WEBVTT - How Facial Recognition Technology Works 0:00:00.320 --> 0:00:02.880 Brought to you by the reinvented two thousand twelve camera. 0:00:03.240 --> 0:00:08.920 It's ready. Are you get in touch with technology with 0:00:09.080 --> 0:00:17.640 tech Stuff from how stuff works dot com. Hello again, everyone, 0:00:17.680 --> 0:00:20.040 and welcome to tech stuff. My name is Chris Poette 0:00:20.040 --> 0:00:21.920 and I am a tech editor here at how stuff 0:00:21.920 --> 0:00:24.600 works dot com. Sitting across from me, as he always does, 0:00:24.680 --> 0:00:27.880 is senior writer Jonathan Strickland. If your life had a face, 0:00:28.040 --> 0:00:32.239 I would punch it, okay, which brings us to a 0:00:32.240 --> 0:00:40.479 little listener mail. This listener mail comes from Dave, and 0:00:40.600 --> 0:00:43.400 Dave says, Hi, guys, love the show. I've been wondering 0:00:43.440 --> 0:00:46.519 how facial recognition technology works. How does it know what 0:00:46.600 --> 0:00:49.000 a face is and whose face it is? What are 0:00:49.040 --> 0:00:51.960 some of its uses fun and practical? Are there dangers 0:00:52.000 --> 0:00:55.480 and controversies around this technology? What's in the future for 0:00:55.520 --> 0:00:58.360 this stuff? Thanks for all your fascinating discussions. Have a 0:00:58.400 --> 0:01:01.520 good one, and I should point out also this email 0:01:01.840 --> 0:01:04.600 other people have asked us about facial recognition technology, so 0:01:04.800 --> 0:01:07.200 Dave's was the first one that I came across. It 0:01:07.280 --> 0:01:10.400 dated from February of two thousand and ten. It is 0:01:10.760 --> 0:01:14.000 currently as we're recording this October of two thousand and 0:01:14.040 --> 0:01:19.920 ten Dave, I'm sorry, we have lots of other topics 0:01:19.959 --> 0:01:22.680 just like that, so we got plenty of stuff. Keep 0:01:22.760 --> 0:01:24.840 letting us know what you want to hear. Yeah. So, 0:01:25.000 --> 0:01:27.679 but a lot of people have asked about face recognition technology, 0:01:27.680 --> 0:01:29.640 and we thought, I thought it would start with kind 0:01:29.640 --> 0:01:34.160 of just a brief discussion about face detection technology because that, 0:01:34.360 --> 0:01:36.600 you know, you really you build upon that. And I 0:01:36.640 --> 0:01:38.800 think a lot of us have used digital cameras at 0:01:38.800 --> 0:01:43.760 this point to have some face detection technology built into them. Yes, 0:01:44.040 --> 0:01:48.200 it's it's not uncommon, No, not at all. Now here's 0:01:48.240 --> 0:01:51.800 the thing about recognizing and detecting faces. People are really 0:01:51.840 --> 0:01:56.040 good at that. Yeah, well most people are, right, most 0:01:56.040 --> 0:01:58.280 people are. There are exceptions, of course, but the the 0:01:58.360 --> 0:02:01.880 average person who doesn't have any uh problems with his 0:02:02.040 --> 0:02:05.800 or her site or have any problems within their brain 0:02:06.040 --> 0:02:10.160 that makes it difficult to detect and recognize faces they check, 0:02:10.240 --> 0:02:12.640 they usually see it right away, right, you know, you 0:02:12.680 --> 0:02:14.480 just look at a person, you see their face, you 0:02:14.520 --> 0:02:17.040 recognize that person if you've if you've seen this person before, 0:02:17.400 --> 0:02:21.440 more often than not, you'll recognize that person. Computers aren't 0:02:21.800 --> 0:02:24.960 very good at this. Uh, It's one of the things 0:02:25.000 --> 0:02:28.400 that that really is a barrier in artificial intelligence is 0:02:28.400 --> 0:02:32.600 that humans are very good at detecting and recognizing patterns 0:02:32.800 --> 0:02:36.040 and and keeping that information stored so that they recognize 0:02:36.040 --> 0:02:39.239 it when they encounter it another time. Yes, in fact, 0:02:39.280 --> 0:02:42.760 we're so good at it, we sometimes make mistakes. This 0:02:42.840 --> 0:02:46.679 is this is often comes out and paradolia, which is 0:02:46.720 --> 0:02:50.000 where you you see patterns in stuff that there's no 0:02:50.120 --> 0:02:52.600 actual pattern there, Like you look up in the clouds 0:02:52.639 --> 0:02:55.200 and you see, hey, that just looks like my buddy 0:02:55.280 --> 0:02:59.040 Joe's face up there. Well, that's that's our brain creating 0:02:59.240 --> 0:03:02.079 a pattern where there's not really a pattern there, right. 