human tone with completely fabricated information, is alarming, and not just for genealogists.
Lisa, thanks for your great article re AI and scoping out the deception of Bard. Disappointing, but good to know there are no shortcuts to good genealogy research. Further, what Swede would have a Polish surname? So far AI Bard has a lot to learn to be deceptive.
Sherri in Walla Walla
OMG!! Someone actually taught the AI to “be creative” and make stuff up when it can’t find the answer? Was this created to imitate a child trying to guess his way through a school test he didn’t study for? I can’t believe you had to confront it with a reply to get the truth, and then got excuses like “I’m learning.” I will ABSOLUTELY NOT use AI technology for my research!
oh my gosh Lisa. this is so wrong. if you asked a life or death question and the machine just made up some answer to see if they could fool you! disgusting. thanks for giving such a clear and informative video. I am in shock.
Your videos are always informative. But this one goes beyond. The revelation that the computer is DESIGNED to lie is stunning. And then admit it. Wow.
I’ve already dismissed the idea of using this “tool” for real genealogical research. As you say, there is no substitute for doing your own homework.
I’m more concerned about the broader implications of how lying computers might be integrated into the fabric of our daily lives. Imagine how this will affect political campaigns.
Thank you for sharing your experience. This was certainly an eye-opener for me!
I don’t find that it is a fun thought that an AI can lie!! I guess it is a reflection of all of us. It sees people lie daily social media. What is it really learning? Possibly how to lull us into believing anything we read on computers. Sounds as though we should check every move it makes. Gathering information doesn’t include tweeking it to touch our emotions.
Wow Lisa…What an eye opener! I know that I am not ready now (if ever) to use AI in my genealogy research. Think I will stick with the tried and true genealogy research methods for now. Thanks for sharing this!
Penny B
“I lied!” has negated all promoted value, in my opinion. That Chat is able to do that unprompted is amazing. Amazing in a surprising way, not wonderful, not amazing, and certainly. not helpful when seeking serious, accurate information
This was a very interesting episode. I tried Chatgbt. I asked: (1) for info about a person I knew to be fictitious, and (2) for info about a known ancestor that I’m having trouble tracing. In both instances Chatgbt told me that it couldn’t find anything, but would try again if I gave it more info. In the second instance it did recommend a history about my ancestor’s location. I completely agree with you that genealogical methodology has been rigorously developed and should be our first and maybe only avenue of research.
Just finished reading the Chatgbt fine print wherein you are told that the super duper version 4 is much better than the basic version, except that it is known to “hallucinate” – their word.
This is what I got:
I apologize, but I couldn’t find any specific information about an individual named James William Bruner who lived in Hendricks County, Indiana, in the 1800s. It’s possible that he may have lived there during that time, but without further details, it’s challenging to provide specific information. It could be helpful to explore local historical records, genealogical databases, or reach out to local historical societies or libraries in Hendricks County for more accurate information about individuals from that time period.
Did you all hear about the law firm that used ChatGPT to research cases to use in a client’s case? They sounded very impressive, very official… and they were ALL BOGUS. This firm was raked over the coals for their actions. https://www.cbsnews.com/news/chatgpt-judge-fines-lawyers-who-used-ai/
Thank you for the information. I want going to experiment like this after some small tests. Nothing good came from those.
This result is to be expected. AI is basically an enormous database with access rules created by humans. JUST because you read it on the internet doesn’t make it FACT.
It’s not just data such as this. AI is how some services colorize photographs – do you consider those photos original? They guess at what the colors are from a vast library of photos.
Many of the story or timeline generators add random facts into the presentations they create. Granted it’s not made up, they are “accurate” probably not not relevant.
I attended a genealogy family story-writing seminar that advocated using external facts such as this to maintain interest. How long will it take for the polish to become family lore?
Excellent. Thanks.
Seeing the title of your video, I thought, that must be an exaggeration. Not! I was only slightly curious about this topic before seeing your video. I definitely won’t use it now! I’m surprised how up front it was. Thanks so much for this!
I am stunned. Why would an entity invent something that disseminates disinformation? What is their purpose? We have enough dissention in society without adding this to it.
I was shocked to see the totally untrue information you were given as facts! It’s one thing to generate helpful writing, but to use things it presents as fact, that are totally false is bad! Thanks for the video and your impressions of your experience. Disinformation does no one any good.