How AI Facial Recognition Works

a production of I Heart Radios How Stuff Works. Hey, welcome to the podcast. I'm Josh Clark. There's Charles Scooter, Computer Bryant and Jerry Um Matthew Broderick in War Games Rolling and this is Should Know. That's good. Thanks. What's your name, Josh m hmm? Okay, Josh, Ali Shedy in More Games Okay Clark, I know Aaron Cooper is doing right now? Man? How cute was she in that movie? Like? I think they designed that movie for like every thirteen year old boy in America's fall in love with Ali

shed I think you're talking about um short Circuit. I never saw that, believe that what with Johnny five. I mean, I know the movie. You gotta see it. Really, it's pretty awful, especially with um oh, what's his name? Who is the sleeves ball from Fast Times at Ridgemont High? Who is the ticket scalper? Yes, he plays a Indian like Asian Indian character, like full on brown face and everything. It's really bad. The movie is bad enough, but then now when you go back and see that, to your like,

I can't believe this. I can't believe it. I think he's Italian, oh easily, maybe Jewish or maybe just a straight up white guy. He's definitely not Asian Indian. No, he's not. But anyway, go see Short Circuit. Okay, see what you think Ali shed He just keeps looking at the camera going, I'm so sorry that was a big hit though she didn't have anything be sorry about. Um So. I guess War Games is kind of in your your wheelhouse. Yeah,

it was a little old for me. Yeah, because I saw that when I was like twelve, bish uh, and you would have been I don't know how how much younger are you. I would have been seven seven. Yeah, that's a little young for War Games, I would think. I mean, I still watched your wherever, but I was like, yeah, that's hilarious. The computer is called Whopper. Yeah, I mean

it was right in my wheelhouse, right. I remember at the end of war Games, Uh, they lock in, you know, they're decoding the the code one number and letter at a time. Very suspenseful and yeah, very suspenseful. And it finally locks in and me and my friends memerized it so we could go home and plug it into an Apple too to see what happened. Nothing, Okay, how do you plug in a number anyway? What does that even mean? Oh? Yeah,

that's true, you know. Yeah. I remember a very rudimentary program you could run where you could type in like four lines of whatever I don't even know if you call it code with a phrase, and it would run the phrase like a thousand times all over your screen and a big scroll. And then that just thought that

was the coolest thing ever. I feel like, I remember what you're talking just like five lines was like the only part I remember is twenty go to ten and ten was the phrase I think something like I don't remember. Then I was like, man, let's just play Castle Wolfenstein. That was a good one. I remember, did Oregon Trail. I never did as well. I was Castle Wolfenstein, but like wolf and signed on the PC. Oh yeah, like move left arrow, right arrow, shoot uh shoot dash, which

is some sort of a bullet. Yeah. It was fun. That was fun, and I thought it was just like the height of technological gaming. It was it was at the time. But now, Chuck, we've reached the height of technology which is being tracked everywhere you go, look at you all the time by whoever wants to do that. I'm gonna change your name to Josh smooth up a Raider Clark for that transition. Very nice. Yeah, Like I like Ali Shedy Clark, Josh, Ali Shedy and Clark. Yeah,

this is a good one. Uh did you put this together? It was this you and this is Dave. Dave Brews good stuff in high Dave. We finally got to meet Dave and his family, lovely family that we cursed awfully in front of in Seattle. I felt terrible, thanks for adding yourself to the max and he was like, you know whatever, he was fine. His kids were adorable, they were they were great. They couldn't look at me though that really they were probably just him did by your present,

No No, it was because I cursed so badly. So this is good stuff. Though. Facial recognition technology that they've been kind of at since the nineteen fifties, which they rolled out as a test in two thousand two at the Super Bowl in New Orleans, did not go that well. No, it was a little clunky back then, but it's gotten

a lot better since then. Let me explain why. For anyone who's listened to the End of the World with Josh Clark the AI episode in particular, everything associated with artificial intelligence got way better starting around two thousand seven, when neural nets became a viable form of machine learning. Because you don't have to train a computer what constitutes

a human face and what to look for. You just feeded a bunch of pictures of faces and say these are human faces, learn what a human faces and they trained themselves. And so around about two thousand seven, two thousand eight, two thousand nine, everything had to do with machine learning got way smarter because we started using neural nets and facial recognition software is no exception. Yeah, and there were a few things that kind of converged all

at the same time or around the same time. Uh, social meds kind of coming on the scene right in that wheelhouse was a big deal. Um, Facebook, this is staggering. Facebook just by itself processes three hundred and fifty million new photos through its uh facial recognition software every day, and every time one comes through. Mark Zuckerberg goes like, you think it's neat when you go when you put a picture up and it says like would you like

to tag Emily your wife because that's her? And you think, oh, well, that's super easy, thanks Facebook. But then you don't think, like, wait a minute, all right, how do they know that's my wife? And you know, it's like with everything else, Uh, there's privacy people that were like, whoa do you guys realize what's going on? And then of the sheep they're like huh no, no, they're like no, it's great, Like I don't have to go in and like make click two links or two buttons to take the way easier.