0:03:02.440 --> 0:03:06.720 But computers are not traditionally very good at this. It's 0:03:06.720 --> 0:03:10.240 actually a big computing problem and problem in the sense 0:03:10.280 --> 0:03:13.760 of how do you teach a computer to recognize patterns? 0:03:14.320 --> 0:03:17.880 Face detection and face recognition kind of that that's right 0:03:17.960 --> 0:03:21.360 up there with that problem is how do you teach 0:03:21.400 --> 0:03:25.680 a computer what is a face? Well, that's a little 0:03:25.720 --> 0:03:29.079 bit tricky. Of course. This is done with software and 0:03:29.240 --> 0:03:32.680 UM it relies on on algorithms, which are you know, 0:03:32.840 --> 0:03:37.000 a sense of instructions basically for for computers or or 0:03:37.040 --> 0:03:40.440 anything running a kind of software like this. UM and 0:03:41.240 --> 0:03:44.880 what the what the researchers and engineers have had to 0:03:45.320 --> 0:03:48.720 do is install a layer of software that enables these 0:03:48.720 --> 0:03:53.160 devices to uh you know, they had to basically teach 0:03:53.200 --> 0:03:56.120 it what I mean, not literally they we're not talking 0:03:56.160 --> 0:03:59.440 artificial intelligence, but they had to basically teach it what 0:03:59.560 --> 0:04:05.200 is a face? Right, So often software software may or 0:04:05.240 --> 0:04:08.040 may not be the right term, depending upon which device 0:04:08.080 --> 0:04:12.160 you're using. Firmware maybe because often it is hard coded 0:04:12.160 --> 0:04:16.200 directly onto a chip. But yeah, it's the same same principle, right, 0:04:16.240 --> 0:04:19.800 It's it's not it's not hardwired onto the chip itself. 0:04:19.839 --> 0:04:23.080 It's it's a program that exists. There. I apologize for 0:04:23.160 --> 0:04:25.280 my mistake. No, no, no, there's plenty of stuff out 0:04:25.320 --> 0:04:28.560 there where it is a layer of software. It's not firmware. 0:04:28.600 --> 0:04:31.240 It all depends on and we we've talked about those 0:04:31.760 --> 0:04:36.280 definitions being fuzzy anyway, right, so I'm just trying to Yeah, 0:04:36.360 --> 0:04:39.200 so anyway, fuzzy firmware. That'd be a great name for 0:04:39.240 --> 0:04:43.719 a band. That's my Devo cover band name. So so yeah, 0:04:43.800 --> 0:04:48.359 the generally what the this firmware or software, what this 0:04:48.440 --> 0:04:52.680 program is looking for are the basic identifiers of the 0:04:52.720 --> 0:04:56.480 average human face, which would be, uh, your eyes knows 0:04:56.520 --> 0:04:59.680 your your ears, your chin kind of the outline, and 0:04:59.720 --> 0:05:04.960 when it recognizes that basic pattern, the the software identifies 0:05:05.000 --> 0:05:07.880 that as a face. So if you hold a digital 0:05:07.920 --> 0:05:11.640 camera with face identification software on it or or that 0:05:11.760 --> 0:05:16.400 feature is enabled, if it sees a pattern that looks 0:05:16.480 --> 0:05:20.479 like ears, eyes, nose, and chin, it's going to immediately 0:05:20.520 --> 0:05:23.599 assume that that's a face. Which sometimes it can be funny, 0:05:23.720 --> 0:05:26.720 like you can sometimes get face recognition software to recognize 0:05:26.720 --> 0:05:31.719 a face on a on a like a picture, like 0:05:31.760 --> 0:05:33.680 it's not even a person, You're like a mural on 0:05:33.720 --> 0:05:37.280 a building. In fact, I remember Google street View. They 0:05:37.400 --> 0:05:40.679 use an algorithm that's essentially this, you know, it looks 0:05:40.760 --> 0:05:44.440 for those features in order to blur out faces. That's 0:05:44.440 --> 0:05:48.120 part of the privacy uh stance that Google takes is 0:05:48.160 --> 0:05:51.680 you know, people were objecting to having their their pictures 0:05:51.680 --> 0:05:53.840 on Google street View. So what Google did was they 0:05:53.839 --> 0:05:57.320 create this algorithm that looks for the human face and 0:05:57.360 --> 0:06:00.440 then