click to view original photo

Genealogy and Singularity

Artificial intelligence can be a boon and bane to genealogical research.

Share

Content Details

Type: Article
Resource: GenWeekly
Prepared by:
Word Count: 1011 (approx.)
Short URL:

There have been some interesting new developments lately in the availability of computerized records. We all know of the main commercial sites and what offerings they have, and there are many libraries and historical societies doing a lot of digitizing of their records to make them more available. That's great news for researchers.

But we all know that sometimes things do not appear in the records, or they do and are terribly handwritten. And don't even let's get started on the weird and wonderful conclusions that people draw from all kinds of information. That's the human factor. GIGO – "garbage in, gospel out." How many anomalies have you ever seen in various records that you have found, only to be able to set them straight with your personal knowledge?

What I want to talk about now is perhaps what we will see coming forward in the next few years. Ray Kurzweil is a well-regarded inventor who has a lot of wonderful things to his credit – like the machine which reads text to a visually impaired person, et al. Recently I read an interview with him where he talks about "the Singularity," which to shorthand it, is when machines become as smart and then smarter than people. TIME< Magazine did a piece where Kurzweil gives dates for this to occur, and seems to think that it could be no later than 2045. I sure won't be here, but many readers could very well be alive then.

How this relates to genealogy (again, vastly oversimplied) is that machines can comb through records and present results for evaluation. If they are smarter than us, then maybe we won't have to think about the research that we do. Hah!

Where I see this being quite useful is having an artificial intelligence go off and find all sorts of records all the time and get back to us with the results. That's a time saver, like searching through various indexed censuses. I am reminded of the time I spent hours going through the reels of films for Albany, New York only to find the names of the people I was researching, misspelled, on the last page of the last ward. Or, the time that, totally by luck, I read the last page of a Soundex film. The people that I was looking for were there, but only the first two letters of a 10-letter last name were correct. Actually seeing the film showed me the address and family members, which I knew to be right. And the time another ancestor appeared with his name bizarrely misspelled in immigration, Civil War, and burial records. But all that can deduced to be right by human thought.

Of course, we might know that someone is the child of a brother (an uncle in reality), even though the birth certificate says that the husband is the actual father of the child (true story – that is the case with a friend and no one has told or is planning to tell him, and his mother is already dead, as is the father and the birth father). That's something that no artificial intelligence would know.

So where I see this as going is having an indefatigable mechanical research as a help -- I certainly would not take anything as total gospel from such a situation. But what might happen is that it will be marketed as such, and much bogus information will be accepted without critical evaluation. That's not a good thing, especially if people want their trees done quick and easy.

Even if there is the programming – because after all, this is not a flesh and blood research, just a blindingly fast and tireless idiot – it would have to be written in to the program to "think" like a person. Mr. Kurzweil is pretty "out there" on some of his claims – that computers will think like humans do by the end of the 2020s; that the Singularity (when the sum total of all machine knowledge is greater than that of all humanity) will occur in 2045; that after the Singularity, humans will become "functionally immortal" by being able to record all their thoughts and feelings into a machine; that we will be able to bring the dead back to life by resequencing failed organic materials – i.e., bodies; and that human intelligence will conquer the universe. I don't know about you, but I have found some pretty unlovely relatives that I would just as soon stay dead!

I do like the idea of the automatic scanning of lots of resources, but I don't see ever trusting a machine to make decisions that people should. I really like Google alerts, since they often find things for me that I otherwise would miss. Then again, auto-scanning of huge databases would make discovering relatives easier. But conversely, what about the following example – that someone researching a lady with the name Chatt and who thinks it's a British name, but finds her married to an Italian man, and appearing in those church records as Cetta. And that's a true example. She really was Cetta.

So, in a practical sense, is this going to help? Maybe. Wiser minds than mine write the Google and other commercial search algorithms. It's always good to have more sources available. But as long as we can say, and this again is a real life example, give me every Lithuanian-born male between 20 and 30 in this particular census, and then evaluate what the computer finds, we are in good shape. And "propinquity searching," for lack of a better term, would be a nice feature to have. Some brainy coder can write a search scheme that says, D county is also bordered by A, B, and C counties, so search those as well. (A family was found not having moved but a quarter mile, but across county borders in a hilly rural area.)

I look forward to the development and implementation of these Singularities for genealogy – but as with all research done by someone other than myself, with a close eye upon it.

Source Information: GenWeekly, New Providence, NJ, USA: Genealogy Today LLC, 2011.

The views and opinions expressed in this article are those of the author and do not necessarily reflect the views of Genealogy Today LLC.

*Effective May 2010, GenWeekly articles that are more than five years old no longer require a subscription for full access.

<< GenWeekly

<< Helpful Articles