WEBVTT - Amazon Helps Law Enforcement with Facial Recognition.  Drew Harwell talks to Armstrong & Getty.

0:00:00.240 --> 0:00:06.320
<v Speaker 1>Drew Harwell is an actual recognized authority on various things technologically,

0:00:06.480 --> 0:00:09.719
<v Speaker 1>covers technology and artificial intelligence for The Washington Post and

0:00:09.800 --> 0:00:14.000
<v Speaker 1>joins US Now to talk about the coming facial recognition

0:00:14.560 --> 0:00:17.520
<v Speaker 1>wave across America. Hey, Drew, how are you? Hey? Guy's good?

0:00:17.560 --> 0:00:19.920
<v Speaker 1>How are you good? That was an extremely worthy introduction,

0:00:19.960 --> 0:00:23.840
<v Speaker 1>but I had pulled it off. So listen. I think

0:00:23.880 --> 0:00:25.680
<v Speaker 1>a lot of us have followed some of the documentaries

0:00:25.720 --> 0:00:29.960
<v Speaker 1>about facial recognition in China and it's it's pretty cool

0:00:30.040 --> 0:00:33.280
<v Speaker 1>in a weird way, but pretty scary. What's the newest

0:00:33.479 --> 0:00:36.879
<v Speaker 1>on facial recognition in the US? Yeah, I mean you're

0:00:36.920 --> 0:00:40.839
<v Speaker 1>seeing it everywhere. You're seeing it in um schools and companies,

0:00:40.880 --> 0:00:42.760
<v Speaker 1>and you know, it's still sort of rare, but it's

0:00:42.840 --> 0:00:46.000
<v Speaker 1>it's increasing pretty quickly. And the big sort of new

0:00:46.040 --> 0:00:50.240
<v Speaker 1>front is for police departments in sheriff's office, where they

0:00:50.320 --> 0:00:53.479
<v Speaker 1>are allowing sort of deputies and investigators to use it

0:00:53.520 --> 0:00:57.200
<v Speaker 1>to identify people that they meet out on on the

0:00:57.240 --> 0:01:00.240
<v Speaker 1>streets and in the field and in their way maybe

0:01:00.320 --> 0:01:03.640
<v Speaker 1>lead to new arrests and new investigations. How would that

0:01:03.680 --> 0:01:08.199
<v Speaker 1>work exactly in practicality, So you know, if somebody shoplifts

0:01:08.200 --> 0:01:11.000
<v Speaker 1>out of the store. They're often caught on security or

0:01:11.040 --> 0:01:15.000
<v Speaker 1>surveillance cameras. The people who called the police, they sort

0:01:15.000 --> 0:01:16.959
<v Speaker 1>of have the footage, but they don't have any quick

0:01:17.000 --> 0:01:20.920
<v Speaker 1>way to identify the people. So now these deputies can

0:01:20.959 --> 0:01:23.280
<v Speaker 1>just sort of take that image, run it through their

0:01:23.280 --> 0:01:26.880
<v Speaker 1>facial recognition which scans all of their sort of jail

0:01:26.959 --> 0:01:31.560
<v Speaker 1>mug shots, and you know, okay, instead of saying that

0:01:31.640 --> 0:01:33.120
<v Speaker 1>he was a guy, he's I don't know, he's about

0:01:33.160 --> 0:01:35.399
<v Speaker 1>six ft tall, probably way too und nay bonds. Well,

0:01:35.400 --> 0:01:38.560
<v Speaker 1>they just they just do the facial facial recognition thing

0:01:38.560 --> 0:01:40.319
<v Speaker 1>and say it's Jimmy Smith, he lives over here, and

0:01:40.360 --> 0:01:42.679
<v Speaker 1>you go get him. Well, does it have to be

0:01:42.760 --> 0:01:45.840
<v Speaker 1>somebody who's been in the system. Yeah, that's right. So

0:01:45.959 --> 0:01:48.880
<v Speaker 1>it's photos right now, and it's everybody who's been sort

0:01:48.880 --> 0:01:51.120
<v Speaker 1>of arrested. Will all be in the system before you know,

0:01:51.240 --> 0:01:53.440
<v Speaker 1>it will all be in the system before you know,

0:01:53.600 --> 0:01:56.280
<v Speaker 1>if we aren't already. Yeah, that's right. I mean, you

0:01:56.320 --> 0:01:58.240
<v Speaker 1>know we have to think that there are you know,

0:01:58.280 --> 0:01:59.960
<v Speaker 1>the d m V has photos on all of us

0:02:00.040 --> 0:02:02.400
<v Speaker 1>or passport photos of the databases are out there. They're

0:02:02.440 --> 0:02:04.840
<v Speaker 1>just not yet connected to those systems. But you know,

0:02:04.880 --> 0:02:07.040
<v Speaker 1>the question is could they ever be and could we

0:02:07.080 --> 0:02:11.680
<v Speaker 1>ever be identified like that? Yes? And yes, yeah, you know,

0:02:12.760 --> 0:02:18.200
<v Speaker 1>there's a fine line between uh aware and paranoid, but

0:02:19.040 --> 0:02:21.239
<v Speaker 1>you know there and we reference this on the show.

