1 00:00:00,320 --> 00:00:02,880 Speaker 1: Brought to you by the reinvented two thousand twelve camera. 2 00:00:03,240 --> 00:00:08,920 Speaker 1: It's ready. Are you get in touch with technology with 3 00:00:09,080 --> 00:00:17,640 Speaker 1: tech Stuff from how stuff works dot com. Hello again, everyone, 4 00:00:17,680 --> 00:00:20,040 Speaker 1: and welcome to tech stuff. My name is Chris Poette 5 00:00:20,040 --> 00:00:21,920 Speaker 1: and I am a tech editor here at how stuff 6 00:00:21,920 --> 00:00:24,600 Speaker 1: works dot com. Sitting across from me, as he always does, 7 00:00:24,680 --> 00:00:27,880 Speaker 1: is senior writer Jonathan Strickland. If your life had a face, 8 00:00:28,040 --> 00:00:32,239 Speaker 1: I would punch it, okay, which brings us to a 9 00:00:32,240 --> 00:00:40,479 Speaker 1: little listener mail. This listener mail comes from Dave, and 10 00:00:40,600 --> 00:00:43,400 Speaker 1: Dave says, Hi, guys, love the show. I've been wondering 11 00:00:43,440 --> 00:00:46,519 Speaker 1: how facial recognition technology works. How does it know what 12 00:00:46,600 --> 00:00:49,000 Speaker 1: a face is and whose face it is? What are 13 00:00:49,040 --> 00:00:51,960 Speaker 1: some of its uses fun and practical? Are there dangers 14 00:00:52,000 --> 00:00:55,480 Speaker 1: and controversies around this technology? What's in the future for 15 00:00:55,520 --> 00:00:58,360 Speaker 1: this stuff? Thanks for all your fascinating discussions. Have a 16 00:00:58,400 --> 00:01:01,520 Speaker 1: good one, and I should point out also this email 17 00:01:01,840 --> 00:01:04,600 Speaker 1: other people have asked us about facial recognition technology, so 18 00:01:04,800 --> 00:01:07,200 Speaker 1: Dave's was the first one that I came across. It 19 00:01:07,280 --> 00:01:10,400 Speaker 1: dated from February of two thousand and ten. It is 20 00:01:10,760 --> 00:01:14,000 Speaker 1: currently as we're recording this October of two thousand and 21 00:01:14,040 --> 00:01:19,920 Speaker 1: ten Dave, I'm sorry, we have lots of other topics 22 00:01:19,959 --> 00:01:22,680 Speaker 1: just like that, so we got plenty of stuff. Keep 23 00:01:22,760 --> 00:01:24,840 Speaker 1: letting us know what you want to hear. Yeah. So, 24 00:01:25,000 --> 00:01:27,679 Speaker 1: but a lot of people have asked about face recognition technology, 25 00:01:27,680 --> 00:01:29,640 Speaker 1: and we thought, I thought it would start with kind 26 00:01:29,640 --> 00:01:34,160 Speaker 1: of just a brief discussion about face detection technology because that, 27 00:01:34,360 --> 00:01:36,600 Speaker 1: you know, you really you build upon that. And I 28 00:01:36,640 --> 00:01:38,800 Speaker 1: think a lot of us have used digital cameras at 29 00:01:38,800 --> 00:01:43,760 Speaker 1: this point to have some face detection technology built into them. Yes, 30 00:01:44,040 --> 00:01:48,200 Speaker 1: it's it's not uncommon, No, not at all. Now here's 31 00:01:48,240 --> 00:01:51,800 Speaker 1: the thing about recognizing and detecting faces. People are really 32 00:01:51,840 --> 00:01:56,040 Speaker 1: good at that. Yeah, well most people are, right, most 33 00:01:56,040 --> 00:01:58,280 Speaker 1: people are. There are exceptions, of course, but the the 34 00:01:58,360 --> 00:02:01,880 Speaker 1: average person who doesn't have any uh problems with his 35 00:02:02,040 --> 00:02:05,800 Speaker 1: or her site or have any problems within their brain 36 00:02:06,040 --> 00:02:10,160 Speaker 1: that makes it difficult to detect and recognize faces they check, 37 00:02:10,240 --> 00:02:12,640 Speaker 1: they usually see it right away, right, you know, you 38 00:02:12,680 --> 00:02:14,480 Speaker 1: just look at a person, you see their face, you 39 00:02:14,520 --> 00:02:17,040 Speaker 1: recognize that person if you've if you've seen this person before, 40 00:02:17,400 --> 00:02:21,440 Speaker 1: more often than not, you'll recognize that person. Computers aren't 41 00:02:21,800 --> 00:02:24,960 Speaker 1: very good at this. Uh, It's one of the things 42 00:02:25,000 --> 00:02:28,400 Speaker 1: that that really is a barrier in artificial intelligence is 43 00:02:28,400 --> 00:02:32,600 Speaker 1: that humans are very good at detecting and recognizing patterns 44 00:02:32,800 --> 00:02:36,040 Speaker 1: and and keeping that information stored so that they recognize 45 00:02:36,040 --> 00:02:39,239 Speaker 1: it when they encounter it another time. Yes, in fact, 46 00:02:39,280 --> 00:02:42,760 Speaker 1: we're so good at it, we sometimes make mistakes. This 47 00:02:42,840 --> 00:02:46,679 Speaker 1: is this is often comes out and paradolia, which is 48 00:02:46,720 --> 00:02:50,000 Speaker 1: where you you see patterns in stuff that there's no 49 00:02:50,120 --> 00:02:52,600 Speaker 1: actual pattern there, Like you look up in the clouds 50 00:02:52,639 --> 00:02:55,200 Speaker 1: and you see, hey, that just looks like my buddy 51 00:02:55,280 --> 00:02:59,040 Speaker 1: Joe's face up there. Well, that's that's our brain creating 52 00:02:59,240 --> 00:03:02,079 Speaker 1: a pattern where there's not really a pattern there, right. 53 00:03:02,440 --> 00:03:06,720 Speaker 1: But computers are not traditionally very good at this. It's 54 00:03:06,720 --> 00:03:10,240 Speaker 1: actually a big computing problem and problem in the sense 55 00:03:10,280 --> 00:03:13,760 Speaker 1: of how do you teach a computer to recognize patterns? 56 00:03:14,320 --> 00:03:17,880 Speaker 1: Face detection and face recognition kind of that that's right 57 00:03:17,960 --> 00:03:21,360 Speaker 1: up there with that problem is how do you teach 58 00:03:21,400 --> 00:03:25,680 Speaker 1: a computer what is a face? Well, that's a little 59 00:03:25,720 --> 00:03:29,079 Speaker 1: bit tricky. Of course. This is done with software and 60 00:03:29,240 --> 00:03:32,680 Speaker 1: UM it relies on on algorithms, which are you know, 61 00:03:32,840 --> 00:03:37,000 Speaker 1: a sense of instructions basically for for computers or or 62 00:03:37,040 --> 00:03:40,440 Speaker 1: anything running a kind of software like this. UM and 63 00:03:41,240 --> 00:03:44,880 Speaker 1: what the what the researchers and engineers have had to 64 00:03:45,320 --> 00:03:48,720 Speaker 1: do is install a layer of software that enables these 65 00:03:48,720 --> 00:03:53,160 Speaker 1: devices to uh you know, they had to basically teach 66 00:03:53,200 --> 00:03:56,120 Speaker 1: it what I mean, not literally they we're not talking 67 00:03:56,160 --> 00:03:59,440 Speaker 1: artificial intelligence, but they had to basically teach it what 68 00:03:59,560 --> 00:04:05,200 Speaker 1: is a face? Right, So often software software may or 69 00:04:05,240 --> 00:04:08,040 Speaker 1: may not be the right term, depending upon which device 70 00:04:08,080 --> 00:04:12,160 Speaker 1: you're using. Firmware maybe because often it is hard coded 71 00:04:12,160 --> 00:04:16,200 Speaker 1: directly onto a chip. But yeah, it's the same same principle, right, 72 00:04:16,240 --> 00:04:19,800 Speaker 1: It's it's not it's not hardwired onto the chip itself. 