So that was one thing. There's way more photos out there for those machines to learn one like good high quality photos right uty million a day just on Facebook alone, um, which means the machines were getting smarter, They were getting better and better at at training themselves. And then lastly, um, that has led to a ubiquity in in facial recognition. That the better the machines have gotten, the easier it has been to put together data sets for them to

train on, which is lots and lots of pictures of people. Um, the cheaper the technology has gotten, which means the more people that are now using facial recognition than ever. Yeah. Amazon has a service called Recognition with a K, which is not a good look. No, it looks very German. There's something about replacing at sea with the K. It just looks creepy, Like when you spell America with a K, it means something. It means like bad America. Yet they

went right full steam ahead and called it recognition. You have to say it like that, you do, I think, and you have to be like squeezing the air out of a syringe while you're saying it too. Um, so they have this. I didn't even know about this, but it's ubiquitous and it's not super expensive, and that means that the law enforcement agencies agencies they don't have to like create their own they can just say, well, just

to sign up for a recognition right exactly. Because it's there, and because it's it's relatively cheap, you can just get a subscription, not just law enforcement agencies. If you have a photo sharing app or whatever and you want some facial recognition technology, you just contract with with Amazon, and Amazon goes here, you go, here's our data, and you can or um like our code, and you can put

it onto your platform. Anybody can use it. Um. So it is kind of everywhere, and that makes a lot of people, including me, very very nervous because as this guy wood Row hurt Zog, if he's not Werner hurt Zog, brother, I'll be disappointed. But but a wood Row and a Werner come on, you know anyway. Um. Wood Row hurt Zog as a professor of computer science, and I believe privacy,

civil liberty. Um. He basically says, Look, there is no way we're going to reap the benefits of facial recognition without ultimately sliding irreversibly into a dystopian surveillance state where it's happening right now, and if we don't do something

about it, it's never going to change back. We're about to fully give up our privacy because it's one thing to have your phone tracked you can give up your phone, get yourself a burner phone, like you're Jesse Pinkman or something like that, and then you just throw that phone away. You can't be tracked anymore. You can't get a burner face. And if that does become a thing down the line, it'll be very, very expensive. So the average person can't get a burner face. Will be tracked by our face

everywhere we go. And as we add more and more cameras and this technology becomes cheaper and cheaper, it's we will be living in a world where there will be zero privacy and will be monitored and tracked because it will be so easy, and it will be sold to us like it's being sold to us now that it's a law enforcement tool to get the bad guys. But it's eventually going to extend to include everybody. But what do you have to worry about. You're an upstanding citizen.

It doesn't matter if you're tracked. That's not true. That's just not the case, everybody, It's not case. All right, we're gonna call that soap soapbox soliloquy number one of what I guarantee will be probably three or four. Uh, let's talk of a little bit about how it works. It is biometric authentication. Um. It's like a fingerprint or

a retina scan. And basically what it does is, uh, it is precise measurements of a face to calculate every person's very unique visual geometry, like how far apart your eyes are, how how far apart your pupils are from your nostrils. Yeah, your facial geometry, how your face is all set up? I think. Yeah, it's even gotten into things like facial hair, skin tone, um, skin texture. Yeah, I'm sure it'll get just more and more specific. Yeah.

And then you know, because they're getting the machines are getting better and better and easier and easier to train on this stuff, you can just add more and more data to it, and the the the recognition will just become increasingly um um good. Yeah. And if you want to throw off facial recognition software and freak out every human you meet to shave your eyebrows, oh yeah, that would be a little freaky. Have you ever seen that. You've seen it before, I'm sure in movies and stuff.

It's an interesting thing. I remember a kid in industrial arts class did that one year. He was like a little you know, kind of a ninth grade burnout, and he just showed up one day with no eyebrows, like you would I think like not having a nose would be more easily accepted. There's something just uncanny when someone shaves their eyebrows, Like one day they have them, the next day they don't. Was it like immediately recognizable what the thing was, or was it like that's the thing

off today? Whereas if you, you know, you came in the next day without a nose, the first thing you would say is what happened to your nose? What happened to your nose? Todd? Yeah, and Todd would be like, I can't rent of it. I fell off. So those measurements we were talking about. What happens then is they compare that just like a fingerprint, with a database of images.

And depending on what this is for, it could be like just within your company, or it could be the FBI's database of mugshots, or it could be the d m VS database of driver's license photos yeh, which we'll

get into. Yeah. What's interesting is each stage of the way, there's a different algorithm that does you know, each increasingly sophisticated step, until you finally have basically like all of the different data points for that, you know, what makes up that facial geometry, and then you can compare it to all the other data points. We think of like a computer running like a you know, a picture, You've got your input picture and then running you know, all

the pictures next to it. That's not what it's doing. It's it's running the numbers. Basically, it's doing computer stuff. Yeah. I love that first step, which is you have to teach the computer what a face is. So, I mean, seems silly, but of course that's what it is. Well, yeah, because I mean if you show it a picture of a person standing next to a fire hydra shaming on the fire Hydran, Yeah, and say hello, handsome, So this

