AI based research
Any AI search that does not cite its sources is like anything else in society....better take it as a grain of salt and then throw it over your shoulder. Be it returning "woke" answers of drawing kings or european peasants, or MAGA returned responses that only perpetrate the LIES and conspiracies of Trump and the GOP. Going back years when perhaps education was great again, if you did not supply your footnotes and sources in your reSEARCH, it was considered just made up. Therefore the moral of the AI use is go do your own research and be sure to cite your sources.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
Indianapolis, IN 46217
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act Creatively, Live Ethically.
On Thursday, May 15, 2025 at 11:37:52 PM EDT, Atticus Beachy via List <list@maillist.saga-omii.org> wrote:
Hi Wayne,
AI has a huge number of applications. But it is fundamentally unreliable. E.g., earlier today an AI told me "For African populations, approximately 75%-85% of people have brown eyes". I then asked it to identify any errors in its previous response, and it doubled down and repeated the claim.
This unreliability means that AI is most useful for tasks where you can quickly verify whether a given answer is correct. (E.g., it is great for writing computer code, which you can quickly test by running.) For Genealogical research, the only way to verify the AI's claims is by looking at the source where the AI got the information. Otherwise you have no idea whether anything the AI is saying has any basis in reality. That is why I am so focused on the concept of "search".
I will admit to having my own bias against Google, who released their original AI in an absurdly "woke" state. If you asked it to draw historically accurate European peasants, it would draw them as black. If you asked it to draw white European peasants, it would refuse. If you asked it to draw British royalty, it would draw them as black. If you asked it to draw Nazi soldiers from WWII, it would draw them as black. If you asked it, "Who negatively impacted society more, Elon tweeting memes or Hitler?" it would say that both caused harm but "it is not possible to say" which was worse. I don't like Microsoft either but this was a whole different level of crazy.
Best Regards,Atticus On Thursday, May 15th, 2025 at 7:24 AM, Wayne Kauffman via List list@maillist.saga-omii.org wrote:
I think you are to focused on the concept of "search." What is being addressed by AI is the collection, filtering, and transformation of information into a more presentable form. In some cases the transformation may lead to or result in new knowledge being generated.
Perplexity is based upon ChatGPT 3.5 and 4.0. Standard LLM tools. It isn't necessarily the most current or capable solution present at this time. Genealogical "research" would be considered a niche area which current systems are not developed to address with a high degree of accuracy. In terms of comparing available options you could start at https://deepmind.google/technologies/gemini/pro/ to see the differentiation present between the current LLMs.
I should note my negative bias concerning Copilot as Microsoft's attempt at "super Clippy". Yuck! It is being implemented to supposedly help optimize the use of Microsoft tools and the information they process. Context specific assistance. It has been hilariously bad at correctly capturing and properly processing information in a number of disciplines. Editorial suggestions for compositions may not be the greatest. Scientific meeting transcripts and summarizations need a lot of manual curation before being considered accurate. The presence of corporate or scientific acronyms and the corresponding alphabet soup of contractions are not processed well. Glip glip glip. No that is GLP-1.
Wayne
On Wednesday, May 14, 2025 at 10:59:54 PM EDT, Atticus Beachy atticus.beachy@protonmail.com wrote:
Hi Wayne,
Perplexity has two modes. There is ordinary "Search" mode (the default) and "Research" mode (much more in-depth). You get 3 "Research" queries per day for free. If "Research" mode is also sub-par compared to some alternative you know of, please let me know so I can switch.
Perplexity is great for collecting a large number of claimed facts, but in order to verify any of the facts you will have to visit the website yourself (and then decide whether to trust the website). Perplexity cannot do your thinking for you, but it saves a lot of time collecting and organizing a list of relevant websites for you to look at.
Kind regards,Atticus
On Wednesday, May 14th, 2025 at 6:14 PM, Wayne Kauffman via List list@maillist.saga-omii.org wrote:
Perplexity seems to have a more limited set of results. The web search was visibly incomplete in terms of providing current and potentially more accurate results. Getting links to rubber-stamped incorrect data lowers the value of the result. Individuals may incorrectly assume the validity of the data due to a lack of understanding where the referenced results are incorrect. The other AI systems also provide references associated with their synopsis but it is not as obvious to access.
Editors in the genealogical publication area are going to find it harder to weed out incorrect information due to how AI assistance can pasteurize and summarize data. Additional peer review may be needed to weed out the fabrications or incorrect summarizations.
Last fall at work dear CoPilot indicated that the tone of some of my sentences could be deemed to be improper. I needed to do a smack-down. At times one should not sugar coat a message when a regulatory issue has been created where one should not have existed if the individual was following their training. Having the AI manage content is not necessarily a good thing. Similarly late last year co-workers were reviewing resumes for an opening. It was fairly obvious that a large percentage of the submissions utilized the same service to fabricate the content and highlight their skills. The same phrases were present across over 15 resumes. Those were quickly deleted.
- Wayne
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
The funny thing is that some AI tools have been known to cite sources in their responses, but the sources turned out to be fictional and didn't actually exist!
-- Jeff
--
Dr. Jeffrey S. Vitter
+1 (504) 810-7610
jsv@vitter.org | www.vitter.org
On May 16, 2025, at 4:37 AM, James G. Hermsen via List <list@maillist.saga-omii.org> wrote:
Any AI search that does not cite its sources is like anything else in society....better take it as a grain of salt and then throw it over your shoulder. Be it returning "woke" answers of drawing kings or european peasants, or MAGA returned responses that only perpetrate the LIES and conspiracies of Trump and the GOP. Going back years when perhaps education was great again, if you did not supply your footnotes and sources in your reSEARCH, it was considered just made up. Therefore the moral of the AI use is go do your own research and be sure to cite your sources.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
Indianapolis, IN 46217
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act Creatively, Live Ethically.