it applies a blurring layer over it so that 0:06:00.520 --> 0:06:03.760 you can't tell who that is. Right. Well, I've seen 0:06:03.800 --> 0:06:07.160 that work on things that were not human beings, mostly 0:06:07.279 --> 0:06:10.479 on things like like billboards or there was one that 0:06:10.600 --> 0:06:11.880 was a mural that was on the side of a 0:06:11.920 --> 0:06:14.720 building and the the face on the mural had been 0:06:14.720 --> 0:06:17.479 blurred out automatically by Google street View. Well it is 0:06:17.480 --> 0:06:22.120 a face. Yeah, it's just an actual three dimensional face. 0:06:22.160 --> 0:06:24.400 There's a good chance that mural did not object to 0:06:24.520 --> 0:06:30.400 having its privacy, uh review, you know, violated. But anyway, 0:06:30.720 --> 0:06:34.800 so it's looking for that basic set and that's just 0:06:34.920 --> 0:06:38.320 your your your very basic face detection. It it it 0:06:38.320 --> 0:06:41.839 it says, it looks for this this pattern of images 0:06:42.200 --> 0:06:45.680 and says this is what a human face is. Now 0:06:46.000 --> 0:06:49.320 when you go to face recognition, where it's going beyond 0:06:49.360 --> 0:06:52.520 detecting a face, it's actually recognizing a face and and 0:06:52.680 --> 0:06:56.520 setting that to an identity, we get a little more complex, 0:06:56.520 --> 0:07:00.040 actually a lot more complex, true enough. Um, yeah, it 0:07:00.440 --> 0:07:04.080 does rely on those facial landmarks. Yes, you know, your 0:07:04.120 --> 0:07:06.880 eyes and nose chin, the depth of depth of your 0:07:06.880 --> 0:07:11.920 eye sockets, the length of your nose, wid your nose, yeah, nostrils, 0:07:12.000 --> 0:07:15.520 everything like that can be part of this. And and 0:07:15.600 --> 0:07:19.280 for the for the camera in this case, it's recognizing 0:07:19.280 --> 0:07:23.360 a face print sort of like your thumb print or fingerprint. Um. 0:07:23.440 --> 0:07:26.680 And so once it it can keep a record of 0:07:26.720 --> 0:07:30.720 what a particular face looks like, um, then it can 0:07:31.040 --> 0:07:33.800 can you know that? That's I think you would probably 0:07:33.840 --> 0:07:36.480 call that the first step and being able to identify 0:07:36.520 --> 0:07:39.320 a particular face. So you take a person's face and 0:07:39.480 --> 0:07:41.920 you look at these different measurements. It might be the 0:07:42.600 --> 0:07:46.200 distance between the eyes. Uh, it can be things like 0:07:46.280 --> 0:07:50.440 the width of the eyes themselves, um, the where the 0:07:50.560 --> 0:07:55.760 ears are in relation to the the head as a whole, um, 0:07:55.880 --> 0:07:59.160 the jawline, all of these kind of features that you 0:07:59.160 --> 0:08:02.600 really want to focus on. Features that are not easily changeable, right. 0:08:02.680 --> 0:08:06.640 You know, like hairstyle would be a bad measurement because 0:08:06.680 --> 0:08:09.280 you could easily change that, right, right. You want things 0:08:09.320 --> 0:08:12.320 that won't change over time, right, So you know it 0:08:12.400 --> 0:08:14.800 might be things like again the width of the forehead, 0:08:14.880 --> 0:08:18.120 that sort of stuff. These are all called nodal points, 0:08:18.800 --> 0:08:21.720 all right, And uh, we have a great article about 0:08:21.760 --> 0:08:25.640 facial recognition technology on how stuff Works dot com. And 0:08:25.920 --> 0:08:28.240 in that article you learned that that the human face 0:08:28.280 --> 0:08:32.840 has around eighty nodal points. Now, not all facial recognition 0:08:32.920 --> 0:08:35.280 technology is going to rely on all e d of 0:08:35.320 --> 0:08:38.600 those in order to create a face print, but they'll 0:08:39.200 --> 0:08:43.280 rely on some combination of those nodal points and through 0:08:43.320 --> 0:08:46.800 the measurements will come up with a numeric value, which 0:08:47.320 --> 0:08:49.600 is the that's the equivalent of the face print. It's 0:08:49.600 --> 0:08:52.800 a numeric value that is unique to that person. More 0:08:52.880 --> 0:08:59.000 or less, identical twins actually can have the same face print. Wow, yeah, 0:08:59.120 --> 0:09:03.560 