is what a human face looks like. Yeah, or no, that's a butt, And then it starts, you know, closer, closer. All right, now that's a face. You know what a face is. Now move on to step two, which is stop screwing around. Yeah, so now you know what a face is. You've got to normalize it for the photo, which means there are not that many. Well, let's you

have to put it in the dockers. That normalizes, Um, you isolate that face, and then you have to make sure that it's normalized as far as looking at the camera. So if it's a if you get a photo of someone from a CCTV, let's say, and it's sort of a three quarter, they have the ability to make it as if it is looking straight at you. Yeah, the computer can pretty accurately predict what the rest of the face looks like face you know it on, I guess

face on um. And when it normalizes it like that, it makes it much easier to compare to other pictures because, as we'll see, most of the pictures or most of the data points that it's comparing it to, are taken from databases of pictures that have been taking of people face on. Right. So that's why it wants to go to driver's license. Yeah, well just spoiled it. But yes, we already said that. Oh we did, Yeah, I did, Okay, I missed it, I know. So uh from there, do

you have more algorithms still that isolate parts of the face. Um. And this is where my old theory that like, there are only so many sort of facial combinations, so that's why you have doppelgangers. We got to do an episode on doppel Gan. There's only so many things you can do with two eyes to eyebrows and nose in a mouth and cheekbones and in chin. Well, okay, what else? I mean, there's not a whole lot. There's lips, lips, sure, what about um uh uh uh, that's about it. These

are called elevens, the ridges between your eyebrows. Well, if you want to get super specific, but that's what I'm saying. I think they're getting more and more specific. Oh yeah, yeah. But my whole point is, and we'll learn and here and facial recognition. They do use Apple gangers, but put a pin in that so they recognize all these features and then each feature becomes what's called a nodal point or no Dowell point. I think, nodal nodal, I think.

And this is where you're gonna get your super exact uh angles and distances between all these parts as a flat two dimensional thing. Which my question was because below here, you know, it talks about Apple and their iPhone have a three D facial recognition. Is that is two dimensional superior to three D? I don't know, or is it just because that's what all the pictures are in the databasis, so that's what they do. I don't know. All I know is my phone usually unlocks when I look at it.

You know what? I Hey, that's having to take off my sunglasses. Worst so I found I've got some wayfarers that I don't have to take, but my aviators I do have to take off. Interesting do they keep trying to make you into mav when you have on the the go ahead? What is that? That was Tom Cruise laughing and chewing gum. Okay, wow thanks, I feel like okay, um we we we gotta keep going because I was

about to take a break unnecessarily. So um when the when the computer is running through the pictures, it just sis soon goes like no, no, no, no, no, millions and millions of times and then finally goes yes. But when it says yes and it spits out another picture, it's not like this is that person? No you want

it to be. Because we all watch n c I S, we all watch c S I, we all watch um Law and Order, we all watch um uh party down uh Andy Griffith, that Mattlock, the whole CHAMAI um, so we wanted to just spit out and be like, here's your here's your your person of interest, right. But what it's really doing is it's it's producing a similarity score that is probabilistic. It's saying this is there's this percent chance that this is the same person as the picture

or the person in the picture that you uploaded. It's a bit of a guess. It is, so a sophisticated guess it is, and the better computers get at this, the likelier it is that if they say this is a there's a chances the same person that it's the same person. But it's as we'll see, it's up to the human user to determine what is an acceptable threshold of a confidence Is it no? Is it no? Frankly, it really should be about or higher should be the

confidence setting for the confidence level. Didn't that what Amazon's recognition says the threshold should be. I'm glad you said that, man, because it really is creepy and I couldn't put my finger on it. And it's exactly I mean, I knew the k looked weird or whatever, but it hadn't hit me that just how creepy it is and just how off the mark or potentially on the mark that name is. Like, Oh, like, if my name was spelled c h u K, I'm sinister a little bit, you'd be more sinister. Yeah, I

don't think you could ever be truly sinister. I appreciate that. Alright, let's take a break. I'm gonna go work on sinistering up a bit, and we'll talk a little bit more about some of the uses of f R right after this. So, as with all technology, it has to be abbreviated into two letters, the second of which is are do they call it f R? I've seen it, Yes, I was just silly, but it doesn't surprise me. Nope. So in f R facial recognition technology, there are there are some

some beneficial uses for it. Yeah, Like we said, you got to tag people is chief among them. The for people like you and me, that's the pinnacle as it stands. You don't have to tag people yourself. Facebook does it for you. That's what we're trading everything for. I gotta calm down. Okay. There are some other, like genuinely beneficial

uses too. There's a nonprofit company called Thorn that scans missing person's pictures against um pictures of children in child porn videos, um or suspected human and trafficking to to get matches, and apparently they've rescued a hundred kids so far from using that technology. There's a pretty beneficial use of facial recognition software. Uh dating apps. Let's say you want to you can get pretty specific on what kind

of face you find attractive, which is interesting. But you can say I really think, um, I like guys with high cheekbones, and but no, you would go find mall lips. It would be more like, um, somebody could be like, oh, I really find Christian Bale attractive, and they get a picture of Christian Bale into this dating app and I would come up. But I wouldn't because I wouldn't be in the dating app because I'm happily married. Do you

think you look like Christian Bale? I'm told that a lot. Really, Yeah, that's weird. Don't think you look anything like it. I don't either, but people say interesting. I'm I don't know what I would do if I was dating now. I guess I would just go to a service and say and aliety type sure for more games era, But they'd be like, okay, sir, you would just upload the picture.

