Facial Recognition Tech Wrongly Identifies Thousands as Potential Criminals

Megalith

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
Police in Wales adopted facial recognition last year in an attempt to catch more criminals, but it hasn’t been particularly accurate: data on the agency’s own website revealed that 92% (2,297) of those were false positives. Despite the high error rate, the force still managed to make 450 valid arrests using the technology.

The force said it had considered privacy issues “from the outset” and had built in checks to ensure its approach was justified and proportionate. However, the civil liberties campaign group Big Brother Watch criticized the technology. In a post on Twitter, the group said: “Not only is real-time facial recognition a threat to civil liberties, it is a dangerously inaccurate policing tool.”
 
Lot of the same news lately but no one should be surprised. All liberties, abilities to conduct transactions, move freely in society will be removed if your not a one world ecumenical citizen. I remember a few leaders that used false enemies and terror to motivate its citizens, Nero being one of the many.
 
Some day it'll be as easy as typing: Computer arrest everyone at the protest yesterday
 
Yea....... if people don't think they are constantly being viewed by CCTV, I've got news for ya. They already have you in their DB doesn't matter who you are, they will trip you up if you some how break the law.
 
Instead of saying - "this technology doesn't work, it's too expensive, and there are better ways to do law enforcement that doesn't infringe on personal privacy" - They'll just say "The technology shows promise and just needs some refinement. We'll just have to deal with the roadbumps as we come to them."
 
i would love to hear the real justifications for such broad and intrusive monitoring systems.

it could be even as simple as .. yes we are too lazy to get up out of our chairs, and this makes it so we dont have to.. kinda thing
 
viper1152012 said:
Wow 8% accurate... I could just watch monitors and guess and have higher percentages
Click to expand...
I highly doubt it. I'm humble enough to admit that a computer can go thru a large dataset much faster than my human self. The UK has about 65 million people. There are just over 30,000 people wanted for arrest on outstanding warrants in the UK according to google. That's about 0.05% of the population. 8% vs 0.05% is about 160:1 ratio. So if you just randomly guessed.... you'd be highly unlikely to reach 8% unless you really can judge a book by its cover and all criminals look alike.

Some people with autism and certain interesting brain developments are geniuses at image recognition and reproduction though. So I can't discount your ability to do so for sure. First of all, if I ask you to spot in this crowd, just one person's face, I would be curious to see how you do. There's 100 people or so.

otball-game-of-ukraine-championship-at-valery-lobanovskyi-stadium-on-may-10-2009-in-kyiv-ukraine.jpg


Now find the 3 men who most look like this guy:

0fe0c87273b6540713fc3bb76151efa9.jpg


Maybe, you can find a guy or three who looks almost identical. However, that's keeping just 'one' image in your head as you scan a crowd. If I said, here's 1,500 mug shots of all the people who have outstanding warrants...I think there's much less chance you'd be able to keep all 1,500 mug shots clear in your head as you scan the crowd and accurately compare each individual face to all 1,500 mug shots to see if any one of those people with warrants are in this Stadium. Even after a full year, I am not sure if after a full year you'd have found 450 of them even if all 1,500 attended at least 3 events in the Stadium. As you'd be looking thru 30,000 faces on maybe 365 separate days or have 10,950,000 faces to examine against the 1,500 images.

I imagine many humans would see someone and it would spark a faint memory because they are similar enough to one or two of the mug shouts and go ... 'wait a sec...he's in this pile of mugshots, I think. Just detain him for a sec while I find it...I'm sure I saw him...huh...flip, flip, flip, flip thru the mug book you know...no, I'm sure he was here somewhere near the middle? maybe? ah, perhaps front more...no no no no no, no...huh, umm nm. They look very similar but it's not him. Please let him go.'

It's a miracle the machine learning system is able to even reach 8% and do 450 arrests in a year. It's probably far far better than I could do. I don't think I could pick out 450 faces out of a set of 10,950,000 that accurately if you just gave me a book containing 1,500 photos of the people I'm looking for and a lot of effectively still photos of random places of people.

Like any machine learning system, it would probably reach 10%, with more data, the following year and I'd still be at 0.004%. Then 12% as I move up to 0.005% and then it would get a hardware upgrade able to do twice as many operations plus software developments and reach 20% while I'd only be at 0.006%. Etc, etc.

It is definitely a slippery slope of the liberties of people and the scary nightmarish scenario is warrants issued and followed thru using this system for people who are labelled criminals for different political views than the current party in power.
 
I'll leave this here in case anyone has a run-in with the cops with tech like this.
 
But it is 100% accurate in China, they never get it wrong. /s

China plans on having over half a billion CCTVs in the next few years (already have over 100 million). The Communist Party is doing it's damnest to stay in power and keep dissension down. I wouldn't be surprised to see another revolution there in my lifetime once the people decide enough is enough and tell the government to fuck off.
 
I have had experience working in surveillance at a casino and was tasked with finding people on wanted lists so it not random street images, so I'll grant you that. But I never had a false apprehension so I got that going.
I am sure they are pulling numbers out of their ass like using population stats with bias.
The biggest issue I saw was arrests vs false arrests.... One night of wrongful imprisonment can screw up a lot.

Hit a nerve for me.
 
You must log in or register to reply here.
Back
Top