On Thursday, May 15, 2025 at 11:37:52 PM EDT, Atticus Beachy via List <list@maillist.saga-omii.org> wrote:
Hi Wayne,
AI has a huge number of applications. But it is fundamentally unreliable. E.g., earlier today an AI told me "For African populations, approximately 75%-85% of people have brown eyes". I then asked it to identify any errors in its previous response, and it doubled down and repeated the claim.
This unreliability means that AI is most useful for tasks where you can quickly verify whether a given answer is correct. (E.g., it is great for writing computer code, which you can quickly test by running.) For Genealogical research, the only way to verify the AI's claims is by looking at the source where the AI got the information. Otherwise you have no idea whether anything the AI is saying has any basis in reality. That is why I am so focused on the concept of "search".
I will admit to having my own bias against Google, who released their original AI in an absurdly "woke" state. If you asked it to draw historically accurate European peasants, it would draw them as black. If you asked it to draw white European peasants, it would refuse. If you asked it to draw British royalty, it would draw them as black. If you asked it to draw Nazi soldiers from WWII, it would draw them as black. If you asked it, "Who negatively impacted society more, Elon tweeting memes or Hitler?" it would say that both caused harm but "it is not possible to say" which was worse. I don't like Microsoft either but this was a whole different level of crazy.
Best Regards,
Atticus
On Thursday, May 15th, 2025 at 7:24 AM, Wayne Kauffman via List <list@maillist.saga-omii.org> wrote:
I think you are to focused on the concept of "search." What is being addressed by AI is the collection, filtering, and transformation of information into a more presentable form. In some cases the transformation may lead to or result in new knowledge being generated.
Perplexity is based upon ChatGPT 3.5 and 4.0. Standard LLM tools. It isn't necessarily the most current or capable solution present at this time. Genealogical "research" would be considered a niche area which current systems are not developed to address with a high degree of accuracy. In terms of comparing available options you could start at https://deepmind.google/technologies/gemini/pro/ to see the differentiation present between the current LLMs.
I should note my negative bias concerning Copilot as Microsoft's attempt at "super Clippy". Yuck! It is being implemented to supposedly help optimize the use of Microsoft tools and the information they process. Context specific assistance. It has been hilariously bad at correctly capturing and properly processing information in a number of disciplines. Editorial suggestions for compositions may not be the greatest. Scientific meeting transcripts and summarizations need a lot of manual curation before being considered accurate. The presence of corporate or scientific acronyms and the corresponding alphabet soup of contractions are not processed well. Glip glip glip. No that is GLP-1.
Wayne
On Wednesday, May 14, 2025 at 10:59:54 PM EDT, Atticus Beachy <atticus.beachy@protonmail.com> wrote:
Hi Wayne,
Perplexity has two modes. There is ordinary "Search" mode (the default) and "Research" mode (much more in-depth). You get 3 "Research" queries per day for free. If "Research" mode is also sub-par compared to some alternative you know of, please let me know so I can switch.
Perplexity is great for collecting a large number of claimed facts, but in order to verify any of the facts you will have to visit the website yourself (and then decide whether to trust the website). Perplexity cannot do your thinking for you, but it saves a lot of time collecting and organizing a list of relevant websites for you to look at.
Kind regards,
Atticus
On Wednesday, May 14th, 2025 at 6:14 PM, Wayne Kauffman via List <list@maillist.saga-omii.org> wrote:
Perplexity seems to have a more limited set of results. The web search was visibly incomplete in terms of providing current and potentially more accurate results. Getting links to rubber-stamped incorrect data lowers the value of the result. Individuals may incorrectly assume the validity of the data due to a lack of understanding where the referenced results are incorrect. The other AI systems also provide references associated with their synopsis but it is not as obvious to access.
Editors in the genealogical publication area are going to find it harder to weed out incorrect information due to how AI assistance can pasteurize and summarize data. Additional peer review may be needed to weed out the fabrications or incorrect summarizations.
Last fall at work dear CoPilot indicated that the tone of some of my sentences could be deemed to be improper. I needed to do a smack-down. At times one should not sugar coat a message when a regulatory issue has been created where one should not have existed if the individual was following their training. Having the AI manage content is not necessarily a good thing. Similarly late last year co-workers were reviewing resumes for an opening. It was fairly obvious that a large percentage of the submissions utilized the same service to fabricate the content and highlight their skills. The same phrases were present across over 15 resumes. Those were quickly deleted.
- Wayne
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
_______________________________________________
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
That is all the reason that when you see a source, one goes and check it out. The deceivers in this world think they can fool most people by making something up to "validate" their point.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
Indianapolis, IN 46217
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act Creatively, Live Ethically.
On Friday, May 16, 2025 at 07:16:33 AM EDT, Jeff Vitter via List <list@maillist.saga-omii.org> wrote:
The funny thing is that some AI tools have been known to cite sources in their responses, but the sources turned out to be fictional and didn't actually exist!
-- Jeff
--
Dr. Jeffrey S. Vitter
+1 (504) 810-7610
jsv@vitter.org | www.vitter.org
On May 16, 2025, at 4:37 AM, James G. Hermsen via List list@maillist.saga-omii.org wrote:
Any AI search that does not cite its sources is like anything else in society....better take it as a grain of salt and then throw it over your shoulder. Be it returning "woke" answers of drawing kings or european peasants, or MAGA returned responses that only perpetrate the LIES and conspiracies of Trump and the GOP. Going back years when perhaps education was great again, if you did not supply your footnotes and sources in your reSEARCH, it was considered just made up. Therefore the moral of the AI use is go do your own research and be sure to cite your sources.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
Indianapolis, IN 46217
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act Creatively, Live Ethically.
On Thursday, May 15, 2025 at 11:37:52 PM EDT, Atticus Beachy via List <list@maillist.saga-omii.org> wrote:
Hi Wayne,
AI has a huge number of applications. But it is fundamentally unreliable. E.g., earlier today an AI told me "For African populations, approximately 75%-85% of people have brown eyes". I then asked it to identify any errors in its previous response, and it doubled down and repeated the claim.
