1 00:00:03,320 --> 00:00:05,080 Speaker 1: If you are hearing the sound of my voice, then 2 00:00:05,080 --> 00:00:08,840 Speaker 1: you are not actually hearing my voice at all. What 3 00:00:08,880 --> 00:00:10,520 Speaker 1: I mean is that the voice you are hearing is 4 00:00:10,560 --> 00:00:13,160 Speaker 1: actually an assistive text of voice cloning tool that my 5 00:00:13,240 --> 00:00:17,760 Speaker 1: company created, and it is completely powered by AI. There 6 00:00:17,760 --> 00:00:20,120 Speaker 1: are many different types of AI tools to help people 7 00:00:20,120 --> 00:00:23,600 Speaker 1: who are differently abled. For me, it has restored my voice, 8 00:00:23,600 --> 00:00:26,280 Speaker 1: but there is so much more it can do. I'm 9 00:00:26,320 --> 00:00:31,880 Speaker 1: excited to see how it grows. Hey, there, I'm grain 10 00:00:31,960 --> 00:00:35,920 Speaker 1: class and this is technically speaking an Intel podcast. The 11 00:00:35,960 --> 00:00:39,199 Speaker 1: show is dedicated to highlighting the way technology is revolutionizing 12 00:00:39,320 --> 00:00:43,120 Speaker 1: the way we live, work and move. In every episode, 13 00:00:43,240 --> 00:00:46,400 Speaker 1: we'll connect with innovators in areas like artificial intelligence to 14 00:00:46,479 --> 00:00:50,440 Speaker 1: better understand the human centered technology they've developed. As early 15 00:00:50,479 --> 00:00:53,120 Speaker 1: as the discovery of fire and the invention of the wheel, 16 00:00:53,640 --> 00:00:57,280 Speaker 1: technology has always been an innovation to improve people's lives. However, 17 00:00:57,920 --> 00:01:01,760 Speaker 1: sometimes leaders and technology unintentionally exclude those who may deal 18 00:01:01,760 --> 00:01:07,759 Speaker 1: with uncommon issues such as physical immobility, neurodivergence, visual impairments, 19 00:01:07,880 --> 00:01:11,320 Speaker 1: or even old age. While governments usually put systems in 20 00:01:11,360 --> 00:01:13,959 Speaker 1: place to acknowledge and care for these communities, it has 21 00:01:14,000 --> 00:01:16,840 Speaker 1: been the role of technology to create advancements necessary for 22 00:01:16,880 --> 00:01:20,080 Speaker 1: those dealing with disabilities to thrive just as much as 23 00:01:20,080 --> 00:01:24,280 Speaker 1: they abled counterparts. With the revelation of artificial intelligence, there 24 00:01:24,280 --> 00:01:27,400 Speaker 1: are many new advancements that are providing accessibility to these 25 00:01:27,440 --> 00:01:30,920 Speaker 1: communities in the ways we never thought possible until now. 26 00:01:31,480 --> 00:01:33,399 Speaker 1: And I have two experts with me who are leading 27 00:01:33,440 --> 00:01:38,319 Speaker 1: the charge to more accessible future. Lama and Ackmann is 28 00:01:38,360 --> 00:01:42,440 Speaker 1: a visionary leader at the intersection of technology and human experience. 29 00:01:43,080 --> 00:01:46,720 Speaker 1: With a distinguished career spanning academia and industry, Lama has 30 00:01:46,800 --> 00:01:50,200 Speaker 1: consistently pushed the boundaries of technology to enhance our daily 31 00:01:50,200 --> 00:01:53,840 Speaker 1: lives and redefine the way we interact with computers. Her 32 00:01:53,840 --> 00:01:56,840 Speaker 1: innovative work has not only advanced the field of AI, 33 00:01:57,320 --> 00:01:59,600 Speaker 1: but has also paved the way for more intuitive and 34 00:02:00,440 --> 00:02:04,200 Speaker 1: human machine interfaces. As an Intel Fellow and director of 35 00:02:04,240 --> 00:02:07,240 Speaker 1: its Human and AI Systems Research Lab, she leads the 36 00:02:07,280 --> 00:02:10,880 Speaker 1: team defining and executing the research for contextually aware and 37 00:02:10,919 --> 00:02:15,920 Speaker 1: personalized computing, developing sensing systems, algorithms, and applications to make 38 00:02:15,960 --> 00:02:16,840 Speaker 1: it all possible. 