73 00:04:19,839 --> 00:04:23,080 Speaker 1: It's it's a program that exists. There. I apologize for 74 00:04:23,160 --> 00:04:25,280 Speaker 1: my mistake. No, no, no, there's plenty of stuff out 75 00:04:25,320 --> 00:04:28,560 Speaker 1: there where it is a layer of software. It's not firmware. 76 00:04:28,600 --> 00:04:31,240 Speaker 1: It all depends on and we we've talked about those 77 00:04:31,760 --> 00:04:36,280 Speaker 1: definitions being fuzzy anyway, right, so I'm just trying to Yeah, 78 00:04:36,360 --> 00:04:39,200 Speaker 1: so anyway, fuzzy firmware. That'd be a great name for 79 00:04:39,240 --> 00:04:43,719 Speaker 1: a band. That's my Devo cover band name. So so yeah, 80 00:04:43,800 --> 00:04:48,359 Speaker 1: the generally what the this firmware or software, what this 81 00:04:48,440 --> 00:04:52,680 Speaker 1: program is looking for are the basic identifiers of the 82 00:04:52,720 --> 00:04:56,480 Speaker 1: average human face, which would be, uh, your eyes knows 83 00:04:56,520 --> 00:04:59,680 Speaker 1: your your ears, your chin kind of the outline, and 84 00:04:59,720 --> 00:05:04,960 Speaker 1: when it recognizes that basic pattern, the the software identifies 85 00:05:05,000 --> 00:05:07,880 Speaker 1: that as a face. So if you hold a digital 86 00:05:07,920 --> 00:05:11,640 Speaker 1: camera with face identification software on it or or that 87 00:05:11,760 --> 00:05:16,400 Speaker 1: feature is enabled, if it sees a pattern that looks 88 00:05:16,480 --> 00:05:20,479 Speaker 1: like ears, eyes, nose, and chin, it's going to immediately 89 00:05:20,520 --> 00:05:23,599 Speaker 1: assume that that's a face. Which sometimes it can be funny, 90 00:05:23,720 --> 00:05:26,720 Speaker 1: like you can sometimes get face recognition software to recognize 91 00:05:26,720 --> 00:05:31,719 Speaker 1: a face on a on a like a picture, like 92 00:05:31,760 --> 00:05:33,680 Speaker 1: it's not even a person, You're like a mural on 93 00:05:33,720 --> 00:05:37,280 Speaker 1: a building. In fact, I remember Google street View. They 94 00:05:37,400 --> 00:05:40,679 Speaker 1: use an algorithm that's essentially this, you know, it looks 95 00:05:40,760 --> 00:05:44,440 Speaker 1: for those features in order to blur out faces. That's 96 00:05:44,440 --> 00:05:48,120 Speaker 1: part of the privacy uh stance that Google takes is 97 00:05:48,160 --> 00:05:51,680 Speaker 1: you know, people were objecting to having their their pictures 98 00:05:51,680 --> 00:05:53,840 Speaker 1: on Google street View. So what Google did was they 99 00:05:53,839 --> 00:05:57,320 Speaker 1: create this algorithm that looks for the human face and 100 00:05:57,360 --> 00:06:00,440 Speaker 1: then it applies a blurring layer over it so that 101 00:06:00,520 --> 00:06:03,760 Speaker 1: you can't tell who that is. Right. Well, I've seen 102 00:06:03,800 --> 00:06:07,160 Speaker 1: that work on things that were not human beings, mostly 103 00:06:07,279 --> 00:06:10,479 Speaker 1: on things like like billboards or there was one that 104 00:06:10,600 --> 00:06:11,880 Speaker 1: was a mural that was on the side of a 105 00:06:11,920 --> 00:06:14,720 Speaker 1: building and the the face on the mural had been 106 00:06:14,720 --> 00:06:17,479 Speaker 1: blurred out automatically by Google street View. Well it is 107 00:06:17,480 --> 00:06:22,120 Speaker 1: a face. Yeah, it's just an actual three dimensional face. 108 00:06:22,160 --> 00:06:24,400 Speaker 1: There's a good chance that mural did not object to 109 00:06:24,520 --> 00:06:30,400 Speaker 1: having its privacy, uh review, you know, violated. But anyway, 110 00:06:30,720 --> 00:06:34,800 Speaker 1: so it's looking for that basic set and that's just 111 00:06:34,920 --> 00:06:38,320 Speaker 1: your your your very basic face detection. It it it 112 00:06:38,320 --> 00:06:41,839 Speaker 1: it says, it looks for this this pattern of images 113 00:06:42,200 --> 00:06:45,680 Speaker 1: and says this is what a human face is. Now 114 00:06:46,000 --> 00:06:49,320 Speaker 1: when you go to face recognition, where it's going beyond 115 00:06:49,360 --> 00:06:52,520 Speaker 1: detecting a face, it's actually recognizing a face and and 116 00:06:52,680 --> 00:06:56,520 Speaker 1: setting that to an identity, we get a little more complex, 117 00:06:56,520 --> 00:07:00,040 Speaker 1: actually a lot more complex, true enough. Um, yeah, it 118 00:07:00,440 --> 00:07:04,080 Speaker 1: does rely on those facial landmarks. Yes, you know, your 119 00:07:04,120 --> 00:07:06,880 Speaker 1: eyes and nose chin, the depth of depth of your 120 00:07:06,880 --> 00:07:11,920 Speaker 1: eye sockets, the length of your nose, wid your nose, yeah, nostrils, 121 00:07:12,000 --> 00:07:15,520 Speaker 1: everything like that can be part of this. And and 122 00:07:15,600 --> 00:07:19,280 Speaker 1: for the for the camera in this case, it's recognizing 123 00:07:19,280 --> 00:07:23,360 Speaker 1: a face print sort of like your thumb print or fingerprint. Um. 124 00:07:23,440 --> 00:07:26,680 Speaker 1: And so once it it can keep a record of 125 00:07:26,720 --> 00:07:30,720 Speaker 1: what a particular face looks like, um, then it can 126 00:07:31,040 --> 00:07:33,800 Speaker 1: can you know that? That's I think you would probably 127 00:07:33,840 --> 00:07:36,480 Speaker 1: call that the first step and being able to identify 128 00:07:36,520 --> 00:07:39,320 Speaker 1: a particular face. So you take a person's face and 129 00:07:39,480 --> 00:07:41,920 Speaker 1: you look at these different measurements. It might be the 130 00:07:42,600 --> 00:07:46,200 Speaker 1: distance between the eyes. Uh, it can be things like 131 00:07:46,280 --> 00:07:50,440 Speaker 1: the width of the eyes themselves, um, the where the 132 00:07:50,560 --> 00:07:55,760 Speaker 1: ears are in relation to the the head as a whole, um, 133 00:07:55,880 --> 00:07:59,160 Speaker 1: the jawline, all of these kind of features that you 134 00:07:59,160 --> 00:08:02,600 Speaker 1: really want to focus on. Features that are not easily changeable, right. 