You don't have to come into the office, which is really not even open to the public, and just tell us you're interested in Ali sheety type like a weirdo. You mean, dating apps don't have offices where they just feel complaints and interested parties. You sit down and they they videotape you with the VHS camera, put you on with some other guys on the tape. That's how they used to do it. Oh yeah, that was one of

the subplots of Singles, the Cameron Crow movie. Oh yeah, it was Expect the Best was the name of the dating service, and you would make a videotape and like a watch video tapes of people you know, saying who they are. How do you remember that? I was a big singles fan. That's not a bunch I got you. Yeah, Expect the Best. You me pack of cigarettes and some coffee. We don't need anything else becausezoom tight. So what else here?

This was? UM Taylor Swift and and her security team on tour used it to scan the audience to see if any of the creeps who have harassed and doctor were in the audience. That's super beneficial. No one should have to go through that. UM also, cops use it in myriad ways, but in particular especially beneficial or um when they use facial recognition to identify people who can't

identify themselves. Somebody in the midst of a psychotic break, perhaps somebody wasted on troom's um, somebody you're not Jesus Christ, who who has uh amnesia, Our friend Benjamin Kyle, who apparently knows who he is now but he's decided not

to disclose it publicly. Remember the guy he was found behind a burger king near a dumpster, had zero recollection of who he was or how he got there, and like there was this international publicity publicizing like who he wasn't that he couldn't remember who he was, and somebody finally came forward and identified him. So now he knows who he was, but he went like a decade without knowing. By the way, when I said, sorry, you're not Jesus Christ, I was making fun of the guy on mushrooms, not

someone in the midst of a psychotic break. I just want to be very specific. I think that was very clear a right to make sure everybody knows that. So uh. Those are some of the good ways that it can

be used UM. Now let's talk about all the bad ways. Yeah, I mean when you're talking about the government, you're talking about law enforcement, when you're talking about things like what's going on allegedly in China with CCTVs everywhere trained to UM to single out ethnic minorities and religious groups just walking down the street going about their day. Yeah, that's it gets into much different territory than tagging people in

dating apps. Yeah, it's pretty difficult to attend your religious service if you're not allowed to attend your religious service and you're being tracked everywhere you go. Yeah, that's why places like and this is the most predictable thing in the world San Francisco, Oakland and Berkeley and then Somerville, Maine. I knew the Manners would be in there. They're not

into this, that's right. They have banned law enforcement UM from using facial recognition altogether in California UM as a state, and the ACE has put a three or moratorium on the use of it UM body cams, which and the a c l U is basically I know this is jumping ahead, but they're at the point where they're like, we need to tap the brakes here for a few years and like because there's no legislation about this yet

and it's just going full steam ahead. Yeah, I really I don't want to like like run past that there is aside from Berkeley, San Francisco, who's the other one, Oakland and Somerville, Maine. UM, there are no laws, state, local, or federal governing the use of facial recognition technology by law enforcement. It's just happening very fast. Whatever they want to do, they can do UM and in some cases they do all sorts of stuff with it. They will use it, like the NYPD very famously used UM what

you were talking about with doppelgangers. There was a guy who was caught stealing beer at a CVS. Not even a duane read a CVS. UM, and they said, but this guy looks a lot like Woody Harrelson. We don't have a good we don't have a good shot of him to use in official recognitions. So they went and got a pick of Woody Harrelson, and there they came up with a match, and they think it was the

guy on video and CVS and so UM. The Georgetown School of Law UH produced a study called Garbage in, Garbage Out, and they were basically like, that's not okay, you really shouldn't be doing that. But that's the level of legality as it stands right now. There's it's just open season, um, and uh, it's just basically whatever you want to do, you can do as far as facial recognition is concerned. In that story in particular, it's like some people are like, awesome, the system works. Sure. Other

people are like, what about poor Woody Harrelson. He was really in danger right then of being implicated in this beer stealing scheme from CVS and what he said? What dude, I love that guy man. True Detective the first season, Yeah, first four episodes just amazing. That's called using a probe photo when you use when you say, hey, that looks like someone. They also did the same with one of the New York Knicks. Apparently I could not, for the

life of me find out who. Yeah, it's like he's being protected or something that maybe no one said who it was. Um, a couple of numbers for you, though. Uh. The FBI receives about fifty thousand facial recognition search submissions a month for their database. So that's the other thing. If you don't have even the money for a subscription to Amazon Recognition, or you don't have an I T person who's capable of assembling and putting it, you know, using it UM. You can just submit these requests to

the FBI. So there's a lot of different avenues you could take his law enforcement to to UM to use facial recognition technology to catch suspected criminals. Yeah. I was about to say bad guys, but who knows. We'll see it's not always the case. So here's some more numbers though, because you know it needs to be regulated. But when it works, it really works. Yeah, it really does though.