39 00:02:17,160 --> 00:02:19,880 Speaker 2: Welcome Lama, thank you very nice to be here. 40 00:02:20,320 --> 00:02:24,040 Speaker 1: We're also joined by Jagadesh Mahendram, a visionary entrepreneur and 41 00:02:24,080 --> 00:02:27,240 Speaker 1: tech innovator who has made significant contributions to the fields 42 00:02:27,240 --> 00:02:32,080 Speaker 1: of artificial intelligence, renewable energy, and sustainable development. With a 43 00:02:32,120 --> 00:02:36,080 Speaker 1: relentless passion for cutting edge technology to address global challenges, 44 00:02:36,560 --> 00:02:39,520 Speaker 1: Jakodesh has emerged as a driving force in shaping a 45 00:02:39,560 --> 00:02:44,040 Speaker 1: more sustainable and inconnected world. Most recently, he joined Camera 46 00:02:44,520 --> 00:02:47,040 Speaker 1: LLC with his founding partners and a team of visually 47 00:02:47,080 --> 00:02:51,000 Speaker 1: impaired volunteers. He uses AI to develop solutions and assistance 48 00:02:51,040 --> 00:02:55,760 Speaker 1: for those dealing with site lost and low visibility. Welcome, Jaggedesh, 49 00:02:56,040 --> 00:03:02,240 Speaker 1: Thank you very much. It's so honor to be here. Okay, 50 00:03:02,280 --> 00:03:05,520 Speaker 1: I've just start with Lama. Lama, how did you get 51 00:03:05,520 --> 00:03:07,639 Speaker 1: your start in tech and AI? 52 00:03:08,480 --> 00:03:12,480 Speaker 2: And I would say that I've been in love with 53 00:03:12,600 --> 00:03:17,120 Speaker 2: tech probably since I was like two years old, you know. 54 00:03:17,160 --> 00:03:19,640 Speaker 2: I've always been into kind of like the latest and 55 00:03:19,680 --> 00:03:24,400 Speaker 2: the greatest technology growing up. But then after I graduated 56 00:03:24,720 --> 00:03:29,639 Speaker 2: from UW Madison, I actually joined Intel out of college 57 00:03:29,919 --> 00:03:32,240 Speaker 2: and then I worked there for a while. I went 58 00:03:32,280 --> 00:03:35,680 Speaker 2: out and did a few startups and then came back 59 00:03:35,720 --> 00:03:40,920 Speaker 2: to Intel specifically focused on that intersection of sensing and 60 00:03:41,080 --> 00:03:45,240 Speaker 2: understanding the world through that to create really compelling technology. 61 00:03:45,320 --> 00:03:47,240 Speaker 2: So that's been kind of like almost like a very 62 00:03:47,320 --> 00:03:50,080 Speaker 2: long career progression that brought me back to what I 63 00:03:50,160 --> 00:03:51,000 Speaker 2: was excited about. 64 00:03:51,560 --> 00:03:54,760 Speaker 1: Excellent, and then in terms of the AI component, how 65 00:03:54,760 --> 00:03:56,360 Speaker 1: did you start to get involved in that? 66 00:03:57,040 --> 00:03:59,600 Speaker 2: Yeah, so early on when I went back to Intel, 67 00:03:59,640 --> 00:04:02,680 Speaker 2: actually in two thousand and three, I started to look at, 68 00:04:02,800 --> 00:04:04,640 Speaker 2: you know, how do we make sense out of the 69 00:04:04,680 --> 00:04:08,480 Speaker 2: world around us? So to be able to understand a 70 00:04:08,480 --> 00:04:11,320 Speaker 2: lot of that sensor data that we were processing, whether 71 00:04:11,360 --> 00:04:15,040 Speaker 2: it's vision or audio or tex or whatever, right, that 72 00:04:15,200 --> 00:04:18,240 Speaker 2: really required work and AI to actually make sense out 73 00:04:18,240 --> 00:04:20,200 Speaker 2: of that data. So that's where it kind of started 74 00:04:20,240 --> 00:04:23,200 Speaker 2: around two thousand and four, and then since then it's 75 00:04:23,279 --> 00:04:28,000 Speaker 2: kind of looking at different ways of intersecting AI and 76 00:04:28,560 --> 00:04:33,840 Speaker 2: HCI to actually bring about really compelling experiences for users 77 00:04:33,920 --> 00:04:36,839 Speaker 2: and helping them perform all sorts of things in their lives. 