135 00:08:02,680 --> 00:08:06,640 Speaker 1: You know, like hairstyle would be a bad measurement because 136 00:08:06,680 --> 00:08:09,280 Speaker 1: you could easily change that, right, right. You want things 137 00:08:09,320 --> 00:08:12,320 Speaker 1: that won't change over time, right, So you know it 138 00:08:12,400 --> 00:08:14,800 Speaker 1: might be things like again the width of the forehead, 139 00:08:14,880 --> 00:08:18,120 Speaker 1: that sort of stuff. These are all called nodal points, 140 00:08:18,800 --> 00:08:21,720 Speaker 1: all right, And uh, we have a great article about 141 00:08:21,760 --> 00:08:25,640 Speaker 1: facial recognition technology on how stuff Works dot com. And 142 00:08:25,920 --> 00:08:28,240 Speaker 1: in that article you learned that that the human face 143 00:08:28,280 --> 00:08:32,840 Speaker 1: has around eighty nodal points. Now, not all facial recognition 144 00:08:32,920 --> 00:08:35,280 Speaker 1: technology is going to rely on all e d of 145 00:08:35,320 --> 00:08:38,600 Speaker 1: those in order to create a face print, but they'll 146 00:08:39,200 --> 00:08:43,280 Speaker 1: rely on some combination of those nodal points and through 147 00:08:43,320 --> 00:08:46,800 Speaker 1: the measurements will come up with a numeric value, which 148 00:08:47,320 --> 00:08:49,600 Speaker 1: is the that's the equivalent of the face print. It's 149 00:08:49,600 --> 00:08:52,800 Speaker 1: a numeric value that is unique to that person. More 150 00:08:52,880 --> 00:08:59,000 Speaker 1: or less, identical twins actually can have the same face print. Wow, yeah, 151 00:08:59,120 --> 00:09:03,560 Speaker 1: it's it's so some facial recognition technology cannot discern between 152 00:09:04,080 --> 00:09:07,360 Speaker 1: identical twins. There are other kinds that rely on even 153 00:09:07,400 --> 00:09:10,840 Speaker 1: more specific data that can discern between the two. But 154 00:09:11,040 --> 00:09:16,079 Speaker 1: a basic facial recognition camera, Um, if you had identical 155 00:09:16,120 --> 00:09:21,440 Speaker 1: twins and they really were identical, yeah, like they didn't 156 00:09:21,480 --> 00:09:23,720 Speaker 1: have one of them didn't have some sort of facial 157 00:09:23,760 --> 00:09:28,400 Speaker 1: feature that was remarkably different. Uh, this software might identify 158 00:09:28,520 --> 00:09:32,720 Speaker 1: both as being the same person without without additional layers. 159 00:09:32,720 --> 00:09:34,760 Speaker 1: And we'll get into that in a little bit. So 160 00:09:35,200 --> 00:09:37,559 Speaker 1: you've got this face print, you create a database of 161 00:09:37,640 --> 00:09:40,680 Speaker 1: face prints. Then when you take a picture of someone 162 00:09:41,280 --> 00:09:46,000 Speaker 1: the again, the camera will measure the noal points on 163 00:09:46,040 --> 00:09:49,680 Speaker 1: this person's face, compare it against the database and see 164 00:09:49,679 --> 00:09:54,040 Speaker 1: if there's a match. And it's probably I'm guessing we 165 00:09:54,080 --> 00:09:56,079 Speaker 1: don't really go into a whole detail in our article, 166 00:09:56,080 --> 00:09:59,200 Speaker 1: but I'm guessing it's kind of like fingerprints in a way, 167 00:09:59,240 --> 00:10:04,080 Speaker 1: you look for a percentage of probability that this this 168 00:10:04,240 --> 00:10:07,640 Speaker 1: particular face matches one that's in the database, because you 169 00:10:07,720 --> 00:10:11,280 Speaker 1: got to remember, not all of these images are going 170 00:10:11,320 --> 00:10:15,480 Speaker 1: to be exactly the same. Early facial recognition technology was 171 00:10:15,600 --> 00:10:18,440 Speaker 1: very limited. You had to have someone looking directly into 172 00:10:18,480 --> 00:10:20,720 Speaker 1: the camera, and then you would have to have that 173 00:10:20,800 --> 00:10:24,280 Speaker 1: same person look directly into the camera again later on 174 00:10:24,760 --> 00:10:27,120 Speaker 1: and compare that to the database in order to find 175 00:10:27,120 --> 00:10:29,400 Speaker 1: a match. If the person was looking a little to 176 00:10:29,440 --> 00:10:31,640 Speaker 1: the left or to the right, the technology wasn't good 177 00:10:31,800 --> 00:10:35,160 Speaker 1: enough to to compensate for that and to make a 178 00:10:35,200 --> 00:10:38,120 Speaker 1: model of that person's face to really get the right 179 00:10:38,120 --> 00:10:42,120 Speaker 1: measurements right. So, and you've got to think of all 180 00:10:42,120 --> 00:10:44,520 Speaker 1: the other factors that play into this. It's the lighting. 181 00:10:45,520 --> 00:10:49,760 Speaker 1: If the lighting is bad, those early facial recognition technologies 182 00:10:49,880 --> 00:10:52,280 Speaker 1: weren't very effective. Or if the person was at a 183 00:10:52,320 --> 00:10:56,000 Speaker 1: different distance. The camera has to know how far away 184 00:10:56,000 --> 00:10:59,800 Speaker 1: you are in order to make valid measurements for things 185 00:10:59,800 --> 00:11:02,240 Speaker 1: like how far apart your eyes are. If the camera 186 00:11:02,320 --> 00:11:05,079 Speaker 1: thinks you're fifteen ft away but you're really twelve feet away, 187 00:11:05,360 --> 00:11:07,360 Speaker 1: those measurements are not going to be accurate, and it's 188 00:11:07,400 --> 00:11:10,080 Speaker 1: not going to find the right match in the database. Right, 189 00:11:10,920 --> 00:11:12,880 Speaker 1: So this is this was definitely one of those things 190 00:11:12,880 --> 00:11:14,600 Speaker 1: that was a big learning curve you had to be 191 00:11:14,640 --> 00:11:17,520 Speaker 1: able to build. You had to develop the digital technology 192 00:11:17,520 --> 00:11:22,840 Speaker 1: to detect distance and then accurately measure as many Noble 193 00:11:22,920 --> 00:11:25,360 Speaker 1: points as possible and as little time as possible. And 194 00:11:25,360 --> 00:11:28,120 Speaker 1: we're talking hundreds of a second here. Uh that that 195 00:11:28,400 --> 00:11:32,400 Speaker 1: a chip is scanning a person's face and identifying those 196 00:11:32,440 --> 00:11:34,160 Speaker 1: nobal points. I mean, it takes no time at all 197 00:11:34,200 --> 00:11:38,200 Speaker 1: for this to happen, but it took engineers time to 198 00:11:38,280 --> 00:11:43,000 Speaker 1: develop that technology. And yeah, there there are some really 199 00:11:43,040 --> 00:11:48,360 Speaker 1: fascinating technologies built into facial recognition, including you know, technologies 200 00:11:48,400 --> 00:11:53,280 Speaker 1: such as a surface texture analysis. It's basically creating a 201 00:11:52,240 --> 00:11:55,720 Speaker 1: a another not a face print, but a skin print. 202 00:11:56,040 --> 00:11:58,599 Speaker 1: Where they are doing, um, the system is doing a 203 00:11:59,000 --> 00:12:03,960 Speaker 1: mathematical analogy of sections of your skin if you come 204 00:12:04,000 --> 00:12:08,439 Speaker 1: before the camera. So um, if you had to say 205 00:12:08,480 --> 00:12:11,280 Speaker 1: a birthmark or a mole or something, that would probably 206 00:12:11,320 --> 00:12:14,400 Speaker 1: help it track down who you are because it's going 207 00:12:14,440 --> 00:12:16,440 Speaker 1: to say, well, we know that this is you know 208 00:12:16,520 --> 00:12:20,760 Speaker 1: in sector number seventeen. Um, you know there is this 209 00:12:21,280 --> 00:12:23,720 Speaker 1: different coloration than there there would be in the rest 210 00:12:23,760 --> 00:12:27,160 Speaker 1: of of the face, so um that that you know, 211 00:12:27,200 --> 00:12:31,360 Speaker 1: they're very sophisticated in breaking them down into um, you know, 212 00:12:31,400 --> 00:12:36,760 Speaker 1: all kinds of mathematical UH constructs to enable this to work. Yeah, 213 00:12:36,760 --> 00:12:39,240 Speaker 1: And the service texture analysis is that layer I was 214 00:12:39,280 --> 00:12:42,920 Speaker 1: talking about earlier about how to identify between identical twins. 