It is the thing. Yeah, there was one department where they said it lowered the average time required for an officer to identify a subject from an image from thirty days to three minutes, which kind of brings home the point there's another number in here. This interesting, but uh, it brings on the point that like, this is something that human policemen were doing. Officers were doing with their eyeballs by flipping through books, yes, for thirty days straight,

saying like it doesn't look like this person. This is like a chance to really speed up that process and to spend more time in theory catching bad guys. Yes, I'm not arguing for it. I'm just saying they were doing this anyway, just through manpower. Right. I think the thing is is anytime you add artificial intelligence automatically makes the user of the artificial intelligence aside unfair unfairly advantaged.

It's not like the criminals are able to use AI to steal beer from cvs more effectively, but the cops are using AI to catch them stealing beer more effectively. And it's kind of like, yes, it makes sense to catch like child pornographers and um human traffickers and rapists and murderers and violent criminals with this stuff, but using that kind of technology catch somebody who stole beer from a CVS. If that's when it starts to feel like what kind of society are we moving towards? You know? Well,

I think someone not. And let me keep going here for a second, because I don't want people be like, what do you in favor of the guy stealing beer from CVS? No, I'm not. I think you're a scumbag if you steal beer from CVS. But I also think that it's overkilled to use facial recognition technology to catch that person. Use old fashioned police tactics or don't catch them. Yes, it's just kind of the fairness of the old western

New York City. I think I might be on the other side because I don't think we need to set a fair playing ground between criminals and cops and saying like it's unfair that cops can use this stuff and criminals are just out there not able to use these same techniques. Okay, So my the fairness thing doesn't just end at the law and order thing, right like, Like, it's not just with cops using it. They they have this huge advantage. I totally get how people would be like, no,

give the cops that huge advantage. I don't have an issue with that in and of itself. I think my issue comes a step or two down the road, sure, where the government or the cops, acting on behalf of the government, use that against everyday citizens who have no recourse whatsoever. That that that that lopsidedness that's so evident when you're using AI to catch somebody's stealing beer from

the CBS. It's really easy to kind of follow that a little further across to the horizon and see just how unfair life could be and how oppressive that could be using that technology. I think that's ultimately what I'm saying. I hopefully dug myself out of that hole by now so.

Uh And and this gets into some of the controversies and the arguments if you're if you're scanning mug shots uh for rapists and arsonists and murderers and violent criminals, and you're catching people, You're not gonna find a lot of people that say, well, that's not fair, go back and use take a month to look through a mug shot book instead, and waste a bunch of time and

don't be efficient. So I think most people would say if you're looking at mug shots, although we should point out that a mug shot doesn't mean that just means you're arrested. That doesn't mean you were guilty of anything. Um So, there are plenty of opportunities for false positives

and people being put in jail that shouldn't be. But but there's not a lot of people who are like, no, not use mugshot databases right Exactly, If you're scanning driver's license databases or other just general public databases, that's when it gets super tricky because we can't avoid the fact that what that means is in in the Center on Privacy and Technology, Um kind of stated very plainly what

that means is everyone is in a perpetual lineup. Essentially, if you have a driver's license, you're part of a police lineup. Yeah, whether you like it or not, whether you know it or not. And if that computer says here's the guy, it's Chuck Bryant, Uh, they will say, oh, he doesn't strike me as very sinister, or in the computer would be like, trust me, this is the guy. Uh,

with like an eighty something percent confidence interval. Um, Chuck, Suddenly you're going to get visited by the cops and maybe you'll even get arrested because you were a little KG when they talked to you and you set off their cop radar or whatever. And then the next thing you know, you're in court being charged with the crime that you didn't commit because the computer implicated you and

the cops thought that you were acting KG. And let's say that you were a very very poor person and you don't have any money to mount a decent defense. The best you can afford is a free public defender who has fifty other cases is not really paying very much attention to you and you're in jail now because you got convicted wrongly because you were putting a lineup

just because you had a driver's license. Yeah. I think for me, um, and this is total my privilege coming through as well, Like I'd want to see some numbers and if one of every ten thousand arrest and conviction of a real criminal or an a rapist and a murderer, and there's three people that get falsely identified and have to go through the system and may or it may

not be acquitted, I'd want to see those numbers. But again that's coming from a privileged position as someone who could afford a legal defense who uh white, Yeah exactly. That's another one too, is that people of color uh bear a inordinate burden, disproportionate burden when it comes to facial recognition technologies. We'll see well, I mean you might as well go ahead and talk about that. It's um.