0:02:21.280 --> 0:02:24.079
<v Speaker 1>Semi regularly drew that there are so many federal crimes

0:02:24.080 --> 0:02:27.480
<v Speaker 1>nobody even knows how to count them, including the regulations

0:02:27.520 --> 0:02:29.200
<v Speaker 1>you can violate which end you up in jail, and

0:02:29.200 --> 0:02:31.480
<v Speaker 1>the rest of it. And I could see a push

0:02:31.560 --> 0:02:33.800
<v Speaker 1>made that if you get a parking ticket, if you

0:02:33.840 --> 0:02:37.880
<v Speaker 1>have an overdue library book, if you do anything, you

0:02:37.919 --> 0:02:41.360
<v Speaker 1>step out a line, we just we're gonna take your picture.

0:02:42.000 --> 0:02:44.880
<v Speaker 1>That's it, and you get entered into the system, and

0:02:44.919 --> 0:02:49.079
<v Speaker 1>then everybody's in the system. Yeah, and that's not really

0:02:49.160 --> 0:02:51.639
<v Speaker 1>science fiction anymore. I mean that a way. A part

0:02:51.639 --> 0:02:53.560
<v Speaker 1>of that happens in China right now, where they have

0:02:53.840 --> 0:02:56.960
<v Speaker 1>sort of facial recognition camera set up at intersections. If

0:02:57.000 --> 0:02:59.320
<v Speaker 1>people jaywalk, it sort of takes a picture of on

0:02:59.400 --> 0:03:03.120
<v Speaker 1>it has their identity, and then they're they're named and shamed. Effectively,

0:03:03.440 --> 0:03:05.760
<v Speaker 1>their their pictures and identities are put up in these

0:03:05.800 --> 0:03:08.600
<v Speaker 1>like public squares. Is sort of like you know, moments

0:03:08.600 --> 0:03:11.480
<v Speaker 1>of embarrassment. So that's not happening here obviously, But that's

0:03:11.520 --> 0:03:14.160
<v Speaker 1>just a sign of you know, how this technology could

0:03:14.160 --> 0:03:15.919
<v Speaker 1>be used. I mean, it's it's it's a way that

0:03:16.120 --> 0:03:19.560
<v Speaker 1>deputies and police could identify any of us, um without

0:03:19.600 --> 0:03:22.280
<v Speaker 1>us really knowing or consenting. And so there's all sorts

0:03:22.320 --> 0:03:25.320
<v Speaker 1>of questions of privacy and whether we really want that

0:03:25.320 --> 0:03:27.720
<v Speaker 1>that future ahead of us. How good is it right now?

0:03:27.800 --> 0:03:31.560
<v Speaker 1>The technology on facial recognition? Is it pretty accurate most

0:03:31.600 --> 0:03:35.000
<v Speaker 1>of the time. It all depends on the quality of

0:03:35.000 --> 0:03:38.280
<v Speaker 1>the image and the images they're looking at, like their

0:03:38.360 --> 0:03:42.080
<v Speaker 1>database of facial images. Right, So if all of a

0:03:42.120 --> 0:03:45.280
<v Speaker 1>sudden does it not work, yeah, I mean that's a

0:03:45.280 --> 0:03:48.119
<v Speaker 1>good quay it really, I mean, yes, it can still

0:03:48.160 --> 0:03:49.920
<v Speaker 1>work in that and a lot of the deputies will

0:03:49.960 --> 0:03:51.920
<v Speaker 1>say like, oh, you know if even if somebody puts

0:03:51.920 --> 0:03:54.600
<v Speaker 1>on glasses or they grow out a beard, you know,

0:03:54.680 --> 0:03:57.360
<v Speaker 1>this can still tell. I mean, it really depends on

0:03:57.520 --> 0:03:59.520
<v Speaker 1>how good the images. And you know a lot of

0:03:59.560 --> 0:04:02.200
<v Speaker 1>these survey and lets photos are taken from in the

0:04:02.240 --> 0:04:06.720
<v Speaker 1>ceilings at weird angles, so those tend to be less accurate, um,

0:04:06.840 --> 0:04:09.040
<v Speaker 1>and you know they're they're taken in the darker by

0:04:09.240 --> 0:04:12.520
<v Speaker 1>you know, crappy smartphones so it all depends. But you know,

0:04:12.840 --> 0:04:16.599
<v Speaker 1>in this scenario, anything less than perfect accuracy is going

0:04:16.640 --> 0:04:21.760
<v Speaker 1>to lead to deputies potentially arresting the wrong right. Horrified

0:04:21.800 --> 0:04:24.240
<v Speaker 1>by all this, but the upsides, you know, for the

0:04:24.279 --> 0:04:26.640
<v Speaker 1>things for good. If we could do away with keys

0:04:26.720 --> 0:04:28.880
<v Speaker 1>more or less, I'd be happy that you could walk

0:04:28.920 --> 0:04:31.080
<v Speaker 1>into your building where you work with your face, or