215 00:12:43,920 --> 00:12:48,280 Speaker 1: Service texture analysis is actually the kind of of technology 216 00:12:48,320 --> 00:12:51,280 Speaker 1: you want in order to do that, because it's looking 217 00:12:51,360 --> 00:12:53,920 Speaker 1: much more closely at the texture of your skin. As 218 00:12:53,960 --> 00:12:58,559 Speaker 1: the name would imply, so uh, even identical twins aren't 219 00:12:58,559 --> 00:13:02,960 Speaker 1: going to have identical lines on their faces, you know, laugh, lines, wrinkles, 220 00:13:03,000 --> 00:13:05,040 Speaker 1: that kind of thing. There might be a freckle or 221 00:13:05,080 --> 00:13:08,360 Speaker 1: a mole that's it's slightly different from one twin to 222 00:13:08,400 --> 00:13:11,040 Speaker 1: the other. And this is the sort of technology that's 223 00:13:11,080 --> 00:13:13,200 Speaker 1: going to pick up on that, as opposed to the 224 00:13:13,240 --> 00:13:16,680 Speaker 1: basic facial recognition technology that might not it might be 225 00:13:16,720 --> 00:13:19,160 Speaker 1: close enough between the two twins in order to identify 226 00:13:19,200 --> 00:13:21,880 Speaker 1: that as the same person. But with this layer, it 227 00:13:21,920 --> 00:13:28,400 Speaker 1: can make it more accurate. In fact, according to one company, UH, 228 00:13:28,440 --> 00:13:33,559 Speaker 1: it can, which is called Identics. UH. Surface texture analysis 229 00:13:33,600 --> 00:13:36,760 Speaker 1: can increase the validity or the the reliability of a 230 00:13:36,760 --> 00:13:41,360 Speaker 1: scan by around which is pretty significant. I mean, you know, 231 00:13:41,480 --> 00:13:45,600 Speaker 1: you know, depending upon how accurate your starting point is. 232 00:13:46,200 --> 00:13:49,120 Speaker 1: The early early facial recognition technology, even when it was 233 00:13:49,160 --> 00:13:52,240 Speaker 1: working well, was not working that well. It was like 234 00:13:52,280 --> 00:13:56,600 Speaker 1: around a sixty success rate. And uh, you think about that, 235 00:13:56,640 --> 00:13:59,280 Speaker 1: it means of the time they get it wrong, and 236 00:13:59,760 --> 00:14:01,880 Speaker 1: that can be pretty serious when you consider that. A 237 00:14:01,880 --> 00:14:04,840 Speaker 1: lot of the facial recognition technology that's out there is 238 00:14:04,920 --> 00:14:08,400 Speaker 1: used in law enforcement practices. Yes, it's in order to 239 00:14:08,520 --> 00:14:13,319 Speaker 1: identify people who you know, let's say that there's unknown 240 00:14:13,640 --> 00:14:17,439 Speaker 1: suspect that is on the loose, may be used to 241 00:14:17,440 --> 00:14:19,760 Speaker 1: try and identify someone like that, or it may even 242 00:14:19,760 --> 00:14:23,880 Speaker 1: be used in a static system where if something happens 243 00:14:23,920 --> 00:14:26,200 Speaker 1: within the view of the camera, the camera would be 244 00:14:26,280 --> 00:14:30,160 Speaker 1: able to to identify that person more quickly. You wouldn't 245 00:14:30,200 --> 00:14:32,200 Speaker 1: have to just you know, stare at this picture and say, 246 00:14:32,280 --> 00:14:34,080 Speaker 1: I wonder if this is you know, is this a 247 00:14:34,160 --> 00:14:36,600 Speaker 1: suspect or you would have compared against the database of 248 00:14:36,680 --> 00:14:39,560 Speaker 1: like mug shots or whatever, and um, that can be 249 00:14:39,600 --> 00:14:42,880 Speaker 1: really useful in UM in urban environments that have lots 250 00:14:42,920 --> 00:14:45,880 Speaker 1: and lots of cameras that they use for law enforcement purposes, 251 00:14:46,160 --> 00:14:49,120 Speaker 1: places like London where there are so many cameras on 252 00:14:49,160 --> 00:14:53,680 Speaker 1: all the different street corners, but also in places like banks. Um, 253 00:14:53,760 --> 00:14:58,680 Speaker 1: if somebody were unknown uh you know, unknown bank robber 254 00:14:58,800 --> 00:15:01,880 Speaker 1: and they didn't where you know, some kind of facial 255 00:15:01,920 --> 00:15:05,480 Speaker 1: obscuring gear. Now, I mean it would be easy enough 256 00:15:05,520 --> 00:15:07,800 Speaker 1: to fool a bank camera, probably if you put on 257 00:15:07,800 --> 00:15:09,800 Speaker 1: a fake mustache or something like that, because that's the 258 00:15:09,840 --> 00:15:13,720 Speaker 1: kind of thing that's going to sort off the facial recognition. Yeah, 259 00:15:13,720 --> 00:15:17,440 Speaker 1: there's some facial recognition technologies that are they're sophistical enough 260 00:15:17,480 --> 00:15:20,360 Speaker 1: to ignore things like facial hair. It'll look at the 261 00:15:20,840 --> 00:15:24,000 Speaker 1: shape and size of your face and the relationship of say, 262 00:15:24,120 --> 00:15:26,960 Speaker 1: like again the distance between your eyes, and it ignores 263 00:15:27,000 --> 00:15:28,920 Speaker 1: things like facial hair because again, facial hair is one 264 00:15:28,920 --> 00:15:32,160 Speaker 1: of those things it's easy to to grow or remove, 265 00:15:32,360 --> 00:15:37,200 Speaker 1: right for some of us anyway, ladies, hopefully not for you. 266 00:15:37,800 --> 00:15:42,840 Speaker 1: But the the most facial recognition technology will ignore that 267 00:15:42,920 --> 00:15:45,240 Speaker 1: kind of stuff. But yeah, if you're wearing something that's 268 00:15:45,240 --> 00:15:48,920 Speaker 1: obscuring your face, that definitely will throw off facial has 269 00:15:48,920 --> 00:15:50,960 Speaker 1: to and that's not gonna be able to recognize it 270 00:15:50,960 --> 00:15:53,760 Speaker 1: because I can't see it. Um. So if you walk 271 00:15:53,760 --> 00:15:56,360 Speaker 1: into the bank wearing a balla clava, the camera probably 272 00:15:56,400 --> 00:15:59,680 Speaker 1: won't recognize you, but the security guard might have some issues, right, 273 00:16:00,080 --> 00:16:03,200 Speaker 1: And this also leads us to the question about what 274 00:16:03,320 --> 00:16:06,120 Speaker 1: sort of problems there can be. Well, clearly, privacy is 275 00:16:06,320 --> 00:16:09,480 Speaker 1: a big concern, right. I mean, if you get to 276 00:16:09,520 --> 00:16:12,040 Speaker 1: the point where you have technology that can recognize faces, 277 00:16:12,440 --> 00:16:14,760 Speaker 1: then you've got your your to the point where you 278 00:16:14,800 --> 00:16:20,280 Speaker 1: have to worry about you being viewed wherever you happen 279 00:16:20,320 --> 00:16:23,440 Speaker 1: to be and identified as being there. And I mean 280 00:16:23,480 --> 00:16:25,760 Speaker 1: that's a big problem. I mean even for people who, 281 00:16:26,080 --> 00:16:33,000 Speaker 1: let's say that you are perfectly innocent, upright wonderful citizen 282 00:16:33,280 --> 00:16:36,520 Speaker 1: and you've never done anything wrong, you still probably wouldn't 283 00:16:36,560 --> 00:16:40,240 Speaker 1: necessarily want all these cameras everywhere identifying your your you 284 00:16:40,440 --> 00:16:43,560 Speaker 1: wherever you happen to be. I mean, it's just it's 285 00:16:43,560 --> 00:16:46,480 Speaker 1: it's an invasion of privacy. And because of that, there 286 00:16:46,480 --> 00:16:50,040 Speaker 1: have been some experiments with this sort of system in 287 00:16:50,120 --> 00:16:54,480 Speaker 1: place in various public areas that ended up getting um 288 00:16:55,240 --> 00:16:59,720 Speaker 1: canceled somewhere along the project because either the public was 289 00:17:00,320 --> 00:17:02,720 Speaker 1: in an uproar about it, saying, hey, this violates our 290 00:17:02,720 --> 00:17:06,440 Speaker 1: privacy and I'm not comfortable with any agency tracking me 291 00:17:06,680 --> 00:17:12,199 Speaker 1: like this, or the reliability was low enough so that 292 00:17:12,240 --> 00:17:15,200 Speaker 1: there were concerns of Hey, I could be at home 293 00:17:16,800 --> 00:17:20,119 Speaker 1: asleep and someone who looks enough like me for the 294 00:17:20,119 --> 00:17:23,800 Speaker 1: facial recognition technology to think that that was me could 295 00:17:23,840 --> 00:17:25,920 Speaker 1: commit a crime, and then I could be charged for it. 296 00:17:26,680 --> 00:17:30,600 Speaker 1: And I mean, if the if the reliable reliability is low, 297 00:17:30,760 --> 00:17:34,240 Speaker 1: like if it it wasn't that six, that's a legitimate concern. 298 00:17:35,960 --> 00:17:37,320 Speaker 1: Why if you happen to be involved in one of 299 00:17:37,320 --> 00:17:42,560 Speaker 1: those mistakes, like Jonathan, why did you break into Tiffany's 300 00:17:42,560 --> 00:17:44,600 Speaker 1: and steal all these diamonds? And I'd be like, wait, 301 00:17:44,640 --> 00:17:48,159 Speaker 1: what did I do? Now? When did that? I? I 302 00:17:48,200 --> 00:17:50,200 Speaker 1: am certain I did not do that. It does not 303 00:17:50,320 --> 00:17:52,680 Speaker 1: say I ever went to Tiffany's on four square, So 304 00:17:52,880 --> 00:17:57,520 Speaker 1: clearly that wasn't me because I check in everywhere, right. 305 00:17:58,840 --> 00:18:02,040 Speaker 1: But there can it can be a lot of positive 306 00:18:02,119 --> 00:18:04,679 Speaker 1: uses of course for facial recognition. Oh yeah, no, there 307 00:18:04,840 --> 00:18:06,679 Speaker 1: there are plenty of really good ones. I mean not 308 00:18:06,720 --> 00:18:09,760 Speaker 1: that those aren't positive, but I mean more fun let's 309 00:18:09,760 --> 00:18:13,679 Speaker 1: see uses. Of course. You know, with cameras being as 310 00:18:13,720 --> 00:18:16,720 Speaker 1: sophisticated as they are, photos are geo tagged, and now 311 00:18:17,200 --> 00:18:19,840 Speaker 1: you know the facial recognition software built into them, you 312 00:18:19,880 --> 00:18:24,560 Speaker 1: can auto tag different photos as basically as you are, 313 00:18:24,600 --> 00:18:28,159 Speaker 1: you know, importing them into your files. Yeah, that to 314 00:18:28,200 --> 00:18:33,160 Speaker 1: me is absolutely amazing stuff. I was amazed back when 315 00:18:33,200 --> 00:18:36,280 Speaker 1: cameras first started being able to detect faces. That to 316 00:18:36,359 --> 00:18:38,280 Speaker 1: me was that to me was really really cool. I 317 00:18:38,320 --> 00:18:40,800 Speaker 1: was like, hey, awesome. And then you had the next 318 00:18:40,840 --> 00:18:44,320 Speaker 1: step beyond detecting faces was detecting when someone was smiling, 319 00:18:44,840 --> 00:18:49,040 Speaker 1: because then they were measuring the person's mouth right, and 320 00:18:49,119 --> 00:18:52,320 Speaker 1: so if a person was smiling or making some sort 321 00:18:52,359 --> 00:18:56,320 Speaker 1: of facial expression that was akin to smiling, because sometimes 322 00:18:56,320 --> 00:18:59,000 Speaker 1: it was like more of a grimace. Uh, your camera 323 00:18:59,119 --> 00:19:02,520 Speaker 1: would merely take the picture. There's some cameras that had 324 00:19:02,520 --> 00:19:04,320 Speaker 1: it where it would it would be ready to take 325 00:19:04,320 --> 00:19:06,200 Speaker 1: the photo and as soon as the person smiled, that's 326 00:19:06,200 --> 00:19:08,399 Speaker 1: when it would take the shot. So that way you 327 00:19:08,440 --> 00:19:11,600 Speaker 1: would get the smile right. And then the next step 328 00:19:11,600 --> 00:19:14,160 Speaker 1: beyond that is what you were talking about, where you 329 00:19:14,200 --> 00:19:16,639 Speaker 1: would tag a photo. You take a picture of someone, 330 00:19:17,119 --> 00:19:19,919 Speaker 1: you tag that that photo with the person's name, and 331 00:19:19,960 --> 00:19:23,280 Speaker 1: then from that point forward, your camera compares the pictures 332 00:19:23,320 --> 00:19:26,280 Speaker 1: you take against the people you've already tagged and says, 333 00:19:26,280 --> 00:19:28,640 Speaker 1: oh wait, this is a person he's tagged already I'm 334 00:19:28,680 --> 00:19:30,880 Speaker 1: just gonna go ahead and throw the tag on there. Yeah, 335 00:19:30,920 --> 00:19:35,000 Speaker 1: that's a good point because just as the manufacturers roly 336 00:19:35,280 --> 00:19:40,360 Speaker 1: on the software engineers to build the technology in so 337 00:19:40,400 --> 00:19:43,840 Speaker 1: that these devices can recognize faces, you have to teach 338 00:19:43,880 --> 00:19:49,280 Speaker 1: it who is who? Um? Who is whom? I should say, um, 339 00:19:49,400 --> 00:19:52,120 Speaker 1: so you know, it won't know that when I take 340 00:19:52,160 --> 00:19:55,199 Speaker 1: a photo of Jonathan that is Jonathan until I tell it, 341 00:19:55,440 --> 00:19:57,520 Speaker 1: you know, this is who this is, and then therefore 342 00:19:57,680 --> 00:19:59,880 Speaker 1: after that it will take photos right well, I reckon 343 00:20:00,040 --> 00:20:03,040 Speaker 1: nize his face. It's Jonathan Strickland. And you know when 344 00:20:03,040 --> 00:20:06,800 Speaker 1: I upload those embarrassing photos of him to Facebook, it will, 345 00:20:06,840 --> 00:20:10,840 Speaker 1: you know, have all the impertinent information included properly, which 346 00:20:10,880 --> 00:20:12,600 Speaker 1: is why every evening I have to sit down and 347 00:20:12,760 --> 00:20:15,600 Speaker 1: untagged photos. Yeah, but you know, I was thinking of 348 00:20:15,640 --> 00:20:21,119 Speaker 1: another uh technology facial recognition application I should say, Um, 349 00:20:21,160 --> 00:20:26,160 Speaker 1: that is also pretty fun, which is uh Microsoft's connect. Ah. Yes, 350 00:20:26,359 --> 00:20:28,720 Speaker 1: that's a good point. Yeah, connect is of course, that's 351 00:20:28,800 --> 00:20:36,400 Speaker 1: the motion detecting right always, because that's what the that's 352 00:20:36,440 --> 00:20:38,679 Speaker 1: that's what the project was called before it hit the public. 