78 00:04:37,240 --> 00:04:42,039 Speaker 1: Excellent and Jackets, how did you get your start in technology? 79 00:04:43,120 --> 00:04:45,800 Speaker 3: I have a very different story here. I was not 80 00:04:45,880 --> 00:04:49,440 Speaker 3: interested in technology or all. Actually wanted to become a doctor, 81 00:04:49,760 --> 00:04:52,760 Speaker 3: but you know, it's very competitive in India, so I 82 00:04:52,760 --> 00:04:55,279 Speaker 3: didn't really get good ranking, so I couldn't join the 83 00:04:55,320 --> 00:04:58,559 Speaker 3: college that I wanted to join, and the second option 84 00:04:58,720 --> 00:05:03,000 Speaker 3: was engineering, and I chose to do the computer science. 85 00:05:03,320 --> 00:05:05,520 Speaker 3: I think the very best turn out to me is 86 00:05:05,600 --> 00:05:10,920 Speaker 3: good that we're enjoying artificial intelligence and more than doctor. 87 00:05:11,040 --> 00:05:13,520 Speaker 3: I think I'm a bettery engineer than a doctor. 88 00:05:16,000 --> 00:05:18,800 Speaker 1: Oh that kind of makes sense with your I guess 89 00:05:18,839 --> 00:05:21,559 Speaker 1: love of medicine and the type of projects that you've 90 00:05:21,839 --> 00:05:23,680 Speaker 1: come up with. So we'll talk about that in a 91 00:05:23,680 --> 00:05:26,279 Speaker 1: little bit. But one thing I'll go back to Lama 92 00:05:26,560 --> 00:05:30,440 Speaker 1: in terms of AI improving the human experience you've mentioned HI. 93 00:05:30,520 --> 00:05:32,200 Speaker 1: First of all, if you can define for the audience 94 00:05:32,200 --> 00:05:37,680 Speaker 1: what HDI is and also how do you address solutions 95 00:05:38,120 --> 00:05:40,279 Speaker 1: and maybe you could educate me around what's the difference 96 00:05:40,279 --> 00:05:43,880 Speaker 1: between accessibility and accommodation when you're designing a system. 97 00:05:44,720 --> 00:05:48,400 Speaker 2: Yeah, so, first of all, HDI is a human computer interaction, 98 00:05:48,880 --> 00:05:53,120 Speaker 2: So it's really trying to understand how would people directly 99 00:05:53,160 --> 00:05:57,160 Speaker 2: interact with technology. And sometimes that technology is something that's 100 00:05:57,200 --> 00:06:01,719 Speaker 2: actually physical that you're picking something on a computer or whatever. 101 00:06:01,800 --> 00:06:03,359 Speaker 2: But a lot of times, you know, some of the 102 00:06:03,400 --> 00:06:06,560 Speaker 2: work that we really work on is embedding it into 103 00:06:06,600 --> 00:06:11,120 Speaker 2: the environment so that it almost becomes like invisible. And 104 00:06:11,160 --> 00:06:13,120 Speaker 2: that's kind of one of the most interesting things is 105 00:06:13,120 --> 00:06:17,640 Speaker 2: like to really architect for interactions of things that are invisible. Honestly, 106 00:06:17,640 --> 00:06:19,960 Speaker 2: if you think about any technology that you're developing, you 107 00:06:20,040 --> 00:06:23,320 Speaker 2: have to think about how you're making it accessible, how 108 00:06:23,320 --> 00:06:26,400 Speaker 2: the interfaces are accessible, how different people with different types 109 00:06:26,440 --> 00:06:30,919 Speaker 2: of disabilities and abilities can actually interact with your technology 110 00:06:31,400 --> 00:06:36,520 Speaker 2: throughout like the development cycle. In some sense, part of 111 00:06:36,560 --> 00:06:41,200 Speaker 2: what I've really been focused on is creating technologies for 112 00:06:41,320 --> 00:06:46,479 Speaker 2: people who are severely disabled, where you really need very 113 00:06:46,480 --> 00:06:50,640 Speaker 2: different ways of interacting with the technology to enable that 114 00:06:50,720 --> 00:06:54,200 Speaker 2: to happen. Really, that focus specifically on the work with 115 00:06:54,240 --> 00:06:57,800 Speaker 2: ACAT and the workforce even Hawking, has really been about 116 00:06:57,839 --> 00:07:01,280 Speaker 2: how do you get around these on strengths to enable 117 00:07:01,600 --> 00:07:05,200 Speaker 2: people to access the technologies just like all of us. 118 00:07:06,800 --> 00:07:11,080 Speaker 1: Lama mentions Intel's ACAT, which stands for Assistive Context Aware talkit. 119 00:07:11,720 --> 00:07:14,880 Speaker 1: This technology was key in enabling Stephen Hawking's ability to 120 00:07:14,920 --> 00:07:19,440 Speaker 1: continue to communicate and inspire people around the world. Listening 121 00:07:19,520 --> 00:07:23,200 Speaker 1: to Lama speak about the human computer interaction processes, she 122 00:07:23,280 --> 00:07:27,000 Speaker 1: sounds less like a tech person and more like an anthropologist. 123 00:07:27,760 --> 00:07:30,240 Speaker 1: We often think of data and algorithms as being this 124 00:07:30,480 --> 00:07:34,120 Speaker 1: cold and personal assessment of people. But Lama has such 125 00:07:34,120 --> 00:07:37,360 Speaker 1: a passion for her programs, it makes me wonder just 126 00:07:37,400 --> 00:07:40,520 Speaker 1: how impactful that passion is to the way AI tools 127 00:07:40,680 --> 00:07:43,920 Speaker 1: interpret how to assist us. While she has spent so 128 00:07:44,000 --> 00:07:47,040 Speaker 1: much time learning how to program and manage computers, it 129 00:07:47,120 --> 00:07:51,040 Speaker 1: seems her real passion is in trying to understand humanity. 130 00:07:52,520 --> 00:07:57,280 Speaker 2: My passion has really been focused on how do we 131 00:07:57,360 --> 00:08:02,400 Speaker 2: bring more equity with technology. The work towards specifically extreme 132 00:08:02,440 --> 00:08:07,280 Speaker 2: disability really came about from my interaction with Professor Hawking. 133 00:08:07,800 --> 00:08:09,680 Speaker 2: So before that, a lot of the work that I 134 00:08:09,720 --> 00:08:13,080 Speaker 2: had focused on in terms of accessibility is really bridging 135 00:08:13,840 --> 00:08:17,480 Speaker 2: where people's needs were as they're doing different aspects of 136 00:08:17,520 --> 00:08:20,320 Speaker 2: their lives. Right you're driving, for example, how can that 137 00:08:20,480 --> 00:08:23,920 Speaker 2: be contextually aware so it can help support you without 138 00:08:24,200 --> 00:08:27,440 Speaker 2: assuming that you have all of your abilities there. But 139 00:08:27,600 --> 00:08:31,600 Speaker 2: once I started working with Professor Stephen Hawking, it became 140 00:08:31,760 --> 00:08:36,319 Speaker 2: very obvious to me that to bridge that extreme disability 141 00:08:36,440 --> 00:08:39,880 Speaker 2: you really have to think very differently about how technology 142 00:08:39,920 --> 00:08:42,280 Speaker 2: comes in. And that's what really got me excited about 143 00:08:42,320 --> 00:08:42,760 Speaker 2: that work. 144 00:08:43,240 --> 00:08:46,320 Speaker 1: Okay, and in terms of the involvement you had with 145 00:08:46,920 --> 00:08:50,559 Speaker 1: Professor Hawking's technology to help him interact with the world, 146 00:08:50,920 --> 00:08:53,000 Speaker 1: What were the areas that you looked at? 147 00:08:53,360 --> 00:08:56,000 Speaker 2: The lab I lead is