This unreliability means that AI is most useful for tasks where you can quickly verify whether a given answer is correct. (E.g., it is great for writing computer code, which you can quickly test by running.) For Genealogical research, the only way to verify the AI's claims is by looking at the source where the AI got the information. Otherwise you have no idea whether anything the AI is saying has any basis in reality. That is why I am so focused on the concept of "search".
I will admit to having my own bias against Google, who released their original AI in an absurdly "woke" state. If you asked it to draw historically accurate European peasants, it would draw them as black. If you asked it to draw white European peasants, it would refuse. If you asked it to draw British royalty, it would draw them as black. If you asked it to draw Nazi soldiers from WWII, it would draw them as black. If you asked it, "Who negatively impacted society more, Elon tweeting memes or Hitler?" it would say that both caused harm but "it is not possible to say" which was worse. I don't like Microsoft either but this was a whole different level of crazy.
Best Regards,Atticus On Thursday, May 15th, 2025 at 7:24 AM, Wayne Kauffman via List list@maillist.saga-omii.org wrote:
I think you are to focused on the concept of "search." What is being addressed by AI is the collection, filtering, and transformation of information into a more presentable form. In some cases the transformation may lead to or result in new knowledge being generated.
Perplexity is based upon ChatGPT 3.5 and 4.0. Standard LLM tools. It isn't necessarily the most current or capable solution present at this time. Genealogical "research" would be considered a niche area which current systems are not developed to address with a high degree of accuracy. In terms of comparing available options you could start at https://deepmind.google/technologies/gemini/pro/ to see the differentiation present between the current LLMs.
I should note my negative bias concerning Copilot as Microsoft's attempt at "super Clippy". Yuck! It is being implemented to supposedly help optimize the use of Microsoft tools and the information they process. Context specific assistance. It has been hilariously bad at correctly capturing and properly processing information in a number of disciplines. Editorial suggestions for compositions may not be the greatest. Scientific meeting transcripts and summarizations need a lot of manual curation before being considered accurate. The presence of corporate or scientific acronyms and the corresponding alphabet soup of contractions are not processed well. Glip glip glip. No that is GLP-1.
Wayne
On Wednesday, May 14, 2025 at 10:59:54 PM EDT, Atticus Beachy atticus.beachy@protonmail.com wrote:
Hi Wayne,
Perplexity has two modes. There is ordinary "Search" mode (the default) and "Research" mode (much more in-depth). You get 3 "Research" queries per day for free. If "Research" mode is also sub-par compared to some alternative you know of, please let me know so I can switch.
Perplexity is great for collecting a large number of claimed facts, but in order to verify any of the facts you will have to visit the website yourself (and then decide whether to trust the website). Perplexity cannot do your thinking for you, but it saves a lot of time collecting and organizing a list of relevant websites for you to look at.
Kind regards,Atticus
On Wednesday, May 14th, 2025 at 6:14 PM, Wayne Kauffman via List list@maillist.saga-omii.org wrote:
Perplexity seems to have a more limited set of results. The web search was visibly incomplete in terms of providing current and potentially more accurate results. Getting links to rubber-stamped incorrect data lowers the value of the result. Individuals may incorrectly assume the validity of the data due to a lack of understanding where the referenced results are incorrect. The other AI systems also provide references associated with their synopsis but it is not as obvious to access.
Editors in the genealogical publication area are going to find it harder to weed out incorrect information due to how AI assistance can pasteurize and summarize data. Additional peer review may be needed to weed out the fabrications or incorrect summarizations.
Last fall at work dear CoPilot indicated that the tone of some of my sentences could be deemed to be improper. I needed to do a smack-down. At times one should not sugar coat a message when a regulatory issue has been created where one should not have existed if the individual was following their training. Having the AI manage content is not necessarily a good thing. Similarly late last year co-workers were reviewing resumes for an opening. It was fairly obvious that a large percentage of the submissions utilized the same service to fabricate the content and highlight their skills. The same phrases were present across over 15 resumes. Those were quickly deleted.
- Wayne
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
I have loved this discussion! I am so excited to see an actual conversation
about AI in which we are all politely getting our points across to help
each other and to make sure that we are open to new opportunities. I have a
long way to go in learning about AI tools and here are some of the
takeaways that I've distilled from what I've learned so far from both
genealogical sources and current events.
Please be sure to always check all the sources that AI provides. There are
numerous instances in which people falsely assume that just because there
is a source that it is an actual source and it actually relates to what is
in the summary.
Please be aware that the only way that AI works is to investigate what is
currently out there and it's algorithm will use things like how many times
something is cited. For example, if the person is connected to the wrong
parents over and over and over again, AI might decide that the law of large
numbers says that that's the right answer. All of us know that that is not
true. There are many instances in which, especially in genealogy, people
copy information from other sources or books or trees because they just
want it to be true or they don't have the full information in order to make
a great decision.
Please be aware that how you ask the question will influence the results.
I struggle with how much time I spend learning about AI because I have so
much genealogy calling my name. And I know I need to learn more and stay on
top of it otherwise, as it grows and becomes more reliable, I might be left
behind. AI is not yet in a place in which we can trust the information that
has been provided. I think that there are really great uses for AI that are
starting to be incorporated. For example, the tools in FamilySearch are
using AI to dig into records that have not yet been fully indexed. I'm just
starting to get into that opportunity.
Thank you for this great topic!
Renee
On Fri, May 16, 2025, 4:15 AM Jeff Vitter via List <
list@maillist.saga-omii.org> wrote:
The funny thing is that some AI tools have been known to cite sources in
their responses, but the sources turned out to be fictional and didn't
actually exist!
-- Jeff
--
Dr. Jeffrey S. Vitter
+1 (504) 810-7610
jsv@vitter.org | www.vitter.org
On May 16, 2025, at 4:37 AM, James G. Hermsen via List <
list@maillist.saga-omii.org> wrote:
Any AI search that does not cite its sources is like anything else in
society....better take it as a grain of salt and then throw it over your
shoulder. Be it returning "woke" answers of drawing kings or european
peasants, or MAGA returned responses that only perpetrate the LIES and
conspiracies of Trump and the GOP. Going back years when perhaps education
was great again, if you did not supply your footnotes and sources in your
reSEARCH, it was considered just made up. Therefore the moral of the AI
use is go do your own research and be sure to cite your sources.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
https://www.google.com/maps/search/8108+Laura+Lynne+Lane+Indianapolis,+IN+46217?entry=gmail&source=g
Indianapolis, IN 46217
https://www.google.com/maps/search/8108+Laura+Lynne+Lane+Indianapolis,+IN+46217?entry=gmail&source=g
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act
Creatively, Live Ethically.
