On 9/18/2015 10:55 AM, Oren wrote:
On Fri, 18 Sep 2015 10:37:21 -0700, Don Y
wrote:

On 9/18/2015 10:07 AM, Oren wrote:
On Fri, 18 Sep 2015 09:54:29 -0700, Don Y
wrote:

On 9/18/2015 5:24 AM, Roger Blake wrote:
On 2015-09-18, Don Y wrote:
OTOH, I've had friends who've been spontaneously notified that they
must have new photos taken (!) and had to sit through the process.

It seems that most if not all states are amassing huge databases of facial
recognition data. Real police-state stuff. To do this they need high-res
digital photos. (Of course they assure us the data won't be misused,
will be protected, etc. Maybe they have a bridge in Brooklyn to sell
us as well.)

Yes, that was the assumption I made. A neighbor was told she had to
have her photo retaken "because she was smiling" (?) Perhaps they
assume most people RARELY smile and want the recognizers to work with
your face in its "normal", grumpy state! :

I was told to remove my cap for the photo. DMV is not the only
database law enforcement can access for photos. Passports and a few
others can be used by LEO.

Your photo exists in *lots* of places -- credit cards, passports,
DL's, "membership clubs", etc.

Casinos take ~80 photos per minute on the Las Vegas Strip. Street
cameras, red light cameras and such, too.

But those don't bind a unique identifier to the photo. My fingerprints
exist on thousands of "public" (i.e., freely accessible without a warrant)
items every day. But, even if someone wanted to "lift" them, they still
have to be tied to my name, address, etc.

Nevada keeps a Black Book of cheaters that are banned by law. If a

But the cheater is uniquely identified WHEN he is banned. They don't
just decide, on their own, that "the guy in the red hat at table 6
will be forever banned". I.e., if someone *looking* like him (where
"looking like" is taken as "the uncertainty in the recognition
algorithm") comes in and is refused service, that person *could*
present a credential (passport, etc.) that "proves" he is not
"John Smythe" but, rather, just happens to resemble him (perhaps
not even *well*!)

[There's a formal process by which he identifies himself and is *then*
banned]

guy walks in, they know him immediately. Whether government or private
business photos are used, facial recognition can match faces. That is
or might be the identifier, no?

Businesses (e.g., Target) use cameras to monitor individual shopper's
travels through the store. Presumably, they can (and would WANT TO)
track each (as yet nameless) "individual" up to the checkout counter.
If the individual uses a credit card to pay, now they can bind a
name to a face (and any other characteristics that they can observe
from the surveillance video -- e.g., "gait" may have some value as
an identifier or disambiguator).

Windows 10 likes to listen to you. So, all those speech samples
sit on a server, somewhere (ditto for your TV and other similar
"appliances" -- "Echo"?). Google voice knows what you sound
like from all the phone calls you make.

I.e., there is a sh*tload of biometric data available along with
ways of accurately binding a name to those data -- or, coming up
with a probabilistic assessment that "this is Oren" (even if they
don't know for sure that it's you)

The downside of this (from the G-man perspective) is that you end up
with *massive* amounts of data. So, doing anything to actually
track individuals (not *an* individual) is beyond the capabilities of
any sort of processing power that is likely to exist in our
lifetimes -- and beyond.

E.g., if *I* suddenly appeared 2000 miles from here and there was
no "biometric trail" that *led* an observer to FOLLOW MY TRAVELS,
it would be hard for someone to discover that I was now no longer
"here" but, rather, "there".

I track the position of "occupants" in this house (ignore the details).
But, I rely on *watching* their movements. If you suddenly appear
somewhere else (in the blink of an eye), you'll confuse my softwa
"He *can't* possibly be over there -- cuz he was at the other end of the
house just a moment ago!"

[Given time, the software accepts the new reality. But, it defies the
assumptions that are built into it (to make the algorithms simpler and
more reliable!) because you've done something (i.e., the data that
represents your location) that is "impossible"]