353 00:20:39,080 --> 00:20:40,919 Speaker 1: How you associate a name with something and then it's 354 00:20:40,960 --> 00:20:43,600 Speaker 1: sort of hard to unassociated with that. I still call 355 00:20:43,720 --> 00:20:50,840 Speaker 1: segways ginger. That's it. That's going way back. So anyway, Yeah, 356 00:20:50,920 --> 00:20:56,040 Speaker 1: Connect has facial recognition built into it, and it's the implementation. 357 00:20:56,119 --> 00:21:00,200 Speaker 1: Like you were saying, pullet is a great idea weather 358 00:21:00,359 --> 00:21:02,400 Speaker 1: weather or not it works well, I can't say because 359 00:21:02,400 --> 00:21:05,439 Speaker 1: I haven't used Connect yet, but the I love the 360 00:21:05,440 --> 00:21:08,360 Speaker 1: concept that when you step in front of your your 361 00:21:08,480 --> 00:21:13,320 Speaker 1: entertainment center, the camera in the Connect peripheral looks at 362 00:21:13,359 --> 00:21:16,640 Speaker 1: you and then analyzes your face. Does this this process 363 00:21:16,680 --> 00:21:19,280 Speaker 1: that we're talking about where it measures, takes these measurements 364 00:21:19,359 --> 00:21:23,000 Speaker 1: very quickly compares that to the information and database and 365 00:21:23,040 --> 00:21:26,000 Speaker 1: then identifies you. So if you have previously set up 366 00:21:26,040 --> 00:21:30,360 Speaker 1: and a Connect account, let's say with your Xbox, it knows, Hey, 367 00:21:30,359 --> 00:21:33,600 Speaker 1: this is Jonathan. Jonathan likes to play at this particular 368 00:21:33,640 --> 00:21:37,440 Speaker 1: skill level. Jonathan likes these particular games. Um, I'm going 369 00:21:37,520 --> 00:21:42,520 Speaker 1: to present this this block of information and features and 370 00:21:42,640 --> 00:21:46,760 Speaker 1: games to Jonathan because we already know what his preferences are. 371 00:21:47,600 --> 00:21:50,080 Speaker 1: Then when someone else, let's say, Chris, steps in front 372 00:21:50,080 --> 00:21:52,760 Speaker 1: of Connect, it will identify Chris and say, oh, well 373 00:21:52,800 --> 00:21:55,480 Speaker 1: that's Chris. Chris like some of the games Jonathan likes, 374 00:21:55,480 --> 00:21:57,960 Speaker 1: but he also likes these other games, and he prefers 375 00:21:58,040 --> 00:22:00,680 Speaker 1: these kind of movies to the movies that Jonathan likes, 376 00:22:00,720 --> 00:22:03,960 Speaker 1: like Chris prefers the three movies they've seen before to 377 00:22:04,119 --> 00:22:07,320 Speaker 1: the vast database of movies that Jonathan has seen. So 378 00:22:07,400 --> 00:22:09,080 Speaker 1: I'm just going to show them these three movies because 379 00:22:09,080 --> 00:22:10,640 Speaker 1: there's no point in trying to get him to watch 380 00:22:10,680 --> 00:22:14,520 Speaker 1: anything else. And then, um, but that's the thing is 381 00:22:14,560 --> 00:22:18,600 Speaker 1: that it'll it'll tailor the experience to the person based 382 00:22:18,640 --> 00:22:21,680 Speaker 1: upon that person's profile. Now again, like Plett was saying, 383 00:22:21,760 --> 00:22:23,760 Speaker 1: you have to create a profile. You have to tell 384 00:22:23,800 --> 00:22:26,440 Speaker 1: connect Hey, this is who I am. Whenever you see 385 00:22:26,480 --> 00:22:30,119 Speaker 1: this face, this is the profile you should use. Now, 386 00:22:30,680 --> 00:22:32,520 Speaker 1: I'm sorry you're going to say something, Well, I was 387 00:22:32,560 --> 00:22:34,679 Speaker 1: just if you were going to extrapolate from that I 388 00:22:34,680 --> 00:22:37,760 Speaker 1: thought of I just thought of another application similar to 389 00:22:37,960 --> 00:22:43,120 Speaker 1: that that Apparently, as I look it up quickly, here, Um, 390 00:22:43,240 --> 00:22:45,680 Speaker 1: other people have our way ahead of me, and doesn't 391 00:22:45,680 --> 00:22:49,359 Speaker 1: surprise me in the least. Um, smart homes can do 392 00:22:49,400 --> 00:22:52,280 Speaker 1: the same thing, because I know that. I remember reading 393 00:22:52,320 --> 00:22:55,960 Speaker 1: a long time ago that when Bill Gates had built 394 00:22:56,040 --> 00:23:00,320 Speaker 1: his you know, massive home with the Microsoft technology state 395 00:23:00,359 --> 00:23:02,479 Speaker 1: of the art, and it would tell as soon as 396 00:23:02,480 --> 00:23:05,080 Speaker 1: you walked into a room, uh, oh, well, you know 397 00:23:05,160 --> 00:23:07,560 Speaker 1: this is Bill. He likes the temperature at you know, 398 00:23:07,600 --> 00:23:10,320 Speaker 1: seventy two degrees, he likes this kind of music, and 399 00:23:10,320 --> 00:23:12,480 Speaker 1: it would automatically like you could walk around the house 400 00:23:12,520 --> 00:23:14,840 Speaker 1: and and I remember reading this. I don't know if 401 00:23:14,880 --> 00:23:16,560 Speaker 1: it's still true or not, but they could follow you 402 00:23:16,640 --> 00:23:18,560 Speaker 1: with he's like, oh, well, you know, we'll turn on 403 00:23:18,600 --> 00:23:20,520 Speaker 1: the music in this room, we'll turn it off in 404 00:23:20,520 --> 00:23:23,080 Speaker 1: that room because he's no longer there, because you know, 405 00:23:23,200 --> 00:23:24,960 Speaker 1: we know where he is. But you would have to 406 00:23:25,680 --> 00:23:28,280 Speaker 1: as I remember correctly, Um, like I said, I just 407 00:23:28,280 --> 00:23:29,960 Speaker 1: thought of this on the fly and didn't research it. 408 00:23:30,000 --> 00:23:32,680 Speaker 1: But as I remember correctly, it relied on some kind 409 00:23:32,680 --> 00:23:34,879 Speaker 1: of r F I D technology. You had to be 410 00:23:34,920 --> 00:23:38,200 Speaker 1: wearing something like like a little name tag that you wore, 411 00:23:38,440 --> 00:23:43,320 Speaker 1: but you could you could use facial recognition technology instead, sure, 412 00:23:43,480 --> 00:23:45,560 Speaker 1: and not have to carry anything with you. And then 413 00:23:45,600 --> 00:23:47,800 Speaker 1: you wouldn't have to go, oh, well, you know, dude, 414 00:23:47,800 --> 00:23:50,680 Speaker 1: I left my card in the in the laundry room, 415 00:23:51,080 --> 00:23:52,360 Speaker 1: and now I'm in the living room and I really 416 00:23:52,359 --> 00:23:54,120 Speaker 1: don't feel like it not right. You could tell where 417 00:23:54,119 --> 00:23:57,280 Speaker 1: Pallette was based upon the game show and commercial music 418 00:23:57,320 --> 00:23:59,920 Speaker 1: that you heard throughout the place. Oh nice, thank you, 419 00:24:00,080 --> 00:24:02,800 Speaker 1: welcome much. For me, it would be musicals. So I mean, 420 00:24:02,800 --> 00:24:06,679 Speaker 1: I'm gonna go ahead and say, like, who's doing Fossey. 