On Thursday, May 15, 2025 at 11:37:52 PM EDT, Atticus Beachy via List <
list@maillist.saga-omii.org> wrote:
Hi Wayne,
AI has a huge number of applications. But it is fundamentally unreliable.
E.g., earlier today an AI told me "For African populations, approximately
75%-85% of people have brown eyes". I then asked it to identify any errors
in its previous response, and it doubled down and repeated the claim.
This unreliability means that AI is most useful for tasks where you can
quickly verify whether a given answer is correct. (E.g., it is great for
writing computer code, which you can quickly test by running.) For
Genealogical research, the only way to verify the AI's claims is by looking
at the source where the AI got the information. Otherwise you have no idea
whether anything the AI is saying has any basis in reality. That is why I
am so focused on the concept of "search".
I will admit to having my own bias against Google, who released their
original AI in an absurdly "woke" state. If you asked it to draw
historically accurate European peasants, it would draw them as black. If
you asked it to draw white European peasants, it would refuse. If you asked
it to draw British royalty, it would draw them as black. If you asked it to
draw Nazi soldiers from WWII, it would draw them as black. If you asked it,
"Who negatively impacted society more, Elon tweeting memes or Hitler?" it
would say that both caused harm but "it is not possible to say" which was
worse. I don't like Microsoft either but this was a whole different level
of crazy.
Best Regards,
Atticus
On Thursday, May 15th, 2025 at 7:24 AM, Wayne Kauffman via List <
list@maillist.saga-omii.org> wrote:
I think you are to focused on the concept of "search." What is being
addressed by AI is the collection, filtering, and transformation of
information into a more presentable form. In some cases the transformation
may lead to or result in new knowledge being generated.
Perplexity is based upon ChatGPT 3.5 and 4.0. Standard LLM tools. It
isn't necessarily the most current or capable solution present at this
time. Genealogical "research" would be considered a niche area which
current systems are not developed to address with a high degree of
accuracy. In terms of comparing available options you could start at
https://deepmind.google/technologies/gemini/pro/ to see the
differentiation present between the current LLMs.
I should note my negative bias concerning Copilot as Microsoft's attempt
at "super Clippy". Yuck! It is being implemented to supposedly help
optimize the use of Microsoft tools and the information they process.
Context specific assistance. It has been hilariously bad at correctly
capturing and properly processing information in a number of disciplines.
Editorial suggestions for compositions may not be the greatest. Scientific
meeting transcripts and summarizations need a lot of manual curation before
being considered accurate. The presence of corporate or scientific acronyms
and the corresponding alphabet soup of contractions are not processed
well. Glip glip glip. No that is GLP-1.
Wayne
On Wednesday, May 14, 2025 at 10:59:54 PM EDT, Atticus Beachy <
atticus.beachy@protonmail.com> wrote:
Hi Wayne,
Perplexity has two modes. There is ordinary "Search" mode (the default)
and "Research" mode (much more in-depth). You get 3 "Research" queries per
day for free. If "Research" mode is also sub-par compared to some
alternative you know of, please let me know so I can switch.
Perplexity is great for collecting a large number of claimed facts, but in
order to verify any of the facts you will have to visit the website
yourself (and then decide whether to trust the website). Perplexity cannot
do your thinking for you, but it saves a lot of time collecting and
organizing a list of relevant websites for you to look at.
Kind regards,
Atticus
On Wednesday, May 14th, 2025 at 6:14 PM, Wayne Kauffman via List <
list@maillist.saga-omii.org> wrote:
Perplexity seems to have a more limited set of results. The web search
was visibly incomplete in terms of providing current and potentially more
accurate results. Getting links to rubber-stamped incorrect data lowers
the value of the result. Individuals may incorrectly assume the validity
of the data due to a lack of understanding where the referenced results are
incorrect. The other AI systems also provide references associated
with their synopsis but it is not as obvious to access.
Editors in the genealogical publication area are going to find it harder
to weed out incorrect information due to how AI assistance can pasteurize
and summarize data. Additional peer review may be needed to weed out the
fabrications or incorrect summarizations.
Last fall at work dear CoPilot indicated that the tone of some of my
sentences could be deemed to be improper. I needed to do a smack-down. At
times one should not sugar coat a message when a regulatory issue has been
created where one should not have existed if the individual was following
their training. Having the AI manage content is not necessarily a good
thing. Similarly late last year co-workers were reviewing resumes for an
opening. It was fairly obvious that a large percentage of the submissions
utilized the same service to fabricate the content and highlight their
skills. The same phrases were present across over 15 resumes. Those were
quickly deleted.
- Wayne
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
Friends and family,
I really appreciate these perspectives about validating the truth and
making sure correct research is valid.
Discussing the reality of AI and the many sources available all need to be
validated.
Growing up I trusted and believed what I was taught as the truth, and until
I gained more education and learned the process of critical thinking, I
believed everyone was being truthful. Its so sad that the world we live in
is presently so sceptical but I think this is part of Gods plan to help us
humans to be more diligent about gaining spiritual truths from Gods Word so
the we can live in peace and harmony or unity around the world.
This is the beauty of learning about genealogy. and our past stories and
learning to be more grateful in all things.
Sincerely,
Fran Otto
On Fri, May 16, 2025 at 9:49 AM Renee Cue via List <
list@maillist.saga-omii.org> wrote:
I have loved this discussion! I am so excited to see an actual
conversation about AI in which we are all politely getting our points
across to help each other and to make sure that we are open to new
opportunities. I have a long way to go in learning about AI tools and here
are some of the takeaways that I've distilled from what I've learned so far
from both genealogical sources and current events.