it's it's so some facial recognition technology cannot discern between 0:09:04.080 --> 0:09:07.360 identical twins. There are other kinds that rely on even 0:09:07.400 --> 0:09:10.840 more specific data that can discern between the two. But 0:09:11.040 --> 0:09:16.079 a basic facial recognition camera, Um, if you had identical 0:09:16.120 --> 0:09:21.440 twins and they really were identical, yeah, like they didn't 0:09:21.480 --> 0:09:23.720 have one of them didn't have some sort of facial 0:09:23.760 --> 0:09:28.400 feature that was remarkably different. Uh, this software might identify 0:09:28.520 --> 0:09:32.720 both as being the same person without without additional layers. 0:09:32.720 --> 0:09:34.760 And we'll get into that in a little bit. So 0:09:35.200 --> 0:09:37.559 you've got this face print, you create a database of 0:09:37.640 --> 0:09:40.680 face prints. Then when you take a picture of someone 0:09:41.280 --> 0:09:46.000 the again, the camera will measure the noal points on 0:09:46.040 --> 0:09:49.680 this person's face, compare it against the database and see 0:09:49.679 --> 0:09:54.040 if there's a match. And it's probably I'm guessing we 0:09:54.080 --> 0:09:56.079 don't really go into a whole detail in our article, 0:09:56.080 --> 0:09:59.200 but I'm guessing it's kind of like fingerprints in a way, 0:09:59.240 --> 0:10:04.080 you look for a percentage of probability that this this 0:10:04.240 --> 0:10:07.640 particular face matches one that's in the database, because you 0:10:07.720 --> 0:10:11.280 got to remember, not all of these images are going 0:10:11.320 --> 0:10:15.480 to be exactly the same. Early facial recognition technology was 0:10:15.600 --> 0:10:18.440 very limited. You had to have someone looking directly into 0:10:18.480 --> 0:10:20.720 the camera, and then you would have to have that 0:10:20.800 --> 0:10:24.280 same person look directly into the camera again later on 0:10:24.760 --> 0:10:27.120 and compare that to the database in order to find 0:10:27.120 --> 0:10:29.400 a match. If the person was looking a little to 0:10:29.440 --> 0:10:31.640 the left or to the right, the technology wasn't good 0:10:31.800 --> 0:10:35.160 enough to to compensate for that and to make a 0:10:35.200 --> 0:10:38.120 model of that person's face to really get the right 0:10:38.120 --> 0:10:42.120 measurements right. So, and you've got to think of all 0:10:42.120 --> 0:10:44.520 the other factors that play into this. It's the lighting. 0:10:45.520 --> 0:10:49.760 If the lighting is bad, those early facial recognition technologies 0:10:49.880 --> 0:10:52.280 weren't very effective. Or if the person was at a 0:10:52.320 --> 0:10:56.000 different distance. The camera has to know how far away 0:10:56.000 --> 0:10:59.800 you are in order to make valid measurements for things 0:10:59.800 --> 0:11:02.240 like how far apart your eyes are. If the camera 0:11:02.320 --> 0:11:05.079 thinks you're fifteen ft away but you're really twelve feet away, 0:11:05.360 --> 0:11:07.360 those measurements are not going to be accurate, and it's 0:11:07.400 --> 0:11:10.080 not going to find the right match in the database. Right, 0:11:10.920 --> 0:11:12.880 So this is this was definitely one of those things 0:11:12.880 --> 0:11:14.600 that was a big learning curve you had to be 0:11:14.640 --> 0:11:17.520 able to build. You had to develop the digital technology 0:11:17.520 --> 0:11:22.840 to detect distance and then accurately measure as many Noble 0:11:22.920 --> 0:11:25.360 points as possible and as little time as possible. And 0:11:25.360 --> 0:11:28.120 we're talking hundreds of a second here. Uh that that 0:11:28.400 --> 0:11:32.400 a chip is scanning a person's face and identifying those 0:11:32.440 --> 0:11:34.160 nobal points. I mean, it takes no time at all 0:11:34.200 --> 0:11:38.200 for this to happen, but it took engineers time to 0:11:38.280 --> 0:11:43.000 develop that technology. And yeah, there there are some really 0:11:43.040 --> 0:11:48.360 fascinating technologies built into facial recognition, including you know, technologies 0:11:48.400 --> 0:11:53.280 such as a surface texture analysis. It's basically creating a 0:11:52.240 --> 0:11:55.720 a another not a face print, but a skin print. 