actually a multi disciplinary lab, right, 148 00:08:56,040 --> 00:08:59,719 Speaker 2: so we bring social science, design, and AI together. So 149 00:09:00,240 --> 00:09:03,280 Speaker 2: the first place you start is we needed to understand 150 00:09:03,360 --> 00:09:06,559 Speaker 2: how Stephen interacts with the world, what he is trying 151 00:09:06,559 --> 00:09:10,160 Speaker 2: to accomplish, and where are his bottomnecks in terms of 152 00:09:10,200 --> 00:09:12,440 Speaker 2: being able to do that with existing technology that he 153 00:09:12,559 --> 00:09:14,760 Speaker 2: was using. So there was a lot of observation to 154 00:09:14,800 --> 00:09:18,280 Speaker 2: try to understand how do we define the problem and 155 00:09:18,400 --> 00:09:21,160 Speaker 2: from there for people who are not aware of this, right, 156 00:09:21,280 --> 00:09:25,080 Speaker 2: Professor Hawking really didn't have an ability to speak, and 157 00:09:25,160 --> 00:09:27,800 Speaker 2: he didn't really have an ability to move, so he 158 00:09:27,840 --> 00:09:31,160 Speaker 2: couldn't really utilize many of the technologies that are available. 159 00:09:31,559 --> 00:09:35,200 Speaker 2: You couldn't do, for example, ASR where he could speak 160 00:09:35,240 --> 00:09:38,240 Speaker 2: and then the computer could be controlled by speech, Nor 161 00:09:38,280 --> 00:09:41,240 Speaker 2: could he type because he had no control over his hands. 162 00:09:41,720 --> 00:09:44,920 Speaker 2: So then we started to basically look at if we 163 00:09:45,080 --> 00:09:48,600 Speaker 2: really had a very very tiny signal, and in this 164 00:09:48,640 --> 00:09:51,480 Speaker 2: specific case for Professor Hawking, it was actually his ability 165 00:09:51,520 --> 00:09:55,160 Speaker 2: to move his cheek. Can we get access to that 166 00:09:55,400 --> 00:10:00,160 Speaker 2: one signal and then turn that into a complete access 167 00:10:00,200 --> 00:10:02,719 Speaker 2: for his whole machine? And then we went onto that 168 00:10:02,840 --> 00:10:07,959 Speaker 2: path of essentially building a software platform and a sensing 169 00:10:08,040 --> 00:10:11,319 Speaker 2: subsystem that allowed for that to happen. All he can 170 00:10:11,360 --> 00:10:13,839 Speaker 2: do is confirm something with the movement of a cheek, 171 00:10:13,880 --> 00:10:17,360 Speaker 2: and now he can type, he can email, he can 172 00:10:17,800 --> 00:10:20,000 Speaker 2: serve the web, he can give lectures, he could do all. 173 00:10:19,960 --> 00:10:22,719 Speaker 1: Of that in What year was that that you were 174 00:10:22,720 --> 00:10:23,160 Speaker 1: working on? 175 00:10:24,080 --> 00:10:27,560 Speaker 2: So we started our interaction with Stephen and twenty eleven 176 00:10:28,200 --> 00:10:31,240 Speaker 2: and it kind of lasted throughout his life until he 177 00:10:31,280 --> 00:10:35,320 Speaker 2: passed away, which was twenty eighteen. So we after a 178 00:10:35,320 --> 00:10:37,560 Speaker 2: couple of years we were able to put together a 179 00:10:37,600 --> 00:10:39,720 Speaker 2: system that he could use that he could switch to, 180 00:10:39,880 --> 00:10:42,640 Speaker 2: and then over the years we just continued to enhance 181 00:10:42,679 --> 00:10:45,160 Speaker 2: it and add more capabilities. We open sourced it so 182 00:10:45,200 --> 00:10:46,560 Speaker 2: that we could take it into the world. 183 00:10:46,600 --> 00:10:48,800 Speaker 1: And yeah, that was my next question in terms of 184 00:10:48,840 --> 00:10:52,120 Speaker 1: the technology that was developed. Have you seen it applied 185 00:10:52,559 --> 00:10:54,120 Speaker 1: more broadly to others. 