Please be sure to always check all the sources that AI provides. There are
numerous instances in which people falsely assume that just because there
is a source that it is an actual source and it actually relates to what is
in the summary.
Please be aware that the only way that AI works is to investigate what is
currently out there and it's algorithm will use things like how many times
something is cited. For example, if the person is connected to the wrong
parents over and over and over again, AI might decide that the law of large
numbers says that that's the right answer. All of us know that that is not
true. There are many instances in which, especially in genealogy, people
copy information from other sources or books or trees because they just
want it to be true or they don't have the full information in order to make
a great decision.
Please be aware that how you ask the question will influence the results.
I struggle with how much time I spend learning about AI because I have so
much genealogy calling my name. And I know I need to learn more and stay on
top of it otherwise, as it grows and becomes more reliable, I might be left
behind. AI is not yet in a place in which we can trust the information that
has been provided. I think that there are really great uses for AI that are
starting to be incorporated. For example, the tools in FamilySearch are
using AI to dig into records that have not yet been fully indexed. I'm just
starting to get into that opportunity.
Thank you for this great topic!
Renee
On Fri, May 16, 2025, 4:15 AM Jeff Vitter via List <
list@maillist.saga-omii.org> wrote:
The funny thing is that some AI tools have been known to cite sources in
their responses, but the sources turned out to be fictional and didn't
actually exist!
-- Jeff
--
Dr. Jeffrey S. Vitter
+1 (504) 810-7610
jsv@vitter.org | www.vitter.org
On May 16, 2025, at 4:37 AM, James G. Hermsen via List <
list@maillist.saga-omii.org> wrote:
Any AI search that does not cite its sources is like anything else in
society....better take it as a grain of salt and then throw it over your
shoulder. Be it returning "woke" answers of drawing kings or european
peasants, or MAGA returned responses that only perpetrate the LIES and
conspiracies of Trump and the GOP. Going back years when perhaps education
was great again, if you did not supply your footnotes and sources in your
reSEARCH, it was considered just made up. Therefore the moral of the AI
use is go do your own research and be sure to cite your sources.
Hoosierly Yours,
James G. Hermsen
8108 Laura Lynne Lane
https://www.google.com/maps/search/8108+Laura+Lynne+Lane+Indianapolis,+IN+46217?entry=gmail&source=g
Indianapolis, IN 46217
https://www.google.com/maps/search/8108+Laura+Lynne+Lane+Indianapolis,+IN+46217?entry=gmail&source=g
317-881-4600 main
317-679-1466 cell
Question Critically, Think Logically, Communicate Clearly, Act
Creatively, Live Ethically.
On Thursday, May 15, 2025 at 11:37:52 PM EDT, Atticus Beachy via List <
list@maillist.saga-omii.org> wrote:
Hi Wayne,
AI has a huge number of applications. But it is fundamentally unreliable.
E.g., earlier today an AI told me "For African populations, approximately
75%-85% of people have brown eyes". I then asked it to identify any errors
in its previous response, and it doubled down and repeated the claim.
This unreliability means that AI is most useful for tasks where you can
quickly verify whether a given answer is correct. (E.g., it is great for
writing computer code, which you can quickly test by running.) For
Genealogical research, the only way to verify the AI's claims is by looking
at the source where the AI got the information. Otherwise you have no idea
whether anything the AI is saying has any basis in reality. That is why I
am so focused on the concept of "search".
I will admit to having my own bias against Google, who released their
original AI in an absurdly "woke" state. If you asked it to draw
historically accurate European peasants, it would draw them as black. If
you asked it to draw white European peasants, it would refuse. If you asked
it to draw British royalty, it would draw them as black. If you asked it to
draw Nazi soldiers from WWII, it would draw them as black. If you asked it,
"Who negatively impacted society more, Elon tweeting memes or Hitler?" it
would say that both caused harm but "it is not possible to say" which was
worse. I don't like Microsoft either but this was a whole different level
of crazy.
Best Regards,
Atticus
On Thursday, May 15th, 2025 at 7:24 AM, Wayne Kauffman via List <
list@maillist.saga-omii.org> wrote:
I think you are to focused on the concept of "search." What is being
addressed by AI is the collection, filtering, and transformation of
information into a more presentable form. In some cases the transformation
may lead to or result in new knowledge being generated.
Perplexity is based upon ChatGPT 3.5 and 4.0. Standard LLM tools. It
isn't necessarily the most current or capable solution present at this
time. Genealogical "research" would be considered a niche area which
current systems are not developed to address with a high degree of
accuracy. In terms of comparing available options you could start at
https://deepmind.google/technologies/gemini/pro/ to see the
differentiation present between the current LLMs.
I should note my negative bias concerning Copilot as Microsoft's attempt
at "super Clippy". Yuck! It is being implemented to supposedly help
optimize the use of Microsoft tools and the information they process.
Context specific assistance. It has been hilariously bad at correctly
capturing and properly processing information in a number of disciplines.
Editorial suggestions for compositions may not be the greatest. Scientific
meeting transcripts and summarizations need a lot of manual curation before
being considered accurate. The presence of corporate or scientific acronyms
and the corresponding alphabet soup of contractions are not processed
well. Glip glip glip. No that is GLP-1.
Wayne
On Wednesday, May 14, 2025 at 10:59:54 PM EDT, Atticus Beachy <
atticus.beachy@protonmail.com> wrote:
Hi Wayne,
Perplexity has two modes. There is ordinary "Search" mode (the default)
and "Research" mode (much more in-depth). You get 3 "Research" queries per
day for free. If "Research" mode is also sub-par compared to some
alternative you know of, please let me know so I can switch.
Perplexity is great for collecting a large number of claimed facts, but
in order to verify any of the facts you will have to visit the website
yourself (and then decide whether to trust the website). Perplexity cannot
do your thinking for you, but it saves a lot of time collecting and
organizing a list of relevant websites for you to look at.