0:04:31.080 --> 0:04:33.159
<v Speaker 1>your house, or start your car with your face. You

0:04:33.240 --> 0:04:35.600
<v Speaker 1>just don't need keys. That that'd be awesome. Drew Harwell

0:04:35.680 --> 0:04:38.760
<v Speaker 1>covers technology for the Washington Post and is on the line, Yeah,

0:04:38.800 --> 0:04:40.400
<v Speaker 1>that that would be kind of cool. And I think

0:04:40.400 --> 0:04:43.440
<v Speaker 1>we're moving in that direction. Hey, one thing you mentioned

0:04:43.440 --> 0:04:45.760
<v Speaker 1>in your piece, which is quite be you. You were

0:04:45.800 --> 0:04:49.320
<v Speaker 1>too fat, You didn't used to be somebody differently, let

0:04:49.320 --> 0:04:52.480
<v Speaker 1>myself go. One thing in your piece that will have

0:04:52.520 --> 0:04:54.800
<v Speaker 1>a link to so folks can find it easily is

0:04:54.880 --> 0:05:00.760
<v Speaker 1>that they've actually um run artists sketches through a software.

0:05:01.440 --> 0:05:04.599
<v Speaker 1>Mm hmm. Yeah. And that's not a way to see

0:05:04.600 --> 0:05:07.359
<v Speaker 1>if that looks like anybody. Yeah, And I mean you

0:05:07.400 --> 0:05:09.080
<v Speaker 1>have to wonder is that really going to give you

0:05:09.160 --> 0:05:11.560
<v Speaker 1>good results? Like you know, the system is designed to

0:05:11.600 --> 0:05:14.559
<v Speaker 1>compare one photo to the other. When you start getting

0:05:14.560 --> 0:05:17.720
<v Speaker 1>into artist sketches it. It all depends on how is

0:05:17.800 --> 0:05:19.719
<v Speaker 1>the how is the artist feeling that day, and how

0:05:19.800 --> 0:05:22.479
<v Speaker 1>good was the drawing right, and and so you know

0:05:22.640 --> 0:05:25.760
<v Speaker 1>that's gonna up the the danger of it being a

0:05:25.800 --> 0:05:29.080
<v Speaker 1>misidentification to all these people are saying we really shouldn't

0:05:29.120 --> 0:05:31.839
<v Speaker 1>be using them that in that way. But you know, Amazon,

0:05:31.880 --> 0:05:34.000
<v Speaker 1>for their point says whatever, I mean, this is a

0:05:34.000 --> 0:05:36.280
<v Speaker 1>tool for the deputies. They get to decide. It's it's

0:05:36.400 --> 0:05:38.200
<v Speaker 1>it's all human judgment at the end of the day

0:05:38.200 --> 0:05:40.200
<v Speaker 1>because they get to look at the photos and the matches,

0:05:40.240 --> 0:05:42.520
<v Speaker 1>at the system shows up and get to choose. But

0:05:42.920 --> 0:05:44.880
<v Speaker 1>you know, I think there's a question of whether, hey,

0:05:44.960 --> 0:05:47.080
<v Speaker 1>is this system being used for for too much? Is

0:05:47.080 --> 0:05:49.480
<v Speaker 1>there is their mission creep here where we're now just

0:05:49.520 --> 0:05:52.760
<v Speaker 1>sort of throwing photos into the certain because it's easy.

0:05:53.279 --> 0:05:55.760
<v Speaker 1>Are well cover tech for the Washington Post? True? Really

0:05:55.760 --> 0:05:57.720
<v Speaker 1>interesting stuff. Again, we'll have a link to your piece.

0:05:57.760 --> 0:06:01.440
<v Speaker 1>Thanks very much for the conversation. Thank you. There's no

0:06:01.520 --> 0:06:05.320
<v Speaker 1>stopping any of this, all our information unless we fight

0:06:04.880 --> 0:06:07.520
<v Speaker 1>for ZENA or faces. Everything is going to be everywhere,

0:06:07.520 --> 0:06:09.480
<v Speaker 1>all the time. The government will track everything we do

0:06:09.600 --> 0:06:13.239
<v Speaker 1>all the time. That is absolutely gonna happen in my lifetime? Chilling?

0:06:13.440 --> 0:06:15.840
<v Speaker 1>What's coming up in your news? Marshall Philips, President Trump's

0:06:15.839 --> 0:06:19.679
<v Speaker 1>former lawyer Michael Cohen one last verbal blast before checking

0:06:19.760 --> 0:06:24.000
<v Speaker 1>into jail, got a startling admission from Boeing about its

0:06:24.080 --> 0:06:30.640
<v Speaker 1>Max Jets, And yet another organ gets drowned. Another Oregon oregan,

0:06:32.960 --> 0:06:36.680
<v Speaker 1>What is what the fascination with drones? I don't know.

0:06:37.360 --> 0:06:39.120
<v Speaker 1>I don't know if you can work a drone into

0:06:39.160 --> 0:06:40.839
<v Speaker 1>what it makes the new wake me when the novelty

0:06:40.920 --> 0:06:41.480
<v Speaker 1>is worn off