0:11:56.040 --> 0:11:58.599 Where they are doing, um, the system is doing a 0:11:59.000 --> 0:12:03.960 mathematical analogy of sections of your skin if you come 0:12:04.000 --> 0:12:08.439 before the camera. So um, if you had to say 0:12:08.480 --> 0:12:11.280 a birthmark or a mole or something, that would probably 0:12:11.320 --> 0:12:14.400 help it track down who you are because it's going 0:12:14.440 --> 0:12:16.440 to say, well, we know that this is you know 0:12:16.520 --> 0:12:20.760 in sector number seventeen. Um, you know there is this 0:12:21.280 --> 0:12:23.720 different coloration than there there would be in the rest 0:12:23.760 --> 0:12:27.160 of of the face, so um that that you know, 0:12:27.200 --> 0:12:31.360 they're very sophisticated in breaking them down into um, you know, 0:12:31.400 --> 0:12:36.760 all kinds of mathematical UH constructs to enable this to work. Yeah, 0:12:36.760 --> 0:12:39.240 And the service texture analysis is that layer I was 0:12:39.280 --> 0:12:42.920 talking about earlier about how to identify between identical twins. 0:12:43.920 --> 0:12:48.280 Service texture analysis is actually the kind of of technology 0:12:48.320 --> 0:12:51.280 you want in order to do that, because it's looking 0:12:51.360 --> 0:12:53.920 much more closely at the texture of your skin. As 0:12:53.960 --> 0:12:58.559 the name would imply, so uh, even identical twins aren't 0:12:58.559 --> 0:13:02.960 going to have identical lines on their faces, you know, laugh, lines, wrinkles, 0:13:03.000 --> 0:13:05.040 that kind of thing. There might be a freckle or 0:13:05.080 --> 0:13:08.360 a mole that's it's slightly different from one twin to 0:13:08.400 --> 0:13:11.040 the other. And this is the sort of technology that's 0:13:11.080 --> 0:13:13.200 going to pick up on that, as opposed to the 0:13:13.240 --> 0:13:16.680 basic facial recognition technology that might not it might be 0:13:16.720 --> 0:13:19.160 close enough between the two twins in order to identify 0:13:19.200 --> 0:13:21.880 that as the same person. But with this layer, it 0:13:21.920 --> 0:13:28.400 can make it more accurate. In fact, according to one company, UH, 0:13:28.440 --> 0:13:33.559 it can, which is called Identics. UH. Surface texture analysis 0:13:33.600 --> 0:13:36.760 can increase the validity or the the reliability of a 0:13:36.760 --> 0:13:41.360 scan by around which is pretty significant. I mean, you know, 0:13:41.480 --> 0:13:45.600 you know, depending upon how accurate your starting point is. 0:13:46.200 --> 0:13:49.120 The early early facial recognition technology, even when it was 0:13:49.160 --> 0:13:52.240 working well, was not working that well. It was like 0:13:52.280 --> 0:13:56.600 around a sixty success rate. And uh, you think about that, 0:13:56.640 --> 0:13:59.280 it means of the time they get it wrong, and 0:13:59.760 --> 0:14:01.880 that can be pretty serious when you consider that. A 0:14:01.880 --> 0:14:04.840 lot of the facial recognition technology that's out there is 0:14:04.920 --> 0:14:08.400 used in law enforcement practices. Yes, it's in order to 0:14:08.520 --> 0:14:13.319 identify people who you know, let's say that there's unknown 0:14:13.640 --> 0:14:17.439 suspect that is on the loose, may be used to 0:14:17.440 --> 0:14:19.760 try and identify someone like that, or it may even 0:14:19.760 --> 0:14:23.880 be used in a static system where if something happens 0:14:23.920 --> 0:14:26.200 within the view of the camera, the camera would be 0:14:26.280 --> 0:14:30.160 able to to identify that person more quickly. You wouldn't 0:14:30.200 --> 0:14:32.200 have to just you know, stare at this picture and say, 0:14:32.280 --> 0:14:34.080 I wonder if this is you know, is this a 0:14:34.160 --> 0:14:36.600 suspect or you would have compared against the database of 0:14:36.680 --> 0:14:39.560 like mug shots or whatever, and um, that can be 0:14:39.600 --> 0:14:42.880 really useful in UM in urban environments that have lots 0:14:42.920 --> 0:14:45.880 and lots of cameras that they use for law enforcement purposes, 0:14:46.160 --> 0:14:49.120 places like London where there are so many cameras on 0:14:49.160 --> 0:14:53.680 all the different street corners, but also in places like banks. Um, 0:14:53.760 --> 0:14:58.680 if somebody were unknown uh you know, unknown bank robber 0:14:58.800 --> 0:15:01.880 and they didn't where you know, some kind of facial 0:15:01.920 --> 0:15:05.480 obscuring gear. Now, I mean it would be easy enough 0:15:05.520 --> 0:15:07.800 to fool a bank camera, probably if you put on 0:15:07.800 --> 0:15:09.800 a fake mustache or something like that, because that's the 0:15:09.840 --> 0:15:13.720 kind of thing that's going to sort off the facial recognition. Yeah, 0:15:13.720 --> 0:15:17.440 there's some facial recognition technologies that are they're sophistical enough 0:15:17.480 --> 0:15:20.360 to ignore things like facial hair. It'll look at the 0:15:20.840 --> 0:15:24.000 shape and size of your face and the relationship of say, 0:15:24.120 --> 0:15:26.960 like again the distance between your eyes, and it ignores 0:15:27.000 --> 0:15:28.920 things like facial hair because again, facial hair is one 0:15:28.920 --> 0:15:32.160 of those things it's easy to to grow or remove, 0:15:32.360 --> 0:15:37.200 right for some of us anyway, ladies, hopefully not for you. 0:15:37.800 --> 0:15:42.840 But the the most facial recognition technology will ignore that 0:15:42.920 --> 0:15:45.240 kind of stuff. But yeah, if you're wearing something that's 0:15:45.240 --> 0:15:48.920 obscuring your face, that definitely will throw off facial has 0:15:48.920 --> 0:15:50.960 to and that's not gonna be able to recognize it 0:15:50.960 --> 0:15:53.760 because I can't see it. Um. So if you walk 0:15:53.760 --> 0:15:56.360 into the bank wearing a balla clava, the camera probably 0:15:56.400 --> 0:15:59.680 won't recognize you, but the security guard might have some issues, right, 0:16:00.080 --> 0:16:03.200 And this also leads us to the question about what 0:16:03.320 --> 0:16:06.120 sort of problems there can be. Well, clearly, privacy is 0:16:06.320 --> 0:16:09.480 a big concern, right. I mean, if you get to 0:16:09.520 --> 0:16:12.040 the point where you have technology that can recognize faces, 0:16:12.440 --> 0:16:14.760 then you've got your your to the point where you 0:16:14.800 --> 0:16:20.280 have to worry about you being viewed wherever you happen 0:16:20.320 --> 0:16:23.440 to be and identified as being there. And I mean 0:16:23.480 --> 0:16:25.760 that's a big problem. I mean even for people who, 0:16:26.080 --> 0:16:33.000 let's say that you are perfectly innocent, upright wonderful citizen 0:16:33.280 --> 0:16:36.520 and you've never done anything wrong, you still probably wouldn't 0:16:36.560 --> 0:16:40.240 necessarily want all these cameras everywhere identifying your your you 0:16:40.440 --> 0:16:43.560 wherever you happen to be. I mean, it's just it's 0:16:43.560 --> 0:16:46.480 it's an invasion of privacy. And because of that, there 0:16:46.480 --> 0:16:50.040 have been some experiments with this sort of system in 0:16:50.120 --> 0:16:54.480 place in various public areas that ended up getting um 0:16:55.240 --> 0:16:59.720 canceled somewhere along the project because either the public was 0:17:00.320 --> 0:17:02.720 in an uproar about it, saying, hey, this violates our 0:17:02.720 --> 0:17:06.440 privacy and I'm not comfortable with any agency tracking me 0:17:06.680 --> 0:17:12.199 like this, or the reliability was low enough so that 0:17:12.240 --> 0:17:15.200 there were concerns of Hey, I could be at home 0:17:16.800 --> 0:17:20.119 asleep and someone who looks enough like me for the 0:17:20.119 --> 0:17:23.800 facial recognition technology to think that that was me could 0:17:23.840 --> 0:17:25.920 commit a crime, and then I could be charged for it. 0:17:26.680 --> 0:17:30.600 And I mean, if the if the reliable reliability is low, 0:17:30.760 --> 0:17:34.240 like if it it wasn't that six, that's a legitimate concern. 