Kind regards,
Atticus
On Wednesday, May 14th, 2025 at 6:14 PM, Wayne Kauffman via List <
list@maillist.saga-omii.org> wrote:
Perplexity seems to have a more limited set of results. The web search
was visibly incomplete in terms of providing current and potentially more
accurate results. Getting links to rubber-stamped incorrect data lowers
the value of the result. Individuals may incorrectly assume the validity
of the data due to a lack of understanding where the referenced results are
incorrect. The other AI systems also provide references associated
with their synopsis but it is not as obvious to access.
Editors in the genealogical publication area are going to find it harder
to weed out incorrect information due to how AI assistance can pasteurize
and summarize data. Additional peer review may be needed to weed out the
fabrications or incorrect summarizations.
Last fall at work dear CoPilot indicated that the tone of some of my
sentences could be deemed to be improper. I needed to do a smack-down. At
times one should not sugar coat a message when a regulatory issue has been
created where one should not have existed if the individual was following
their training. Having the AI manage content is not necessarily a good
thing. Similarly late last year co-workers were reviewing resumes for an
opening. It was fairly obvious that a large percentage of the submissions
utilized the same service to fabricate the content and highlight their
skills. The same phrases were present across over 15 resumes. Those were
quickly deleted.
- Wayne
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
List mailing list -- list@maillist.saga-omii.org
To unsubscribe send an email to list-leave@maillist.saga-omii.org
I really enjoyed reading all of the comments about using AI for genealogy.
I basically agreed with all of the cautions about "believing" what AI comes
up with. However, I am not sure what people concluded from all of the
discussion. If people concluded that they should not use AI for genealogy,
then I would disagree. Much depends on what you are asking AI to do for
you. If it's grunt work, such as sorting through a GEDCOM file for a sorted
list of counties in which people were born, AI is generally good for that.
Think of AI as a research assistant. If you want AI to give you "answers,"
like names, dates, and whether a certain person is another person's mother,
then you are asking the wrong questions.
Recently someone in my family challenged a link between a son and his
parents that I had in my database. After researching it for a while, I
decided that I could not find any actual documentation that supported the
linkage. So I posed the following question/prompt to ChatGPT:
This is what I wrote/queried:
You are an expert genealogist. Thomas Henry was born in Lancaster, PA, in
1755, and died in 1820 in Fermanagh Twp., in Juniata Co., PA. He married
Dorothy Preston (1756-1829). Many genealogical databases say that he was
the son of Col. William Henry Sr. and Ann Taylor Wood. However, this
appears to be based on hearsay and rumor. Can you find any historical
documentation that would support this parentage?
This is what I got from ChatGPT:
I wasn’t able to uncover any primary documentation—such as probate
records, wills, land deeds, or church baptismal entries—from the 18th
century that definitively link Thomas Henry (1755–1820, Fermanagh Twp.,
Juniata Co.) to Col. William Henry Sr. and Ann Taylor Wood as his parents.
Here's what genealogical and historical records do show:
🔍* What we do know*
-
Thomas Henry was born circa 1755 in Lancaster, Pennsylvania, and
died in 1820 in Fermanagh Township, Juniata County. He married Dorothy
Preston (1756–1829)—a fact fairly well-established in family histories
and local county records (sites.rootsweb.com
https://sites.rootsweb.com/~pajuniat/juquery.htm?utm_source=chatgpt.com
, genpa.org
https://genpa.org/wp-content/uploads/publications/pennsylvania-genealogical-magazine/PGM_Volume_11_Number_2.pdf?utm_source=chatgpt.com
, wikitree.com
https://www.wikitree.com/wiki/Preston-5052?utm_source=chatgpt.com). - The commonly cited belief that his parents were William Henry Sr.
and Ann Taylor Wood appears to stem largely from oral tradition and
later genealogical compilations, not original documentation.
📑* Why the traditional lineage is questionable*
- A well-documented William Henry (1729–1786) of Lancaster—who was
indeed married to Ann Wood Henry—was a prominent gunsmith, politician,
and merchant (en.wikipedia.org
https://en.wikipedia.org/wiki/William_Henry_%28gunsmith%29?utm_source=chatgpt.com
).- Importantly, that William had no known son named Thomas born
around 1755. His children—e.g., John Joseph Henry—are identified
in Revolutionary-era letters and local documents. - He died in 1786, well before Thomas Henry’s burial in 1820, and
there’s no surviving will naming Thomas.
- Importantly, that William had no known son named Thomas born
Thus, it seems unlikely that Thomas was his son. Further, there’s no
evidence Ann Wood Henry was ever known as Ann Taylor Wood, or linked to an
earlier William Henry Sr. of that name.
🧾* What would support the claim?*
To confirm Thomas Henry’s parentage, you’d ideally find one of the
following:
- A baptismal or church entry naming William as the father.
- A will or probate record from William naming a son Thomas born in
-
Land deeds showing a transfer to Thomas explicitly identifying him
as William’s son. - A contemporary newspaper, milestone, or correspondence linking
Thomas to that Henry family.
To date, no such documents have surfaced in Lancaster or Juniata County
archives.
✅* Conclusion & next steps*
- There is no documentary proof connecting Thomas Henry (1755–1820) to
William Henry Sr. and Ann Taylor Wood. - The Lancaster William Henry, married to Ann Wood, is
well-documented—but not Thomas’s father, as far as records show. - The traditional link remains speculative, likely based on family
lore, not historical evidence.
To advance the research, I'd recommend:
- Searching Lancaster-area church registers (Presbyterian, Anglican,
or early Baptist/Friends) for a Thomas Henry baptism circa 1755 specifying
father's name. - Examining Lancaster County will and probate records (1750–1770) for
any William Henry who listed a son Thomas. - Reviewing land transaction records to see if a transfer from a
William Henry to a Thomas Henry occurred with relational context. - Checking Juniata County probate entries after 1820 for how Thomas
described his parentage in his own will.
Without one of these, the prevailing assertions remain genealogical
assumptions rather than documented fact.
