SoylentNews is people
SoylentNews
CNN reports that URL, the porn-sniffing dog, is the newest crime-fighting tool at the Weber County Sheriff's office with a nose that could help put away some of the country's most predatory and dangerous criminals. URL (pronounced Earl) sniffs out electronic storage media. Still just a pup, the 18-month-old K-9 is one of fewer than two dozen such dogs in the United States that hunt the unique chemical compounds emitted from flash drives, memory cards, cell phones, iPads and other similar devices. While dogs like URL can't tell detectives if a device has electronic evidence on it, they are able to find devices that humans might otherwise miss.
Detective Cameron Hartman points to the high-profile case of former Subway spokesman Jared Fogle, who was convicted on child pornography and other charges last year. A K-9 named Bear, who was trained by the same man who trained URL, led investigators to hidden thumb drives inside Fogle's home. The US Attorney's office for Southern Indiana confirmed those devices contained evidence against Fogle. URL has found evidence relating to pornography during the execution of search warrants for the task force in several investigations of child sex crimes and child trafficking. "He actually found a USB that was in this jar that was closed, and the jar was in a box, and the box had stuff in it. The jar itself had stuff in it."
(Score: 5, Informative) by bradley13 on Tuesday September 06 2016, @09:31AM
My wife works with dogs, doing various kinds of scent training. It's a completely different world, one that people cannot really understand. To give a rough idea of the difference, how many people are standing on the far shore, in this photograph [shutterfreaks.com]? If this were a world of smells, a dog could tell you. That's the qualitative difference between our sense of smell, and theirs.
So: is it possible that electronic devices with solid-state memory have a particular smell? Sure, in fact, it's likely. The thing is: we, ourselves, have no idea. Since we cannot perceive the smells, how do we teach a dog what to search for? If we present a dog with sample devices, how can we be sure that we aren't teaching them to hunt for particular types of plastic? Or the smell of solder? Or the smell of items that spent weeks in a shipping container, crossing the ocean? Or - assuming it's police doing the training - items handled by people carrying weapons that contain gunpowder?
Let's assume that we have a gifted trainer, who manages to train dogs to search for the right scents.
Then comes the next set of problems: the handler. The trainer may have done a perfect job, but the handler is now the source of rewards. "Sausage when I find something!" For the trainer, it was SSDs, but for the handler, the above comments apply all over again. In the best case, the training must be continually reinforced, which takes continual effort. In the worst case, dogs want to please people, and they learn to alert whenever the handler wants them to.
How many police departments actually understand all of this? How many send their handlers to be trained, not just pro forma, but where they actually fail if they cannot do it right? How many allocate the time for the continual reinforcement of the training? Anyone want to take bets?
Everyone is somebody else's weirdo.
(Score: 2) by FatPhil on Tuesday September 06 2016, @09:43AM
There was a visual AI story, military - tank detection, less than a decade ago which revealed a huge error in training; the AI was detecting shadows, as every tank it had been trained on was skulking in a shadow.
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 0) by Anonymous Coward on Tuesday September 06 2016, @09:51AM
the AI was detecting shadows, as every tank it had been trained on was skulking in a shadow
He Hates These Cans! [youtube.com]
(Score: 0) by Anonymous Coward on Tuesday September 06 2016, @11:25AM
I don't even need to follow the link to start laughing. That was a pretty funny scene. You obviously know your shit from Shinola.
(Score: 5, Informative) by pTamok on Tuesday September 06 2016, @10:02AM
In the worst case, dogs want to please people, and they learn to alert whenever the handler wants them to.
This is true. See the linked paper.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3078300/ [nih.gov] "Handler beliefs affect scent detection dog outcomes" - Animal Cognition 2011 May; 14(3): 387–394.
Our aim was to evaluate how human beliefs affect working dog outcomes in an applied environment. We asked whether beliefs of scent detection dog handlers affect team performance and evaluated relative importance of human versus dog influences on handlers’ beliefs.
...
Handlers’ beliefs that scent was present potentiated handler identification of detection dog alerts. Human more than dog influences affected alert locations. This confirms that handler beliefs affect outcomes of scent detection dog deployments.
But the USA Federal courts don't seem over-worried:
https://www.washingtonpost.com/news/the-watch/wp/2015/08/04/federal-appeals-court-drug-dog-thats-barely-more-accurate-than-a-coin-flip-is-good-enough/?utm_term=.19f9a976d556 [washingtonpost.com] "Federal appeals court: Drug dog that’s barely more accurate than a coin flip is good enough"
There's no doubt that dogs have a more sensitive sense of smell than humans (they can detect smells at lower concentrations), and have greater discrimination (they can tell the difference between more types of smells than humans); the problem is, they are bred to please their owner, and are very good at reading the non-verbal (and subconscious) cues given by handlers. This means that, although they will find more search items than humans working without dogs, they are prone to a high rate of false positives. This can be worked around, to a limited extent: for example, see the work done on getting dogs to detect cancer:
http://www.mindfully.org/Health/2006/Canine-Cancer-Detection1mar06.htm [mindfully.org]"Diagnostic Accuracy of Canine Scent Detection in Early-and Late-Stage Lung and Breast Cancers" - Integrative Cancer Therapies, v.5, n.1 1mar2006
In this study, a double blinding protocol was used to better evaluate the dog's performance.
(Score: 3, Informative) by Arik on Tuesday September 06 2016, @07:08PM
Search for 'police IQ hiring' you'll find that many departments now have policies barring anyone with an IQ high enough to completely understand your points from ever getting their foot in the door. So it seems likely very few do and even fewer will in the future.
If laughter is the best medicine, who are the best doctors?