1 00:00:00,01 --> 00:00:06,02 (upbeat music) 2 00:00:06,02 --> 00:00:09,00 - [Speaker] I have the solution for the challenge. 3 00:00:09,00 --> 00:00:12,00 I'm curious to see what did you come up with? 4 00:00:12,00 --> 00:00:13,07 What does your list look like? 5 00:00:13,07 --> 00:00:18,07 So it is important to see an image holistically 6 00:00:18,07 --> 00:00:21,03 for all the information it's capturing. 7 00:00:21,03 --> 00:00:23,05 So I had two questions for you in the challenge. 8 00:00:23,05 --> 00:00:26,06 One is to see what information is it collecting 9 00:00:26,06 --> 00:00:30,03 that is privacy-preserving and privacy-invading. 10 00:00:30,03 --> 00:00:34,06 And other question is what information is it collecting 11 00:00:34,06 --> 00:00:37,01 that you should not be sharing in the cloud 12 00:00:37,01 --> 00:00:40,06 and sending around and might be against compliance laws 13 00:00:40,06 --> 00:00:41,09 related to privacy. 14 00:00:41,09 --> 00:00:44,08 Okay, so let us look at the picture. 15 00:00:44,08 --> 00:00:47,05 I'm going to open the same picture. 16 00:00:47,05 --> 00:00:50,02 So it's loaded the picture here 17 00:00:50,02 --> 00:00:53,04 and it is doing the object detection. 18 00:00:53,04 --> 00:00:54,05 That's pretty quick. 19 00:00:54,05 --> 00:00:58,09 So it is recognizing five objects, 20 00:00:58,09 --> 00:01:02,06 all people from the bounding box. 21 00:01:02,06 --> 00:01:06,03 And in this case, Azure is not saying that "Hey, 22 00:01:06,03 --> 00:01:10,09 I see a helmet or I see a board 23 00:01:10,09 --> 00:01:14,03 or lights or anything else." 24 00:01:14,03 --> 00:01:18,05 It seems to actually focus only on the six people. 25 00:01:18,05 --> 00:01:23,08 And so it's easy to think we have a HJI, 26 00:01:23,08 --> 00:01:26,08 we have a camera that's running this HJI, 27 00:01:26,08 --> 00:01:31,01 it's recognizing five, six people. 28 00:01:31,01 --> 00:01:33,00 And that's what this image is about. 29 00:01:33,00 --> 00:01:35,08 This is where you take a step back 30 00:01:35,08 --> 00:01:40,01 and you say, "This is what I want for my use case." 31 00:01:40,01 --> 00:01:43,02 But what is being collected? 32 00:01:43,02 --> 00:01:46,05 What other information is in this image 33 00:01:46,05 --> 00:01:48,07 that the AI is collecting? 34 00:01:48,07 --> 00:01:51,00 So think about this. 35 00:01:51,00 --> 00:01:53,01 AI is not a static thing. 36 00:01:53,01 --> 00:01:55,05 AI is constantly learning. 37 00:01:55,05 --> 00:01:58,00 So today, you feed this information, 38 00:01:58,00 --> 00:02:00,08 especially computer vision is a very hot area 39 00:02:00,08 --> 00:02:02,04 of AI development 40 00:02:02,04 --> 00:02:07,00 where vision algorithms are developing by the day. 41 00:02:07,00 --> 00:02:11,03 And so what we are looking at is what it is seeing today. 42 00:02:11,03 --> 00:02:14,01 The same model could be updated 43 00:02:14,01 --> 00:02:16,04 to see more things in the future. 44 00:02:16,04 --> 00:02:21,08 The same data that is stored could carry more insights 45 00:02:21,08 --> 00:02:24,07 than it is actually showing outwardly. 46 00:02:24,07 --> 00:02:27,05 So the things I'm seeing is I'm definitely seeing 47 00:02:27,05 --> 00:02:31,00 the board and in the board, I'm seeing a lot 48 00:02:31,00 --> 00:02:32,08 of different stickies and notes. 49 00:02:32,08 --> 00:02:36,01 So it is possible that there is sensitive information 50 00:02:36,01 --> 00:02:38,09 that should not leave this particular building 51 00:02:38,09 --> 00:02:40,08 and should not go out into the cloud. 