186 00:10:54,800 --> 00:10:58,240 Speaker 2: Yeah, And initially we were hoping that we could find 187 00:10:58,679 --> 00:11:02,280 Speaker 2: some technology out there we could take and modify slightly 188 00:11:02,360 --> 00:11:04,800 Speaker 2: so that he could use it. And then after being 189 00:11:04,840 --> 00:11:08,120 Speaker 2: proven wrong, we then went onto this path to go 190 00:11:08,160 --> 00:11:10,880 Speaker 2: and develop something from scratch. But from the get go 191 00:11:11,120 --> 00:11:14,200 Speaker 2: our goal was to develop it so that it could 192 00:11:14,280 --> 00:11:17,760 Speaker 2: support a lot of different users and be a platform 193 00:11:17,840 --> 00:11:21,079 Speaker 2: for developers to build on top of, because we realized 194 00:11:21,080 --> 00:11:24,319 Speaker 2: that there was that gap in what existed out there 195 00:11:24,400 --> 00:11:27,920 Speaker 2: in the open world. And Stephen was a huge contributor 196 00:11:27,960 --> 00:11:30,600 Speaker 2: to this project, right he he you know, he was 197 00:11:31,280 --> 00:11:33,800 Speaker 2: a designer, he was a validator, he was you know, 198 00:11:33,840 --> 00:11:36,080 Speaker 2: he gave a lot of his insights. So throughout all 199 00:11:36,080 --> 00:11:39,559 Speaker 2: of this he was really focused on ensuring that that 200 00:11:39,640 --> 00:11:43,240 Speaker 2: actually went to open source because you know, people reached 201 00:11:43,280 --> 00:11:45,720 Speaker 2: out to him all the time because he was, you know, 202 00:11:45,800 --> 00:11:49,320 Speaker 2: an own figure with that extreme disability, and everybody was 203 00:11:49,360 --> 00:11:52,040 Speaker 2: asking him, like, what technology is available to us to 204 00:11:52,120 --> 00:11:54,959 Speaker 2: actually use. So he's been like really focused quite a 205 00:11:55,000 --> 00:11:57,600 Speaker 2: bit on making that available to the world. 206 00:11:59,120 --> 00:12:01,960 Speaker 1: You can really sense how dedicated Lama is to helping 207 00:12:01,960 --> 00:12:06,280 Speaker 1: those with disabilities communicate with others. However, talking is just 208 00:12:06,480 --> 00:12:09,880 Speaker 1: one way we communicate, and moreover, there are a combination 209 00:12:09,960 --> 00:12:13,000 Speaker 1: of ways that we engage and interact with our environments. 210 00:12:14,160 --> 00:12:16,320 Speaker 1: As a way to help people that may struggle with 211 00:12:16,360 --> 00:12:21,600 Speaker 1: another sense is our other guest Jagged Dish originally designing 212 00:12:21,600 --> 00:12:24,240 Speaker 1: a backpack that uses AI to help guide the blind. 213 00:12:24,920 --> 00:12:28,280 Speaker 1: His project expanded into really dissecting what it means to 214 00:12:28,320 --> 00:12:32,040 Speaker 1: be visually impaired. Lama and Jagged Dish both have different 215 00:12:32,040 --> 00:12:35,559 Speaker 1: approaches to their missions. The work compliments each other so well. 216 00:12:38,600 --> 00:12:42,440 Speaker 1: Jagsh I'd like to get you into conversation now, in 217 00:12:42,480 --> 00:12:47,280 Speaker 1: particular the AI powered backpack that has been developed by 218 00:12:47,280 --> 00:12:49,880 Speaker 1: yourself and others. Can you just tell me a little 219 00:12:49,920 --> 00:12:52,880 Speaker 1: bit about I guess the genesis of that idea. 220 00:12:53,600 --> 00:12:58,040 Speaker 3: I've always wanted to do something using the technology that 221 00:12:58,120 --> 00:13:00,720 Speaker 3: can help the society in what way or the other. 