There will be some spectacular failures when erroneous
genealogies are widely replicated and feed too much garbage into
the AI, and when there are uncanny genetic similarities that occur
in isolation of one another, but I'll bet these (sometimes
frightening) machines will fairly quickly develop the ability to
filter out the vast majority of bad genealogies.?
It feels to me like the situation may parallel the time 25-30
years ago when the majority of the human genome was dismissed as
"junk DNA" by scientists, mainly because they couldn't determine
its function.? In the same way, we effectively dismissed autosomal
DNA for serious genealogical purposes because of the all the
recombination.? But rather than rendering it meaningless, maybe
all the mixing just rendered it too complicated for our, uh,
single core processors (brains) to manage.? Whereas if you have
dozens or hundreds of threads that you can hold in your
(electronic) brain at once, relationships appear...
Optimistically,
Edward
On 7/1/22 03:06, Nigel B via groups.io
wrote:
toggle quoted message
Show quoted text
That is interesting. ?Thrulines has predicted some distant links
for me too. ?It is powerful technology but can be thrown by people
inputting accurate trees. ?Unfortunately that is not always the
case, particularly earlier than say 1800. ?Other users copy those
inaccurate trees and they are then given weight as ¡®high
confidence¡¯.
I am occupied with environmental projects too - around a
local stream and also campaigning to reduce sewage ¡®storm
overflows¡¯ into our river system.
Nigel?
?
Sorry, I'll stop now... but imagine when you can bridge
to Sandy Mocilan's other faint contacts that might bear
"Hammond" centimorgans that I've lost.?
Hell, for all we know now, Sandy Mocilan may have a faint
autosomal match to a Turner, or a Braginton.
I bet it doesn't take that long for it to be figured out
how to make those sort of connections 'in silico'.
It's funny, in my day job one of things I am paid to
worry about it all the exabytes of DNA sequences of
plants, animals, microbes, etc... that have no geographic
origin ascribed to them.? I was just talking last week
with a data scientist about the idea that you could
probably, quite soon, unleash an AI on a massive
biodiversity sequence database and have it figure out how
to more or less reliably tell you where each sequence came
from.?
In a similar way...
EH
On 6/30/22 20:01, Edward
Hammond wrote:
There's my best Hammond example of a nice machine
learning match.?
I would never have been able to link Sandy Mocilan to
myself on the basis of a tiny genetic match (10cm) with
anything less than her (him?) having a well-developed
genealogy.?
I haven't run down Sandy's line myself, but it looks
plausible.?
In effect a probable 7th generation link made possible
by AI.
Maybe not revolutionary yet, but I suspect we're at the
early stages.
EH
On 6/30/22 19:25, Edward
Hammond wrote:
On a slightly different note, have you noticed how AI
/ machine learning seems to be producing some pretty
interesting results at ancestry.com?? I'm not crazy
about yet another pay site, but I've been amazed by
its ability to (seemingly ... I haven't fully proofed
very many) accurately match rather distant autosomal
hits.? The sort of people that you scroll by in the
autosomal results at FTDNA and think "that one just
isn't worth pursuing".
EH
On 6/30/22 19:11, Edward
Hammond wrote:
Completely agree that there's not much new to
discuss.? My intent was to simply have a point of
contact should anything interesting arise.? Since
you were frequently the emcee, I thought you might
have the best list and judgment as to who to place
here.? If you simply pass along the list of names
and e-mail addresses you think are appropriate, I
can send an invitation to join to those people.? If
you think its not worth that effort, fair enough,
it's certainly true that we haven't had any news of
note for some time.
I've turned into a very low power - milliwatts -
high frequency radio nut in the absence of anything
genetic to do!? That and journeyman forester, trying
(often in vain) to contain the unbelievably
aggressive invasive species trying to consume the
little shred of forest I own.
Edward
On 6/30/22 18:29, Nigel B
via groups.io wrote:
?
Hi Edward - all good here. ?Apologies for not
sorting our S16264 distribution list. ?I have been
busy with other interests / activities. ?It is now
several years since I had any new Y-DNA matches
and I confess I have lost much of my motivation
for keeping the group going. ?Still I agree we
should not let links lapse as we might all learn
more from future matches. ?Will try harder! ?
Nigel
?Hi Nigel -
Hope your summer is going well.? We're still
all by our lonesome in this group.
Edward