421 00:24:07,000 --> 00:24:10,040 Speaker 1: That's gotta be Jonathan. But yeah, I mean it would be. 422 00:24:10,080 --> 00:24:11,800 Speaker 1: It would be highly useful. And there are people, as 423 00:24:11,840 --> 00:24:14,480 Speaker 1: I you know, run a quick search on the search engine, 424 00:24:14,480 --> 00:24:17,320 Speaker 1: that are already ahead of me on this and have 425 00:24:17,600 --> 00:24:20,320 Speaker 1: you know, started implementing that technology. Of course, it also 426 00:24:20,359 --> 00:24:22,200 Speaker 1: means you have to have cameras everywhere in your house, 427 00:24:22,280 --> 00:24:25,479 Speaker 1: right right, Yeah, they're They're definitely tradeoffs here. It's, like 428 00:24:25,520 --> 00:24:28,320 Speaker 1: I said, the big one being privacy. Um. There's also 429 00:24:28,440 --> 00:24:31,200 Speaker 1: been discussions of using the story of technology for things 430 00:24:31,240 --> 00:24:33,520 Speaker 1: like a t M. A t M so I'm not 431 00:24:33,560 --> 00:24:35,600 Speaker 1: gonna say that, I was gonna say a t M machines. 432 00:24:36,119 --> 00:24:38,720 Speaker 1: I apologize for the redundancy. It's funny how that gets 433 00:24:38,720 --> 00:24:42,240 Speaker 1: into vernacular. But anyway, a t M S could use 434 00:24:42,320 --> 00:24:46,000 Speaker 1: this to identify a person and then theoretically you could 435 00:24:46,160 --> 00:24:49,440 Speaker 1: get money out of your checking Accounter Stavings account without 436 00:24:49,440 --> 00:24:52,240 Speaker 1: having to have a pen number or anything. Did it again? 437 00:24:52,320 --> 00:24:56,280 Speaker 1: I did it again twice? An all pen number. Okay, 438 00:24:57,440 --> 00:25:00,880 Speaker 1: without having to have a pen or any other identification 439 00:25:01,040 --> 00:25:03,560 Speaker 1: identification on you, you could just it would see your 440 00:25:03,600 --> 00:25:05,439 Speaker 1: face and know that that was you and identify you 441 00:25:05,480 --> 00:25:08,920 Speaker 1: with the account. Now, there are some definite problems there 442 00:25:09,000 --> 00:25:11,840 Speaker 1: because if you have my technical twins, Hey, I'm gonna 443 00:25:11,840 --> 00:25:16,080 Speaker 1: go get my brother's cash out of his account today. Uh. Also, 444 00:25:16,400 --> 00:25:17,879 Speaker 1: I'm sorry, No, I was just gonna say, but it 445 00:25:17,920 --> 00:25:21,320 Speaker 1: could cut down on skimming. It's it's a hard it's 446 00:25:21,359 --> 00:25:24,399 Speaker 1: a hard thing to say because I'm reminded of you 447 00:25:24,440 --> 00:25:26,560 Speaker 1: may have heard the story. It was I think a 448 00:25:26,560 --> 00:25:29,240 Speaker 1: couple of years ago. Actually in Japan. Japan is always 449 00:25:29,240 --> 00:25:31,760 Speaker 1: ahead of us on this sort of stuff. Japan had 450 00:25:32,920 --> 00:25:38,120 Speaker 1: cigarette machines. Cigarette machines had facial recognition technology built into 451 00:25:38,119 --> 00:25:41,640 Speaker 1: them to recognize how old someone is. There was this 452 00:25:42,320 --> 00:25:45,200 Speaker 1: The technology was designed to look for things like wrinkles 453 00:25:45,240 --> 00:25:48,080 Speaker 1: and and laugh lines and that sort of stuff to 454 00:25:48,200 --> 00:25:51,040 Speaker 1: identify a person as being old enough to purchase cigarettes, 455 00:25:51,040 --> 00:25:53,320 Speaker 1: because of course it's a vending machine, so otherwise you're 456 00:25:53,359 --> 00:25:55,840 Speaker 1: just kind of working on the honor system. But the 457 00:25:55,840 --> 00:26:00,680 Speaker 1: news broke that kids could easily bypass this just by 458 00:26:00,720 --> 00:26:03,280 Speaker 1: holding up a picture of an old person's face to 459 00:26:03,320 --> 00:26:06,760 Speaker 1: the camera. They just hold up the picture and the 460 00:26:06,840 --> 00:26:09,520 Speaker 1: camera would detect the old person's face and say, yeah, 461 00:26:09,560 --> 00:26:11,680 Speaker 1: this person could totally buy cigarettes, and then the kids 462 00:26:11,720 --> 00:26:14,119 Speaker 1: could buy as much as they wanted to. Um. So 463 00:26:14,160 --> 00:26:16,040 Speaker 1: there is a concern about, well, if you had a 464 00:26:16,160 --> 00:26:19,200 Speaker 1: high enough resolution picture of someone's face and you held 465 00:26:19,240 --> 00:26:22,359 Speaker 1: it up to the camera, with the camera just be 466 00:26:22,520 --> 00:26:25,240 Speaker 1: unable to distinguish the fact that that's a a two 467 00:26:25,280 --> 00:26:28,520 Speaker 1: dimensional representation of a person's face versus an actual three 468 00:26:28,560 --> 00:26:33,240 Speaker 1: dimensional face, and some technology can't do that very well. 469 00:26:33,960 --> 00:26:36,679 Speaker 1: But it's getting better all the time, and uh, you know, 470 00:26:36,760 --> 00:26:38,960 Speaker 1: quite possible by the time this podcast goes live that 471 00:26:39,000 --> 00:26:42,000 Speaker 1: will be solved. I would surprise me. I would imagine 472 00:26:42,040 --> 00:26:43,840 Speaker 1: the best way of doing that would be used to 473 00:26:44,119 --> 00:26:47,679 Speaker 1: be to use a two camera system where you have 474 00:26:48,119 --> 00:26:51,960 Speaker 1: essentially binocular vision using a camera with two lenses, and 475 00:26:52,080 --> 00:26:55,800 Speaker 1: that way you create that whole parallax issue that we 476 00:26:55,880 --> 00:26:58,119 Speaker 1: have natively as human beings. As long as we have 477 00:26:58,160 --> 00:27:01,160 Speaker 1: two eyes that work, we have that parallax issue. That's 478 00:27:01,160 --> 00:27:03,879 Speaker 1: what allows us to to see three D and three 479 00:27:03,960 --> 00:27:07,640 Speaker 1: D movies. That's part of the reason. And uh, if 480 00:27:07,640 --> 00:27:09,719 Speaker 1: you created that parallax, and we're able to create an 481 00:27:09,760 --> 00:27:12,560 Speaker 1: algorithm that compared the two images so that it could 482 00:27:12,560 --> 00:27:16,000 Speaker 1: detect whether something was flat or an actual three dimensional object. 483 00:27:16,640 --> 00:27:19,359 Speaker 1: That would get around that problem. Uh. I just said 484 00:27:19,400 --> 00:27:22,480 Speaker 1: something that sounds like it's simple. It's actually incredibly complex. 485 00:27:22,960 --> 00:27:24,960 Speaker 1: But that would be how that would be my approach. 