I think from the beginning, even with social media, there were certain facial recognition, early facial recognition technologies that admitted like, we're not as good as uh seeing or recognizing faces

with darker skin. It's just not that good. Yeah, I think something like darker skinned men and women were recognized, twelve and thirty were misidentified, compared to one percent and seven percent of light skinned men and him in And they say it's because of the data sets that these machines have been trained on, which is not it's not purposefully, but it makes sense if you live in like a generally like like the white people are in power, and

it's like whiteness is the most celebrated part of the society or whatever. That's why you're going to have more pictures of And when you feed just a bunch of pictures from your society into a machine and say, learn what faces are, it's going to go, oh, white men, I got you. Well, they just are more white people. Numbers wise, probably has something to do with it, right, Um, Yes, that's a really that is an excellent point as well,

for sure. But the fact of the matter is the data sets that the machines are learning on are largely white and largely male, and so they're just not as good at recognizing the differences in faces among um people who aren't white males. Uh, let's read these quotes. There's a couple of good quotes here. The first one is from Woodrow Heart sog. I was going to read it as wurner. I don't know if I can. I should get Nolan here. He does a good burner. The most

uniquely dangerous surveillance mechanism ever intended. It's an irresistible tool for oppression that's perfectly suited for governments to display unprecedented authoritary tane I'm sorry, authoritarian control and an all out privacy eviscerating machine. I just realized it's heart sog, so it's it's spelled differently. It's a j r T hert sog is just h g r z o G. I'm glad that we didn't figure that out beforehand, though. You

want to take the other one though. I also I have to say I detected a hint of Michael Caine in there too. There might have been a little bit. It's hard to get Michael Caine out of my system.

What's the other one? Oh? From Microsoft president Brad Smith? Yeah. So. Brad Smith says that when combined with ubiquitous cameras and massive computing power and storage in the cloud, a government could use facial recognition technology to enable continuous surveillance of specific individuals like they're supposedly doing in China as an aside, it could follow anyone anywhere, or for that matter, everyone everywhere, at any time, or even all the time. And he

wasn't this wasn't a sales pitch. He was speaking out against this to Congress, saying like, guys, we gotta we have to do something about this, because this is the path we're heading down. And that's why uh Seth abrama Witz changed his name to Brad Smith. It sounds like a total like like I just want to blend in. UM. So you've got scanning against mug shots, scanning against driver's licenses, and then um, there's a new one that just came out.

The New York Times just released this expose on January eighteenth, just a few days ago, UM on a company called clear View AI. And apparently even among um Silicon Valley there has been this longstanding kind of unspoken thing where let's steer clear of this facial recognition technology because it's such a tool of oppression potentially. And clear View AI said, hey, we're not from Silicon Valley. Well, we're just going to

do our own thing. So now there's this, there's this tool that's available to law enforcement agencies that they're using. Remember that one that one guy who had a quote saying that, um, it went from thirty days to three minutes. They were almost certainly using clear View AI. And the reason clear View AI has such an advantage because they've gone to this place where everyone else said was off limits,

which is scraping social media. So rather than the forty one million driver's license and mug shot pictures that is available in the FBI's database, clear View AI is this app that you can subscribe to for a year for like two thousand and ten thousand dollars, and they have three billion pictures, including links to the social accounts of the people whose pictures come up, so that you can not only see who it is, you can find out where they're at right then. And it's a hugely invasive thing.

And there's no legislation on this whatsoever. And it's only just recently come out that this this company even exists, or that this app exists, and that law enforcement is using this stuff because again there's basically no laws saying you can do this, you can't do that. Um And again Woodrow Herzog has basically said, there's no way we're going to realize the benefits of this without the incredibly disproportionate drawbacks, and um, he just calls for an all

outband of the technology. He's basically saying it's not worth it. All right, let's take another break. Oh my gosh, we haven't taken our second break yet. Uh, and we'll be right back to talk about the rest of this stuff

right after this. I think we should talk a little bit, like we've talked about the false positives, UM, and I think within Amazon, their contention is that what you're talking about with like the studies out of m I T that said, UM, that there are too many false positive is he's they're saying, wait a minute, you're talking about facial analysis, not facial recognition, and those are two different things. I did not understand this at all. I went and looked it up and I just fully get it. Either

there's it sounds like some tap dancing to me. I looked at and there's like not a distinction between those two aside from this quote. Yeah, it's basically the same thing. And also it doesn't even make sense as a defense. So basically what they're saying is that um, that they were being called out by M I. T. S. Media Lab. They did a two thousand eighteen study. That's the one that found that there's like a twelve and thirty five

misidentification among darker skinned men and women women. I think yeah, um, and Amazon said no, no, no, uh, you guys are using facial analysis, not facial recognition. It's like, no, that's that's not the case at all. They're doing facial recognition. All right. I'm glad it wouldn't just me, because you see I wrote I don't get it next to this. It was a it was a bad, a bad jam,