52 00:02:40,08 --> 00:02:43,07 There's also personal information about people, 53 00:02:43,07 --> 00:02:47,01 about the outfits they're wearing. 54 00:02:47,01 --> 00:02:50,08 Two people in here seem to wear suits. 55 00:02:50,08 --> 00:02:54,02 The others I think are in some kind of T-shirt. 56 00:02:54,02 --> 00:02:55,09 It looks like a uniform. 57 00:02:55,09 --> 00:02:58,08 I don't know if there is some kind of class divide 58 00:02:58,08 --> 00:02:59,08 that maybe their managers 59 00:02:59,08 --> 00:03:04,06 or different levels of work that they are doing. 60 00:03:04,06 --> 00:03:07,07 Some people are wearing helmets, 61 00:03:07,07 --> 00:03:09,08 they're all not wearing helmet in their head, 62 00:03:09,08 --> 00:03:13,01 so it is showing their hair and glasses. 63 00:03:13,01 --> 00:03:15,05 So it's giving a lot more information 64 00:03:15,05 --> 00:03:17,09 and then it is actually giving their faces. 65 00:03:17,09 --> 00:03:22,01 But GDPR we do not save person's face 66 00:03:22,01 --> 00:03:23,07 because you can do facial recognition 67 00:03:23,07 --> 00:03:27,01 and these days, it's becoming a digital identifier 68 00:03:27,01 --> 00:03:30,02 to get you into the airport, to get you into buildings. 69 00:03:30,02 --> 00:03:32,04 And so it is important 70 00:03:32,04 --> 00:03:34,08 that this information doesn't get stored 71 00:03:34,08 --> 00:03:38,01 and passed around in the cloud for the privacy 72 00:03:38,01 --> 00:03:39,08 of these people and also compliance 73 00:03:39,08 --> 00:03:41,08 of the company, which is doing that. 74 00:03:41,08 --> 00:03:45,01 I'm also curious about these little rectangle things 75 00:03:45,01 --> 00:03:48,02 behind this person below the board. 76 00:03:48,02 --> 00:03:49,07 What are those? 77 00:03:49,07 --> 00:03:53,03 So there might be more things than we are actually seeing. 78 00:03:53,03 --> 00:03:54,08 And then if you take a step back 79 00:03:54,08 --> 00:03:56,09 and say it's not just visually what I'm seeing, 80 00:03:56,09 --> 00:04:00,01 the timestamp of this image of when it was captured 81 00:04:00,01 --> 00:04:03,01 and the lighting is actually telling me 82 00:04:03,01 --> 00:04:05,00 it's the night shift. 83 00:04:05,00 --> 00:04:07,04 Are these all night shift people? 84 00:04:07,04 --> 00:04:10,01 There must be something blurred in the background. 85 00:04:10,01 --> 00:04:11,02 We could zoom in maybe 86 00:04:11,02 --> 00:04:13,04 and it can show something else 87 00:04:13,04 --> 00:04:15,05 that should not be in this picture. 88 00:04:15,05 --> 00:04:17,00 So that kind of thing. 89 00:04:17,00 --> 00:04:18,03 So you capture the people, 90 00:04:18,03 --> 00:04:20,04 you capture the information about these people. 91 00:04:20,04 --> 00:04:22,06 It actually shows their eye color. 92 00:04:22,06 --> 00:04:27,01 Now I see this, it shows little bit more profile information 93 00:04:27,01 --> 00:04:30,01 that definitely should not be passed around 94 00:04:30,01 --> 00:04:32,08 and stored beyond identifying, you know, 95 00:04:32,08 --> 00:04:34,08 six people are standing here. 96 00:04:34,08 --> 00:04:36,09 So I hope you were able to capture this. 97 00:04:36,09 --> 00:04:38,02 I hope this was insightful. 98 00:04:38,02 --> 00:04:41,07 I want you to not think of this as, "Okay, I did a class, 99 00:04:41,07 --> 00:04:44,07 I did an exercise and I learned about data privacy. 100 00:04:44,07 --> 00:04:47,04 I learned about Edge AI and building cool stuff." 