Thoughts on the Creation of a Universal Family Tree

One of the outstanding demonstrations of current computer power was the Google-DeepMind’s AI system AlphaGo's defeat of a human professional for the first time last year at the ancient board game Go. Subsequently, Google's AlphaGo Zero surpassed the strength of AlphaGo in three days by winning 100 games to 0, reached the level of AlphaGo Master in 21 days, and exceeded all the old versions in 40 days. See Wikipedia: AlphaGo Zero. If you aren't familiar with the ancient Chinese game of Go, you should read "What is Go?" from the American Go Association.

In a JSTOR Daily post entitled, "The Inevitable Triumph of Iteration over Intellect," Kevin Litman-Navarro observed the following about the AlphaGo Zero achievement,

But how impressive is this really? If you could do anything 4.9 million times, you’d probably become a master too. The ability to iterate a larger number of times on a small time scale is what makes neural networks such useful tools in artificial intelligence.

Artificial Intelligence has been a buzzword for a long time and is generally defined as the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. From a genealogical perspective, the increasingly available "record hints" provided by several online websites such as FamilySearch.org, Ancestry.com, MyHeritage.com, and Findmypast.com, are more or less an implementation of an artificial intelligence program. The main limitation of the hints being the lack of indexed data and the variations in the way genealogical information has been recorded. Obviously, one huge limitation is an accurate and reliable method of handwriting recognition implemented for genealogical research.

With the rise of huge, integrated, online family tree programs that compare and correlate information from millions of users, there is a possibility that more sophisticated artificial intelligence type programs could begin to build a reliable universal family tree. The FamilySearch.org Family Tree has the potential to begin such a project, but there is presently no way to integrate information from the millions of other online family trees on other programs and there is definitely no way to evaluate the reliability of the information presented in all those family trees.

Another basic limitation in undertaking such a project is the unreliability of existing historical data and the fact that all such data is scattered around the world in fragments which are more or less still unavailable. You could be persuaded to believe that current DNA testing could solve part of the problem. Even assuming that universal DNA testing would be possible or desirable, the results from the testing would have to be correlated and evaluated.

What would a universal family tree look like?

I think any universal family tree would need to be focused on the supported connections between individuals now shown on many well-sourced family trees presently available. For example, as I noted above, there are several websites that provide record hints. When these record hints are supported by reliable DNA matching data, there is a good possibility of establishing a basis for the foundation of a universal family tree. For example, let's suppose that existing data supports my relationship to my immediate family and some or all of the connections are supported by DNA testing data assuming further that all of the family members participate. This basic, supported information can then become the basis for constructing additional reliable connections.

Could an artificial intelligence program be written to build such a family tree? Probably, assuming the data was made available and not blocked by current attitudes towards privacy. But these are big assumptions. Presently, the best that we have are the major genealogical database and family tree programs.