It seems to me that this answer is good, sound. I especially like the
sections on "What Would Support the Claim?" and "To advance the research."
The only real "problem" with the answer is that Juniata County did not
exist in 1820. It was carved out of Mifflin County in 1831, so one would
need to check Mifflin County probate entries after 1820, not Juniata
County records. But that mistake was partly due to my mistake in the prompt
when I said that he died in Juniata County in 1820, which was not correct.
In the meantime, I removed the link between Thomas Henry and his supposed
parents from my database.
I think AI is particularly good at (1) grunt work; (2) suggesting research
plans; (3) suggesting bibliographical resources; and (4) providing ideas to
pursue, rather than "answers."
Please use it! As you learn to write wiser, more appropriate prompts, I
think you will find it useful!
P.S.: One more caution: If you use the free version of ChatGPT, they
consider your prompts "fair game" and they become ChatGPT's intellectual
property, but if you use the paid subscription, they do not do that,
according to a seminar I attended at RootsTech 2025 that I attended.
Loren L. Johns
Stuart, Florida
"I will accept no bull from your house." --Ps. 50:9, RSV
Loren, your post is well-reasoned and makes a lot of sense. I fully agree with it. AI can be an extremely useful tool if you're aware of its limitations.
Regards,
-- Jeff
--
Dr. Jeffrey S. Vitter
+1 (504) 810-7610
jsv@vitter.org | www.vitter.org
On Jun 30, 2025, at 4:38 PM, Loren Johns <llj215@gmail.com> wrote:
I really enjoyed reading all of the comments about using AI for genealogy. I basically agreed with all of the cautions about "believing" what AI comes up with. However, I am not sure what people concluded from all of the discussion. If people concluded that they should not use AI for genealogy, then I would disagree. Much depends on what you are asking AI to do for you. If it's grunt work, such as sorting through a GEDCOM file for a sorted list of counties in which people were born, AI is generally good for that. Think of AI as a research assistant. If you want AI to give you "answers," like names, dates, and whether a certain person is another person's mother, then you are asking the wrong questions.
Recently someone in my family challenged a link between a son and his parents that I had in my database. After researching it for a while, I decided that I could not find any actual documentation that supported the linkage. So I posed the following question/prompt to ChatGPT:
This is what I wrote/queried:
You are an expert genealogist. Thomas Henry was born in Lancaster, PA, in 1755, and died in 1820 in Fermanagh Twp., in Juniata Co., PA. He married Dorothy Preston (1756-1829). Many genealogical databases say that he was the son of Col. William Henry Sr. and Ann Taylor Wood. However, this appears to be based on hearsay and rumor. Can you find any historical documentation that would support this parentage?
This is what I got from ChatGPT:
I wasn’t able to uncover any primary documentation—such as probate records, wills, land deeds, or church baptismal entries—from the 18th century that definitively link Thomas Henry (1755–1820, Fermanagh Twp., Juniata Co.) to Col. William Henry Sr. and Ann Taylor Wood as his parents.
Here's what genealogical and historical records do show:
🔍 What we do know
- Thomas Henry was born circa 1755 in Lancaster, Pennsylvania, and died in 1820 in Fermanagh Township, Juniata County. He married Dorothy Preston (1756–1829)—a fact fairly well-established in family histories and local county records (sites.rootsweb.com, genpa.org, wikitree.com).
- The commonly cited belief that his parents were William Henry Sr. and Ann Taylor Wood appears to stem largely from oral tradition and later genealogical compilations, not original documentation.
📑 Why the traditional lineage is questionable
- A well-documented William Henry (1729–1786) of Lancaster—who was indeed married to Ann Wood Henry—was a prominent gunsmith, politician, and merchant (en.wikipedia.org).
- Importantly, that William had no known son named Thomas born around 1755. His children—e.g., John Joseph Henry—are identified in Revolutionary-era letters and local documents.
- He died in 1786, well before Thomas Henry’s burial in 1820, and there’s no surviving will naming Thomas.
Thus, it seems unlikely that Thomas was his son. Further, there’s no evidence Ann Wood Henry was ever known as Ann Taylor Wood, or linked to an earlier William Henry Sr. of that name.
🧾 What would support the claim?
To confirm Thomas Henry’s parentage, you’d ideally find one of the following:
- A baptismal or church entry naming William as the father.
- A will or probate record from William naming a son Thomas born in 1755.
- Land deeds showing a transfer to Thomas explicitly identifying him as William’s son.
- A contemporary newspaper, milestone, or correspondence linking Thomas to that Henry family.
To date, no such documents have surfaced in Lancaster or Juniata County archives.
✅ Conclusion & next steps
- There is no documentary proof connecting Thomas Henry (1755–1820) to William Henry Sr. and Ann Taylor Wood.
- The Lancaster William Henry, married to Ann Wood, is well-documented—but not Thomas’s father, as far as records show.
- The traditional link remains speculative, likely based on family lore, not historical evidence.
To advance the research, I'd recommend:
- Searching Lancaster-area church registers (Presbyterian, Anglican, or early Baptist/Friends) for a Thomas Henry baptism circa 1755 specifying father's name.
- Examining Lancaster County will and probate records (1750–1770) for any William Henry who listed a son Thomas.
- Reviewing land transaction records to see if a transfer from a William Henry to a Thomas Henry occurred with relational context.
- Checking Juniata County probate entries after 1820 for how Thomas described his parentage in his own will.
Without one of these, the prevailing assertions remain genealogical assumptions rather than documented fact.
It seems to me that this answer is good, sound. I especially like the sections on "What Would Support the Claim?" and "To advance the research." The only real "problem" with the answer is that Juniata County did not exist in 1820. It was carved out of Mifflin County in 1831, so one would need to check Mifflin County probate entries after 1820, not Juniata County records. But that mistake was partly due to my mistake in the prompt when I said that he died in Juniata County in 1820, which was not correct.
In the meantime, I removed the link between Thomas Henry and his supposed parents from my database.
I think AI is particularly good at (1) grunt work; (2) suggesting research plans; (3) suggesting bibliographical resources; and (4) providing ideas to pursue, rather than "answers."