0:17:35.960 --> 0:17:37.320 Why if you happen to be involved in one of 0:17:37.320 --> 0:17:42.560 those mistakes, like Jonathan, why did you break into Tiffany's 0:17:42.560 --> 0:17:44.600 and steal all these diamonds? And I'd be like, wait, 0:17:44.640 --> 0:17:48.159 what did I do? Now? When did that? I? I 0:17:48.200 --> 0:17:50.200 am certain I did not do that. It does not 0:17:50.320 --> 0:17:52.680 say I ever went to Tiffany's on four square, So 0:17:52.880 --> 0:17:57.520 clearly that wasn't me because I check in everywhere, right. 0:17:58.840 --> 0:18:02.040 But there can it can be a lot of positive 0:18:02.119 --> 0:18:04.679 uses of course for facial recognition. Oh yeah, no, there 0:18:04.840 --> 0:18:06.679 there are plenty of really good ones. I mean not 0:18:06.720 --> 0:18:09.760 that those aren't positive, but I mean more fun let's 0:18:09.760 --> 0:18:13.679 see uses. Of course. You know, with cameras being as 0:18:13.720 --> 0:18:16.720 sophisticated as they are, photos are geo tagged, and now 0:18:17.200 --> 0:18:19.840 you know the facial recognition software built into them, you 0:18:19.880 --> 0:18:24.560 can auto tag different photos as basically as you are, 0:18:24.600 --> 0:18:28.159 you know, importing them into your files. Yeah, that to 0:18:28.200 --> 0:18:33.160 me is absolutely amazing stuff. I was amazed back when 0:18:33.200 --> 0:18:36.280 cameras first started being able to detect faces. That to 0:18:36.359 --> 0:18:38.280 me was that to me was really really cool. I 0:18:38.320 --> 0:18:40.800 was like, hey, awesome. And then you had the next 0:18:40.840 --> 0:18:44.320 step beyond detecting faces was detecting when someone was smiling, 0:18:44.840 --> 0:18:49.040 because then they were measuring the person's mouth right, and 0:18:49.119 --> 0:18:52.320 so if a person was smiling or making some sort 0:18:52.359 --> 0:18:56.320 of facial expression that was akin to smiling, because sometimes 0:18:56.320 --> 0:18:59.000 it was like more of a grimace. Uh, your camera 0:18:59.119 --> 0:19:02.520 would merely take the picture. There's some cameras that had 0:19:02.520 --> 0:19:04.320 it where it would it would be ready to take 0:19:04.320 --> 0:19:06.200 the photo and as soon as the person smiled, that's 0:19:06.200 --> 0:19:08.399 when it would take the shot. So that way you 0:19:08.440 --> 0:19:11.600 would get the smile right. And then the next step 0:19:11.600 --> 0:19:14.160 beyond that is what you were talking about, where you 0:19:14.200 --> 0:19:16.639 would tag a photo. You take a picture of someone, 0:19:17.119 --> 0:19:19.919 you tag that that photo with the person's name, and 0:19:19.960 --> 0:19:23.280 then from that point forward, your camera compares the pictures 0:19:23.320 --> 0:19:26.280 you take against the people you've already tagged and says, 0:19:26.280 --> 0:19:28.640 oh wait, this is a person he's tagged already I'm 0:19:28.680 --> 0:19:30.880 just gonna go ahead and throw the tag on there. Yeah, 0:19:30.920 --> 0:19:35.000 that's a good point because just as the manufacturers roly 0:19:35.280 --> 0:19:40.360 on the software engineers to build the technology in so 0:19:40.400 --> 0:19:43.840 that these devices can recognize faces, you have to teach 0:19:43.880 --> 0:19:49.280 it who is who? Um? Who is whom? I should say, um, 0:19:49.400 --> 0:19:52.120 so you know, it won't know that when I take 0:19:52.160 --> 0:19:55.199 a photo of Jonathan that is Jonathan until I tell it, 0:19:55.440 --> 0:19:57.520 you know, this is who this is, and then therefore 0:19:57.680 --> 0:19:59.880 after that it will take photos right well, I reckon 0:20:00.040 --> 0:20:03.040 nize his face. It's Jonathan Strickland. And you know when 0:20:03.040 --> 0:20:06.800 I upload those embarrassing photos of him to Facebook, it will, 0:20:06.840 --> 0:20:10.840 you know, have all the impertinent information included properly, which 0:20:10.880 --> 0:20:12.600 is why every evening I have to sit down and 0:20:12.760 --> 0:20:15.600 untagged photos. Yeah, but you know, I was thinking of 0:20:15.640 --> 0:20:21.119 another uh technology facial recognition application I should say, Um, 0:20:21.160 --> 0:20:26.160 that is also pretty fun, which is uh Microsoft's connect. Ah. Yes, 0:20:26.359 --> 0:20:28.720 that's a good point. Yeah, connect is of course, that's 0:20:28.800 --> 0:20:36.400 the motion detecting right always, because that's what the that's 0:20:36.440 --> 0:20:38.679 that's what the project was called before it hit the public. 