Amazon’s Facial Recognition Tool Is Laden With Bugs

Amazon’s celebrity facial recognition tool mistakenly seemed to identify some criminals on FBI’s Most Wanted list as famed personalities such as late painter Bob Ross, BuzzFeed News reported on Friday.

BuzzFeed said it ran 507 photos of criminals on the FBI’s Most Wanted list and 3,248 photos from the National Institute of Standards and Technology’s (NIST) mugshot identification database through Amazon’s celebrity recognition tool, a feature the company introduced about a year ago…

BuzzFeed said the celebrity recognition tool identified 17 criminals on the FBI’s Most Wanted list as celebrities with a 95-100 percent confidence level, including Jerry Garcia, the late Grateful Dead guitarist, “Sons of Anarchy” actor Kim Coates and Ross.

The problem exists with all their Facial recognition products:

The American Civil Liberties Union tested Amazon’s facial recognition system — and the results were not good. To test the system’s accuracy, the ACLU scanned the faces of all 535 members of congress against 25,000 public mugshots, using Amazon’s open Rekognition API. None of the members of Congress were in the mugshot lineup, but Amazon’s system generated 28 false matches, a finding that the ACLU says raises serious concerns about Rekognition’s use by police.

Do you think the government has this problem with its top tier facial recognition?

And why is Amazon getting into facial recognition?

My own assumption is, facial recognition is too powerful to let the plebes have it. Imagine if you could identify anyone around you instantly by their face.

Now imagine if the entire social structure was laden with spies gathering information on everyone. A competent facial recognition program would allow you to turn it loose on the internet, scrubbing for any old breadcrumbs about the people you cross paths with. Maybe an ancient high school friend had up a picture from decades back with a real name. Maybe you can effortlessly locate names and IDs of everyone you cross paths with. That is the start of a database. That could all be trouble.

So Amazon announces it will use its billions to seize the marketplace on facial recognition programs, and then it fills the niche with a defective product.

Here is another thought. If I was Cabal intelligence, and I had control of Amazon, I would plug those civilian facial recognition apps into the confidential master database of Cabal spooks and informants, and program it so any facial search for one of those spooks would return a defective result – with 95-100% certainty.

And here we have searches on some criminals and Congress people returning defective results. Why might that be? Who is most compromisable? Who has the lowest morals and gets into the stickiest situations they would do anything to escape? Criminals and politicians.

I’d love to know who all these people were whose results went wildly divergent.

Tell others about r/K Theory, because everything else is fixed

0 0 votes

Article Rating

Name*

Website

5 Comments

Oldest

Newest Most Voted

Inline Feedbacks

View all comments

Rossa

6 years ago

Rumours are that Apple’s next iPhone will have no home button and will use facial recognition to unlock your phone. I’m not convinced it will work. Especially for women as we change hairstyles, use makeup to ‘contour’ our faces and different lighting causes shadows, plus things like scarves and hats etc will complicate it.

It’s also the latest thing in home security in U.K. with a pin point camera at your front door linked to an app on your phone, with remote access capability to let people into the home or other premises in the owner’s absence. Though not strictly facial recognition it appears to be getting people used to the idea.

U.K. airports were supposed to be using retina scans on people for security reasons but it appears to have been quietly abandoned. Same goes for fingerprints as they can be altered which negates using them for security.

Voice recognition also has downsides as anyone who uses voice control of their phone in a car for instance can testify. Background noise and other factors, like having a cold, can affect the use of these things.

A lot of these things sound good in theory but don’t always work in reality. Bit like autonomous vehicles. They’ve got a long way to go until the tech actually works properly.