I guess. But I think their point was, well, you're trying to tell the gender of somebody, and if you're doing binary gender stuff, like you're trying to say this as male or female, you can't really use facial recognition for that, especially among darker skinned people. And they said that you shouldn't use that, especially in cases of people's

civil liberties or whatever. But it still remains the case that if you are a darker skinned person and you're being looked at by a police department that has their threshold for a confidence level set low, yeah, there's a chance that a false positive is going to be put out there, right, And that's that can be trouble for you if you don't have the money to mount a defense. And even if you do have the money, you shouldn't have to mount a defense to spend money on that

to be equitted of a crime. Just because the computer is not so good at a distinguishing Um black people think it is among white people. Yeah, And what you know when it comes to where this is going to end up legally, Uh, you might want to look at the Fourth Amendment. Um. It gets really dicey on how you interpret the constitution when you talk about illegal search and seizure? Is this a search or a seizure? Probably not, um, because it depends on what we're talking about with the

Supreme Court. Um. You've probably been stopped h in a at a d y checkpoint, and that's stopping everybody. That's sort of the same thing. It's like, if you're in a car, we're gonna stop you and check you out because the couplet the public, the public his said, you know, that's okay, it's reasonable, it's not super invasive. And if you're stopping drunk drivers, it's just putting someone out for

a few minutes. Yeah, the court said, if it's minimally invasive and the public good or the potential for public good, which is in this case getting drunk drivers off the road is high enough, then it's it's okay to basically search everybody without probable cause. Yeah. Same with T s A checkpoints when it comes to official rulings. Obviously we

don't have one in facial recognition yet. But if you look at Carpenter v. United States, the court ruled five four that police violated uh Fourth Amendment rights of a man when they asked for his cell phone location data without a warrant from T Mobile. Um. So hopefully this nuance will prevail, and it just won't. It looks like it probably won't be some blanket ruling that just says, yep, you can use it for whatever you want, right if it even gets to that point, and if the court

hears it, which probably would. So. The other thing that um that is has become worrisome for people, though, is there. It's becoming um our society is becoming increasingly surveiled. Right like ring the ring doorbell. They market to law enforcement basically saying like you can, you can, These people like will pay to have video cameras put on their house and you can go get these videos on neighborhood all the time, people like my car got broken into, who

can help me out with their cameras. So it's being marketed to law enforcement. Your TV has a camera and your smart speaker has a microphone and it So the more that we are um surveiled and the more ubiquitous facial recognition technology gets, the easier will be to not just scan a picture of somebody stealing beard of a CVS against the mugshot database, or but to say this person right here that you're you're looking at, that the cameras following, that's this, that's that's um, that's uh, that's

Chuck Bryant right there. And everywhere you walk there's getting a little icon next to your head Chuck Bryant. You know, if you click it, it'll show you your Facebook page or a map to your house or whatever. They want to know, your police record, it doesn't matter. And that this is what we're increasingly getting closer to. And some people say this is what they're already doing in China. Yeah, and London has has They were one of the first on the CCTV train. Yes, but they use humans, which

is fair right for recognizing phases. Yeah, they have people like actually looking at the at the individual monitors looking for crime. This is this is the idea of this is it's just tracking people who are just doing nothing wrong. Yeah, but there are plenty of people on the other side we should point out that are like, you know what, if you're catching bad guys, that's great. If you're a good guy, you've got nothing to hide, so you shouldn't sweat it. Yeah. I can never remember the name of

the article. I'll try to find it, but there's a man I wish I could remember off the top of my head. But there's there's this amazing article from a few years back, um that that basically says like that's that's a terrible argument that that even if you have nothing to hide, you still, um are a human being.

And if somebody wanted to put together and but if somebody wanted to put together like a dossier on your embarrassing things that you've done or said or thought or whatever, um and and put it all together and condensed it, you can make anybody look bad. No one should want to live in a situation where like that could conceivably happen in the police state. Yeah, police state, good stuff. I guess we'll see how it pans out. I'm not

saying police state stuff. Police date. It's good stuff. Yeah, we'll see what happens right in Woodrow, Hurtzog and let us know what to do. If you want to know more about facial recognition technology, you can go onto the internet and start reading stuff about it. Definitely read the New York Times expose about um clear view AI came out January. Yes, okay, since I said that it's time for a listener mitt all right, now it's not you know what it's time for. Oh yeah, I know it's

time for you ready, Yeah, you say it. It's time for administrative details. All right. This is part two. This is where we thank people on the show that have sent us kindness. Is via snail mail. Siggy s I g D. I sent sent me some hand that it sucks not you for some reason. I don't know why. I got some socks too. Oh really yeah, I didn't know who they were from, so they may be from Siggy.

I think probably what it was is you left him with my desk and I thank you for it, CHUCKR do another one while I'm pulling up my list acting. Julie Shoop made us t shirts. Shoop, this is good stuff. Faux band name tour shirts. Uh, super fun. Thanks a lot of Julie. Very cool. Uh, you're still working, so I'm gonna keep going. Thalia Dawes is our pal from Australia,

said my daughter a couple of books. She's a very lovely lady who has a very adorable and whip smart daughter about the same age, who listens to our show. And um, I was just like, man, I wish she lived here. We could go into play date. They both seemed like lovely humans. There's such thing as plains. Yeah, good Australia for a play date. Um. So at our Portland's main show, Chuck, we had like a lot of um, We've got a lot of neat gifts. Jim Diefenbacher made

us amazing cross hatch portraits of them. Yeah, those were great of us, like of a photo we took I think on like our West Coast tour from two. It brought back some memories saw that. It's just really great stuff. And you can see Jim's work at Jim Diefenbaker dot com, j I M D I E F F E N B A C H E R dot com and they were framed in everything. Yeah, very sweet stuff, Jim. We got some home tapped maple syrup from Andy Huntsberger from