0:20:39.080 --> 0:20:40.919 How you associate a name with something and then it's 0:20:40.960 --> 0:20:43.600 sort of hard to unassociated with that. I still call 0:20:43.720 --> 0:20:50.840 segways ginger. That's it. That's going way back. So anyway, Yeah, 0:20:50.920 --> 0:20:56.040 Connect has facial recognition built into it, and it's the implementation. 0:20:56.119 --> 0:21:00.200 Like you were saying, pullet is a great idea weather 0:21:00.359 --> 0:21:02.400 weather or not it works well, I can't say because 0:21:02.400 --> 0:21:05.439 I haven't used Connect yet, but the I love the 0:21:05.440 --> 0:21:08.360 concept that when you step in front of your your 0:21:08.480 --> 0:21:13.320 entertainment center, the camera in the Connect peripheral looks at 0:21:13.359 --> 0:21:16.640 you and then analyzes your face. Does this this process 0:21:16.680 --> 0:21:19.280 that we're talking about where it measures, takes these measurements 0:21:19.359 --> 0:21:23.000 very quickly compares that to the information and database and 0:21:23.040 --> 0:21:26.000 then identifies you. So if you have previously set up 0:21:26.040 --> 0:21:30.360 and a Connect account, let's say with your Xbox, it knows, Hey, 0:21:30.359 --> 0:21:33.600 this is Jonathan. Jonathan likes to play at this particular 0:21:33.640 --> 0:21:37.440 skill level. Jonathan likes these particular games. Um, I'm going 0:21:37.520 --> 0:21:42.520 to present this this block of information and features and 0:21:42.640 --> 0:21:46.760 games to Jonathan because we already know what his preferences are. 0:21:47.600 --> 0:21:50.080 Then when someone else, let's say, Chris, steps in front 0:21:50.080 --> 0:21:52.760 of Connect, it will identify Chris and say, oh, well 0:21:52.800 --> 0:21:55.480 that's Chris. Chris like some of the games Jonathan likes, 0:21:55.480 --> 0:21:57.960 but he also likes these other games, and he prefers 0:21:58.040 --> 0:22:00.680 these kind of movies to the movies that Jonathan likes, 0:22:00.720 --> 0:22:03.960 like Chris prefers the three movies they've seen before to 0:22:04.119 --> 0:22:07.320 the vast database of movies that Jonathan has seen. So 0:22:07.400 --> 0:22:09.080 I'm just going to show them these three movies because 0:22:09.080 --> 0:22:10.640 there's no point in trying to get him to watch 0:22:10.680 --> 0:22:14.520 anything else. And then, um, but that's the thing is 0:22:14.560 --> 0:22:18.600 that it'll it'll tailor the experience to the person based 0:22:18.640 --> 0:22:21.680 upon that person's profile. Now again, like Plett was saying, 0:22:21.760 --> 0:22:23.760 you have to create a profile. You have to tell 0:22:23.800 --> 0:22:26.440 connect Hey, this is who I am. Whenever you see 0:22:26.480 --> 0:22:30.119 this face, this is the profile you should use. Now, 0:22:30.680 --> 0:22:32.520 I'm sorry you're going to say something, Well, I was 0:22:32.560 --> 0:22:34.679 just if you were going to extrapolate from that I 0:22:34.680 --> 0:22:37.760 thought of I just thought of another application similar to 0:22:37.960 --> 0:22:43.120 that that Apparently, as I look it up quickly, here, Um, 0:22:43.240 --> 0:22:45.680 other people have our way ahead of me, and doesn't 0:22:45.680 --> 0:22:49.359 surprise me in the least. Um, smart homes can do 0:22:49.400 --> 0:22:52.280 the same thing, because I know that. I remember reading 0:22:52.320 --> 0:22:55.960 a long time ago that when Bill Gates had built 0:22:56.040 --> 0:23:00.320 his you know, massive home with the Microsoft technology state 0:23:00.359 --> 0:23:02.479 of the art, and it would tell as soon as 0:23:02.480 --> 0:23:05.080 you walked into a room, uh, oh, well, you know 0:23:05.160 --> 0:23:07.560 this is Bill. He likes the temperature at you know, 0:23:07.600 --> 0:23:10.320 seventy two degrees, he likes this kind of music, and 0:23:10.320 --> 0:23:12.480 it would automatically like you could walk around the house 0:23:12.520 --> 0:23:14.840 and and I remember reading this. I don't know if 0:23:14.880 --> 0:23:16.560