Artificial Intelligence, Genealogy and DNA

Mick Tator

All the big Genealogy companies (Genocos) have for some time been using AI to provide ‘Hints’ for ancestors and relatives. This was a basic service, but it now has become much more refined, primarily because the task has been outsourced to third party companies (e.g. MyHeritage is using Open AI). For me, that is where a big problem arises. Genealogists now should be much more careful about consent and the data they input into AI-based search functions.

If you consent to most Genocos’ latest T’s and C’s, the search engine will firstly search its database (i.e. its proprietary data) followed by data from its other members (e.g. members’ trees) and the inputs will be shared with OpenAI to help refine a response. HOWEVER, the data you input to populate your search will also be retained FOREVER as part of the machine-learning process and it cannot be deleted.

Currently, if you accept new ‘Ts & Cs’ without reading them, you are leaving yourself wide open to (IMO) unacceptable use of personal data. I now have become much more guarded as a result of AI and advise everyone not to simply click the ‘Accept’ button when new changes are proposed. You should READ them and see if you are at ease with what is proposed.

VirginiaB

Thanks very much for this, Mick. Do you have any specific info on Ancestry's use of AI?

pinkypinky

If you don't accept new terms and conditions, it does mean giving using that site. I'm not disagreeing with what Mick is saying - it's a difficult call.

As a professional, I can't just decide not to use these big companies. It's easier to do as a hobbyist.

Mick Tator

I’ve been on the fringes of AI for a long time, feeling my way, seeing it being used and learning.

@Virginia B – Ancestry has been using AI for decades but AFAIK it has been proprietary and not very intrusive. If you have given consent, they share your (anonymized) DNA results. It also has its own AI for ‘reading’ handwriting. Every geno company uses AI- if you input a search for e.g. Mick Tator, born Munster, Ireland between X & Y year, the algorithm (AI) will find all the Mick Tators in that timeframe/province / country. If your tree on Ancestry is private, search results are hidden BUT information about deceased individuals (name, birth year, birthplace, and any documents you have uploaded) will appear . It is up to others to contact you and for you to respond to allow access.

All the other businesses, including FamilySearch have the same or variations, up to now.......

The next step is the intrusive bit ….MyHeritage is using an outside company (Open AI) and before Christmas wrote to subscribers looking for consent to use their data. If agreed, consent allows use of more intrusive AI machine learning algorithms and that data is stored, forever! These ‘teach’ the search engine to learn from the data (input & found) and then use it to predict relationships beyond the usual DNA matches. By mining ‘Big Data’ i.e. millions of records, machine learning will discover links that heretofore were missed. (OK, some will be rubbish, but it will improve over time.)

@Pinky – Yes, In the EU, our GDPR rules do not extend to dead people but we do have rules for data on BMDs, so I’d have concerns. I’m an early adapter (my personal email is simply my name, no numbers) but on this topic I remain quite reticent.