Elgin I A, Okay, what's Iowa? Is that Iowa? Yeah? Yeah, okay, Yeah, I was about to say the wrong state. What are you gonna say? You know, I think I went to say Illinoia. If you ever see Gary Gulman's bid on Nate abbreviating the states, dude, just look it up. One of the great comedy bits I've ever seen. It's hysterical. Let's see. Uh oh. Another at the Portlands Show, we got a letter from tog Braun from Downea's day Boat

from Lloyd Braun. Tog Braun Um and Downey's day Boats mission is to bring sustainable, delicious scalops from Maine to the world, and she said that scalops have varietals like oysters and that main has the best. So check out down East day Boat dot com. Coke Braun feel free to send us some scallops as long as they've been appropriately refrigerated. The entire time. I got another children's book, Are You a Good Egg? And that was from Peter Deutschel,

along with some stuff you should know coasters. Yeah yeah, thanks again Peter. I think we thanked him last episode for the coasters too. I didn't know about the children's book. Then Sarah Law who is an s y s k Army member. Um, she came to the Toronto show and she brought us a bunch of um Canadian goodies, everything from Japanese cheesecakes and tarts from Uncle Tetsus. So good. Um and uh. I think some other stuff too, like coffee crisps, which are my favorite. Um. Yeah, so thanks

a lot. Sarah's always that is everything from Japan awesome. They just it's really good. They don't necessarily invent much. They just take other people's inventions and perfect them. And it seems like they take a lot of pride in like doing things right. I think that he could say that, probably because we got from Matt an assortment of food things from Japan and that came in today, including our beloved QP Man's. I love that stuff. It's been too long, man,

God bless you. Let's see Leah Harrison gave us some amazing goodies to including coffee Crisp and Canadian Smarties, which are way better than American Smarties because they involved chocolate and super smarties. A student named Maria Styling wrote us a letter for an honors English project because she had to write someone who inspired her, and she asked this. I told her we to answer, how do we choose a topic? Maria? We choose a topic. It's not it's

pretty low fih. We just send each other one each week on whatever happens to grab our fancy. We're always looking around our world, uh and thinking I wonder about that. That's as that's as easy as it gets. And we'll just send each other an email and times out of a hundred will say great, let's do it. Yeah more ka,

let's see um oh. Michael C. Learner, who's an attorney at Law and Reno, sent us a letter about getting the word out about the National Consumer Law Center, for which Learner does a lot of pro bono work for people who are poor and getting screwed over because of debt. As he put it so, he pointed to the National Consumer Law Center and the Practicing law Institutes Consumer Financial Services.

Answer book. So if you are in debt and you're getting pushed around, go check those things out, says Michael C. Learner. Good stuff Van Ostro, and we gotta thank him again. Our buddy from Washington sent us a book by his friend Andy Robbins called Field Guide to the North American Jackalothes. Pretty awesome, that's pretty fun. Paul Speth from Mars Community Brewing Company in Chicago gives a bunch of beer at the Chicago show. Thank you for that. Um, I got

one more. Okay, I'll go and finish up and then you can round us out. Man, I have a whole page. Laugh all right. Robert Highland from Wammo, this was just came in today. Okay. He works for Wammo. He sent us each their seventieth anniversary super book. Oh wow, thanks a lot. Like you guys talk a lot about wammo products. Does it bounce? I have not dropped it on the floor. YEA,

let's find out. Give it a try. I'm gonna do a couple more and then well maybe we'll split these up because there for both of us for another episode. It's up to you. He can blaze through them too. Now there's too many Um, so let's see the Crown Royal people again for hooking us up. Very sweet. They've hooked us up many, many times. Um. And they gave us a nice congratulations because we've got the Best Curiosity Award from the Heart Podcast Awards last year. Oh yeah,

that's how old this one is. Mick Sullivan gave us a copy of his book The Meat Shower, which is amazingly illustrated. You can check it out on the Past and the Curious dot com. Yeah that just sounds really greats it really does. Let's see, um and over round everything out with Danielle Dixon, who is a real life marine biologists chuck at the University of Delaware, and she sent us a couple of copies of her kids books

See Stories, children's books based on real science. You can check it out at s E A S T O r Y books dot com. Alright, you're gonna save the rest. I'm gonna save the restival splow up. All right. Thanks everybody who sent us stuff, and and thank you also just for saying hi to anyone who does. You can say hi to us by sending us an email. Wrap it up, spending on the bottom, send it off to Stuff podcast at i heart radio dot com. Stuff you Should Know is a production of iHeart Radio's How Stuff Works.

For more podcasts for my heart Radio, visit the iHeart radio app, Apple Podcasts, or wherever you listen to your favorite shows.