101 00:04:47,04 --> 00:04:49,09 I want you to go past that 102 00:04:49,09 --> 00:04:52,09 to develop a way of thinking 103 00:04:52,09 --> 00:04:56,08 when you look at Edge ai, be it audio sound 104 00:04:56,08 --> 00:04:58,09 and voice AI or images 105 00:04:58,09 --> 00:05:03,02 and videos, what AI is seeing is our environment. 106 00:05:03,02 --> 00:05:06,08 AI is trying to make decisions on her behalf. 107 00:05:06,08 --> 00:05:10,00 So it is our responsibility 108 00:05:10,00 --> 00:05:12,04 to actually be very responsible about 109 00:05:12,04 --> 00:05:15,08 what we are leaving in this data, what we are allowing 110 00:05:15,08 --> 00:05:20,04 for AI to capture and share and process and analyze. 111 00:05:20,04 --> 00:05:22,09 And whether it is kept in the cloud 112 00:05:22,09 --> 00:05:25,00 or whether it is kept in the edge 113 00:05:25,00 --> 00:05:28,03 or it is going to be shared across businesses, 114 00:05:28,03 --> 00:05:29,07 across partnerships. 115 00:05:29,07 --> 00:05:33,01 So think about that and I want you to develop this thinking. 116 00:05:33,01 --> 00:05:34,04 When you look at your environment 117 00:05:34,04 --> 00:05:36,08 and you're looking at use cases, 118 00:05:36,08 --> 00:05:41,00 not just think about the data that the AI is expected 119 00:05:41,00 --> 00:05:44,07 to operate on, but all that the AI can glean. 120 00:05:44,07 --> 00:05:46,03 And it's a very valuable skill 121 00:05:46,03 --> 00:05:49,09 because it's definitely good for ethics and data privacy, 122 00:05:49,09 --> 00:05:55,01 but also a valuable business skill where you can come up 123 00:05:55,01 --> 00:05:58,01 with different ways by which you can create from the data 124 00:05:58,01 --> 00:05:59,07 and then you stop yourself 125 00:05:59,07 --> 00:06:01,05 and say, "Is this good value 126 00:06:01,05 --> 00:06:05,00 that these people in the picture are going to allow me 127 00:06:05,00 --> 00:06:07,03 and gimme permission to go 128 00:06:07,03 --> 00:06:09,04 and create a business on top of it 129 00:06:09,04 --> 00:06:11,04 or go make money out of it? 130 00:06:11,04 --> 00:06:12,05 Or is this something 131 00:06:12,05 --> 00:06:14,04 that is unethical and I should not be doing?" 132 00:06:14,04 --> 00:06:19,02 So it's a way of thinking about value creation from data. 133 00:06:19,02 --> 00:06:20,08 That's a lifelong lesson. 134 00:06:20,08 --> 00:06:23,00 So I'm curious to see how you're going 135 00:06:23,00 --> 00:06:24,08 to apply this entire course learning. 136 00:06:24,08 --> 00:06:28,08 So beyond just this course, if you scroll down, 137 00:06:28,08 --> 00:06:30,07 you can see a link to Sudha Live 138 00:06:30,07 --> 00:06:32,06 that is my monthly live stream 139 00:06:32,06 --> 00:06:34,05 where I meet my students and answer questions. 140 00:06:34,05 --> 00:06:36,05 I would love for you to participate 141 00:06:36,05 --> 00:06:38,03 and come tell us what kind 142 00:06:38,03 --> 00:06:40,05 of things are you building, what have you learned? 143 00:06:40,05 --> 00:06:43,05 What are you applying, what challenges are you facing? 144 00:06:43,05 --> 00:06:45,05 Because I like to help my students. 145 00:06:45,05 --> 00:06:49,05 And outside of that, I think there is also inside LinkedIn, 146 00:06:49,05 --> 00:06:51,02 there is a business school of AI, 147 00:06:51,02 --> 00:06:53,01 which is my learner community. 148 00:06:53,01 --> 00:06:55,06 You can follow that and come and ask questions 149 00:06:55,06 --> 00:06:58,07 and share the projects that you're posting. 150 00:06:58,07 --> 00:07:01,02 We like to showcase our students and their work. 151 00:07:01,02 --> 00:07:03,00 So remember, I'm here to help.