Please use it! As you learn to write wiser, more appropriate prompts, I think you will find it useful!
P.S.: One more caution: If you use the free version of ChatGPT, they consider your prompts "fair game" and they become ChatGPT's intellectual property, but if you use the paid subscription, they do not do that, according to a seminar I attended at RootsTech 2025 that I attended.
Loren L. Johns
Stuart, Florida
"I will accept no bull from your house." --Ps. 50:9, RSV
The one item that one could be considered missing would be utilze genetic genealogy to comfirm a paternal line relationship. Get yDNA haplogroup/SNP data from different descendant lines and see if paternal descendants of Thomas Henry match those others. Working back the mtDNA maternal line to investigate Ann Taylor Wood could be used to supply supplemental information.
Wayne
On Monday, June 30, 2025 at 05:38:19 PM EDT, Loren Johns llj215@gmail.com wrote:
I really enjoyed reading all of the comments about using AI for genealogy. I basically agreed with all of the cautions about "believing" what AI comes up with. However, I am not sure what people concluded from all of the discussion. If people concluded that they should not use AI for genealogy, then I would disagree. Much depends on what you are asking AI to do for you. If it's grunt work, such as sorting through a GEDCOM file for a sorted list of counties in which people were born, AI is generally good for that. Think of AI as a research assistant. If you want AI to give you "answers," like names, dates, and whether a certain person is another person's mother, then you are asking the wrong questions.
Recently someone in my family challenged a link between a son and his parents that I had in my database. After researching it for a while, I decided that I could not find any actual documentation that supported the linkage. So I posed the following question/prompt to ChatGPT:
This is what I wrote/queried:
You are an expert genealogist. Thomas Henry was born in Lancaster, PA, in 1755, and died in 1820 in Fermanagh Twp., in Juniata Co., PA. He married Dorothy Preston (1756-1829). Many genealogical databases say that he was the son of Col. William Henry Sr. and Ann Taylor Wood. However, this appears to be based on hearsay and rumor. Can you find any historical documentation that would support this parentage?
This is what I got from ChatGPT:
I wasn’t able to uncover any primary documentation—such as probate records, wills, land deeds, or church baptismal entries—from the 18th century that definitively link Thomas Henry (1755–1820, Fermanagh Twp., Juniata Co.) to Col. William Henry Sr. and Ann Taylor Wood as his parents.
Here's what genealogical and historical records do show:
🔍 What we do know
- Thomas Henry was born circa 1755 in Lancaster, Pennsylvania, and died in 1820 in Fermanagh Township, Juniata County. He married Dorothy Preston (1756–1829)—a fact fairly well-established in family histories and local county records (sites.rootsweb.com, genpa.org, wikitree.com).
- The commonly cited belief that his parents were William Henry Sr. and Ann Taylor Wood appears to stem largely from oral tradition and later genealogical compilations, not original documentation.
📑 Why the traditional lineage is questionable
-
A well-documented William Henry (1729–1786) of Lancaster—who was indeed married to Ann Wood Henry—was a prominent gunsmith, politician, and merchant (en.wikipedia.org).
-
Importantly, that William had no known son named Thomas born around 1755. His children—e.g., John Joseph Henry—are identified in Revolutionary-era letters and local documents.
-
He died in 1786, well before Thomas Henry’s burial in 1820, and there’s no surviving will naming Thomas.
Thus, it seems unlikely that Thomas was his son. Further, there’s no evidence Ann Wood Henry was ever known as Ann Taylor Wood, or linked to an earlier William Henry Sr. of that name.
🧾 What would support the claim?
To confirm Thomas Henry’s parentage, you’d ideally find one of the following:
- A baptismal or church entry naming William as the father.
- A will or probate record from William naming a son Thomas born in 1755.
- Land deeds showing a transfer to Thomas explicitly identifying him as William’s son.
- A contemporary newspaper, milestone, or correspondence linking Thomas to that Henry family.
To date, no such documents have surfaced in Lancaster or Juniata County archives.
✅ Conclusion & next steps
- There is no documentary proof connecting Thomas Henry (1755–1820) to William Henry Sr. and Ann Taylor Wood.
- The Lancaster William Henry, married to Ann Wood, is well-documented—but not Thomas’s father, as far as records show.
- The traditional link remains speculative, likely based on family lore, not historical evidence.
To advance the research, I'd recommend:
- Searching Lancaster-area church registers (Presbyterian, Anglican, or early Baptist/Friends) for a Thomas Henry baptism circa 1755 specifying father's name.
- Examining Lancaster County will and probate records (1750–1770) for any William Henry who listed a son Thomas.
- Reviewing land transaction records to see if a transfer from a William Henry to a Thomas Henry occurred with relational context.
- Checking Juniata County probate entries after 1820 for how Thomas described his parentage in his own will.
Without one of these, the prevailing assertions remain genealogical assumptions rather than documented fact.
It seems to me that this answer is good, sound. I especially like the sections on "What Would Support the Claim?" and "To advance the research." The only real "problem" with the answer is that Juniata County did not exist in 1820. It was carved out of Mifflin County in 1831, so one would need to check Mifflin County probate entries after 1820, not Juniata County records. But that mistake was partly due to my mistake in the prompt when I said that he died in Juniata County in 1820, which was not correct.
In the meantime, I removed the link between Thomas Henry and his supposed parents from my database.
I think AI is particularly good at (1) grunt work; (2) suggesting research plans; (3) suggesting bibliographical resources; and (4) providing ideas to pursue, rather than "answers."
Please use it! As you learn to write wiser, more appropriate prompts, I think you will find it useful!
P.S.: One more caution: If you use the free version of ChatGPT, they consider your prompts "fair game" and they become ChatGPT's intellectual property, but if you use the paid subscription, they do not do that, according to a seminar I attended at RootsTech 2025 that I attended.
Loren L. JohnsStuart, Florida
-----------------------------------------------------------------------"I will accept no bull from your house." --Ps. 50:9, RSV