486 00:27:25,720 --> 00:27:28,920 Speaker 1: Just you know, you look at how do we see 487 00:27:28,960 --> 00:27:31,480 Speaker 1: these things and interpret them and then how could we 488 00:27:31,600 --> 00:27:37,000 Speaker 1: copy that within the realm of technology? Interesting? Interesting? Yeah, 489 00:27:37,400 --> 00:27:40,800 Speaker 1: good times man. It's um yeah, it's it's one of 490 00:27:40,800 --> 00:27:43,160 Speaker 1: those things like so many of the other discussions we've 491 00:27:43,200 --> 00:27:46,919 Speaker 1: had where there are so many positive applications but you know, 492 00:27:47,000 --> 00:27:50,960 Speaker 1: their privacy concerns and and uh, possible ways to misuse 493 00:27:51,000 --> 00:27:55,440 Speaker 1: the technology. It's you know, but I'm it's really fascinating though, 494 00:27:55,480 --> 00:27:59,960 Speaker 1: how they've been able to develop the technology to recognize 495 00:28:00,119 --> 00:28:02,400 Speaker 1: is a face at all, you know, especially when it's 496 00:28:02,440 --> 00:28:05,359 Speaker 1: somebody far away in a photo you have a group picture. 497 00:28:05,440 --> 00:28:07,480 Speaker 1: Of course, I think my camera only has the ability 498 00:28:07,520 --> 00:28:11,040 Speaker 1: to recognize as many as six faces at once. Yeah. Um, 499 00:28:11,119 --> 00:28:14,600 Speaker 1: so there's still limitations, right, some can can do like 500 00:28:14,760 --> 00:28:17,320 Speaker 1: ten or so, but yeah, it's it's still one of 501 00:28:17,359 --> 00:28:19,400 Speaker 1: those things that the processors to work pretty hard to 502 00:28:19,480 --> 00:28:22,240 Speaker 1: keep scanning and identifying like that. And plus there's a 503 00:28:22,240 --> 00:28:26,000 Speaker 1: point where you know, if you have a far enough 504 00:28:26,040 --> 00:28:28,160 Speaker 1: distance from the camera, it's probably gonna ignore it because 505 00:28:28,160 --> 00:28:30,159 Speaker 1: it doesn't want to try and focus on something in 506 00:28:30,200 --> 00:28:32,600 Speaker 1: the background at the expense of whatever is in the foreground. 507 00:28:33,600 --> 00:28:36,600 Speaker 1: And um, yeah, it is really cool. It's it's the 508 00:28:37,400 --> 00:28:41,480 Speaker 1: it's it's a step toward artificial intelligence. Is really what 509 00:28:41,520 --> 00:28:46,320 Speaker 1: we're talking about, teaching computers how to to uh observe 510 00:28:46,440 --> 00:28:50,800 Speaker 1: things and identify them. They're still not thinking, but they 511 00:28:50,840 --> 00:28:53,840 Speaker 1: are able to identify stuff. And again, it's really just 512 00:28:54,240 --> 00:28:59,760 Speaker 1: creating measurements and then assigning a numeric value to the 513 00:29:00,000 --> 00:29:03,640 Speaker 1: election of measurements and using that as the identifier. So 514 00:29:04,040 --> 00:29:06,520 Speaker 1: you might think, hey, that's my brother Bill, and your 515 00:29:06,560 --> 00:29:10,600 Speaker 1: camera's thinking, hey that's six seven four nine dash three 516 00:29:10,640 --> 00:29:13,239 Speaker 1: to a B or something like that, because it's the 517 00:29:13,360 --> 00:29:19,840 Speaker 1: identifier that coincides with this list potentially very long list 518 00:29:20,480 --> 00:29:24,240 Speaker 1: of measurements of that person's face. Yeah. Now, I wonder 519 00:29:24,240 --> 00:29:27,880 Speaker 1: who's going to go back through history to identify people, 520 00:29:28,000 --> 00:29:30,080 Speaker 1: you know, enter it into a database. People like William 521 00:29:30,120 --> 00:29:34,440 Speaker 1: Shakespeare and Thomas Edison and uh, Nicola Tesla and all 522 00:29:34,440 --> 00:29:36,800 Speaker 1: these other people, so that it just auto tags everything 523 00:29:36,840 --> 00:29:40,040 Speaker 1: on the internet. I'm just wondering how many how many 524 00:29:40,080 --> 00:29:42,440 Speaker 1: cameras are going to go out there and mistakenly identify 525 00:29:42,520 --> 00:29:47,600 Speaker 1: people's fathers as Kenny Rogers. You know that website, right, 526 00:29:48,120 --> 00:29:51,680 Speaker 1: My dad looks like Kenny Rogers. Yeah, so I can't 527 00:29:51,680 --> 00:29:54,160 Speaker 1: believe you brought that into this. Okay, just say it's 528 00:29:54,840 --> 00:29:59,120 Speaker 1: apparently a common face to have when facial recognition technology 529 00:29:59,200 --> 00:30:02,400 Speaker 1: is is everywhere, we will also believe that Kenney Rogers 530 00:30:02,400 --> 00:30:06,120 Speaker 1: is omnipresent. You know, you gotta no one to walk away, 531 00:30:06,320 --> 00:30:09,120 Speaker 1: and I think now is the time to run. Yes, yes, 532 00:30:09,480 --> 00:30:12,680 Speaker 1: don't be a gambler. Alright, guys, that wraps up this 533 00:30:12,720 --> 00:30:16,800 Speaker 1: discussion about facial recognition technology. I hope that you enjoyed it. 534 00:30:16,840 --> 00:30:18,520 Speaker 1: We do have a great article on the side if 535 00:30:18,520 --> 00:30:20,640 Speaker 1: you want to read more, that goes into more detail 536 00:30:20,680 --> 00:30:23,600 Speaker 1: about the different the different measurements that these cameras have 537 00:30:23,680 --> 00:30:26,440 Speaker 1: to take and the methodologies they use. So if you 538 00:30:26,520 --> 00:30:28,680 Speaker 1: really want to dive into it, I recommend that they 539 00:30:28,680 --> 00:30:32,000 Speaker 1: also have some really helpful illustrations in there, And if 540 00:30:32,040 --> 00:30:35,360 Speaker 1: you want to send us any questions or comments, or 541 00:30:35,400 --> 00:30:39,160 Speaker 1: you just wanna be our buddies. You can check us 542 00:30:39,160 --> 00:30:42,360 Speaker 1: out on Facebook and Twitter. Our handle at both is 543 00:30:42,600 --> 00:30:46,840 Speaker 1: text Stuff h s W. Or you can send us 544 00:30:46,880 --> 00:30:50,800 Speaker 1: an email or email addresses text stuff at how stuff 545 00:30:50,840 --> 00:30:53,040 Speaker 1: works dot com. Chris and I will talk to you 546 00:30:53,040 --> 00:30:59,600 Speaker 1: again really soon. If you're a tech stuff and be 547 00:30:59,680 --> 00:31:02,920 Speaker 1: sure to check us out on Twitter tech stuff hs 548 00:31:03,120 --> 00:31:05,880 Speaker 1: WSR handle, and you can also find us on Facebook 549 00:31:05,880 --> 00:31:08,880 Speaker 1: at Facebook dot com slash tech stuff h s W. 550 00:31:10,560 --> 00:31:13,080 Speaker 1: For more on this and thousands of other topics, visit 551 00:31:13,160 --> 00:31:15,560 Speaker 1: how stuff works dot com and be sure to check 552 00:31:15,560 --> 00:31:17,600 Speaker 1: out the new tech stuff blog now on the house 553 00:31:17,640 --> 00:31:25,080 Speaker 1: stuff Works homepage, brought to you by the reinvented two 554 00:31:25,120 --> 00:31:27,640 Speaker 1: thousand twelve camera. It's ready, are you