[ExI] sanity and connectedness, was: RE: list test
Anders Sandberg
anders at aleph.se
Mon Jul 29 07:52:41 UTC 2013
On 2013-07-29 06:38, spike wrote:
>
> Consider for instance this magnificent beauty:
>
> Mt. Rainier Tiger Beetle - Cicindela depressula
>
> I have been hiking for years up at Mount Rainier, and had never seen
> one of these that I recall. The photo above is about 2x actual size.
> I spent some time observing them, after having the good fortune of
> showing up apparently right after they hatched. I yearned to know
> more about these beasts, while I was still there on site to observe.
> Later I found out it is likely a Cicindela depressula. Kewall!
>
Indeed. The dispirited tiger beetle! (the name apparently has to do with
the "broken elbow" in the patterning, rather than any mood in the
beetle) I know the problem of photographing tiger beetles - they refuse
to sit still for a picture.
An automatic species detector would be awesome. But it is tricky to get
the species right. I can imagine software recognizing the picture above
as "a cincidelid beetle", and likely homing in on a few likely species
based on color and location. But to get to Cicindela depressula you need
to check the length of the labrum and how the eybrow bristles look - and
that requires a facial closeup. Many species are even worse, you need to
dissect them to figure out what they are. So the species detector should
have a micro-DNA sample device too, in order to use DNA barcodes.
There are a few projects going in this direction:
http://link.springer.com/article/10.1007%2Fs00371-013-0782-8
http://www.ppgia.pucpr.br/~alekoe/Papers/ISM2011-Koerich.pdf
<http://www.ppgia.pucpr.br/%7Ealekoe/Papers/ISM2011-Koerich.pdf>
http://www.cs.waikato.ac.nz/~ml/publications/2006/MayoSGAI_2006.pdf
<http://www.cs.waikato.ac.nz/%7Eml/publications/2006/MayoSGAI_2006.pdf>
https://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6481468&sortType%3Dasc_p_Sequence%26filter%3DAND(p_IS_Number%3A4358066)
<https://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6481468&sortType%3Dasc_p_Sequence%26filter%3DAND%28p_IS_Number%3A4358066%29>
http://leafsnap.com/
It seems to me that sensor fusion is the way to go: use pictures,
animations, recorded birdsong, whatever to help focus the search. One
could use something like
http://people.csail.mit.edu/torralba/tinyimages/ or
http://googleresearch.blogspot.se/2013/06/fast-accurate-detection-of-100000.html
to do an overall guess at what kind of critter it is, then apply local
expert software to narrow things down. In many cases it will just tell
you "some kind of cincidelid" or "little brown thing" due to lack of
information, but I suspect it will be amazingly good under the right
circumstances.
I think we will get the system eventually. And probably sooner than it
looks.
--
Dr Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20130729/b819106b/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 10701 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20130729/b819106b/attachment.jpe>
More information about the extropy-chat
mailing list