On 2017-03-14 04:00, Daniel wrote:
On 13/03/2017 1:57 AM, Ron Hunter wrote:
On 3/12/2017 4:04 AM, Daniel wrote:
On 12/03/2017 6:53 PM, Ron Hunter wrote:
like vacation pictures directly from my iPhone to Facebook. The
downside of this is that the images are greatly reduced in
quality/resolution, and have the metadata stripped from them.
and, I'm guessing, not just your's, Ron!
Not sure just what you mean. I load the images from my phone to
Facebook, and, yes, all uploaded photos are stripped of their metadata
by Facebook to prevent stalkers from finding people because they forgot
to strip their GPS data from their images.
Some time ago, the Australian Government legislated for all Australian
ISP's to have to hold e-mail metadata for at least two years, so
yesterday I asked my ISP, via ChatZilla, if that law was now in force ....
[18:21] ISP Yes, that is law. And it's WAY WAY WAY more than merely email.
[18:22] Me Oh!! O.K., should I be really, really, worried??
[18:25] Me By your "more than merely email" are you suggesting nntp
and/or http and/or ftp?
[18:27] ISP yes, no and yes. (and more)
So, Big Brother is definitely out there!
Saw an intersting comment, can't recall where, maybe The Register: the
age of Big Ddata has shown up a serious problem: even with AI search
a) you still need humans at some stage to decide whether the data the
algorithm tosses up is "significant";
b) the human has to decide what that significance is. Since almost all
the data has significance if and only if you already have some idea of
whether and what kind of nefariousness is being planned, that's very
difficult. Most of the significant data isn't recognised as such until
after the event. Eg, the Boston marathon bombers, who were under
surveillance. But so were thousands (tens of thousands?) of other
people, almost all of which did (and will do) nothing at all.
It's the effect of error rate and incidence rate. A medical screening
test whose error rate is greater than the incidence rate will always tag
up many more false positives than true ones. In medicine, you can
usually use a second or third test to find the true cases. That's not so
in intelligence, which has basically only one test: "This doesn't smell
right". AI algorithms have a lower error rate than a human, but it's
still higher than the incidence rate.
Have a good day,
It's called "opinion" because it's not knowledge.
general mailing list