222 00:13:01,360 --> 00:13:05,360 Speaker 3: And when it came to Masters in twenty thirteen, one 223 00:13:05,400 --> 00:13:07,120 Speaker 3: of the first things that occurred to me was like, 224 00:13:07,160 --> 00:13:10,439 Speaker 3: you know, we should use EI and use a bunch 225 00:13:10,480 --> 00:13:13,720 Speaker 3: of sensors to help the usually impaired see the world, 226 00:13:13,800 --> 00:13:18,200 Speaker 3: like how sighted people see. And one of the primary 227 00:13:18,280 --> 00:13:20,600 Speaker 3: visions that I used to occur to me was when 228 00:13:20,640 --> 00:13:24,640 Speaker 3: somebody is standing in public places like buz stop, there 229 00:13:24,640 --> 00:13:27,160 Speaker 3: should be a solution in such a way that the 230 00:13:27,200 --> 00:13:31,400 Speaker 3: person who is blind should get totally unnoticed. And around 231 00:13:31,440 --> 00:13:34,800 Speaker 3: that time the technology was not as good as how 232 00:13:34,840 --> 00:13:37,960 Speaker 3: it is now. The real inspiration occurred to me when 233 00:13:37,960 --> 00:13:41,000 Speaker 3: I met my friend. The day when I met her, 234 00:13:41,360 --> 00:13:43,840 Speaker 3: she had a black mark in her face, and I 235 00:13:43,920 --> 00:13:46,440 Speaker 3: was like, you know, what happened to your face? And 236 00:13:46,480 --> 00:13:49,640 Speaker 3: she's usually impair and she was saying, as she was 237 00:13:49,679 --> 00:13:52,760 Speaker 3: walking outside in the sidewalk, she ran into a tree 238 00:13:52,760 --> 00:13:55,520 Speaker 3: branch and then that left a mark in her face. 239 00:13:56,360 --> 00:14:00,520 Speaker 3: And that was such an ironical for me because by 240 00:14:00,600 --> 00:14:04,920 Speaker 3: then I was already a perception engineer, teaching robots to 241 00:14:04,960 --> 00:14:07,800 Speaker 3: see things, you know, do complex us. But at the 242 00:14:07,800 --> 00:14:11,040 Speaker 3: same time, there are so many people who cannot see 243 00:14:11,240 --> 00:14:17,520 Speaker 3: right and that sort of spart my desire to work 244 00:14:17,600 --> 00:14:21,080 Speaker 3: on this project sooner than later. Around the same time, 245 00:14:21,480 --> 00:14:24,400 Speaker 3: this competition of Open Sea Special ai I was going on, 246 00:14:24,560 --> 00:14:28,960 Speaker 3: sponsored by Intel, and I submitted this idea and the 247 00:14:29,000 --> 00:14:31,640 Speaker 3: project ended up winning the first price. And this friend 248 00:14:31,640 --> 00:14:34,360 Speaker 3: has been helping you throughout how to develop a system 249 00:14:34,360 --> 00:14:38,680 Speaker 3: that is more user friendly and actually solves important use cases. 250 00:14:39,440 --> 00:14:43,520 Speaker 3: And through this competition received a lot of attention, and 251 00:14:43,600 --> 00:14:45,280 Speaker 3: this is when we started to think, you know, we 252 00:14:45,280 --> 00:14:49,560 Speaker 3: should probably you know, get incorporated and try to create 253 00:14:49,680 --> 00:14:52,880 Speaker 3: a full fledged open source system so that anybody in 254 00:14:52,880 --> 00:14:56,040 Speaker 3: the world can use it and help in improving the 255 00:14:56,040 --> 00:14:59,320 Speaker 3: lives of the visually impaired. Currently, we are supported by 256 00:15:00,280 --> 00:15:04,800 Speaker 3: Intel's irt A program Intel Rise Technological Initiative program, and 257 00:15:05,200 --> 00:15:08,840 Speaker 3: we are receiving an in collaboration with Accenture. Through this partnership, 258 00:15:08,840 --> 00:15:11,760 Speaker 3: you're able to gain a lot of support both on 259 00:15:11,800 --> 00:15:14,400 Speaker 3: the technical and non technical side. And soon we will 260 00:15:14,440 --> 00:15:18,600 Speaker 3: be releasing our improved version of the system, which we 261 00:15:18,640 --> 00:15:20,640 Speaker 3: call as Phoenix in a few months. 262 00:15:21,240 --> 00:15:24,480 Speaker 1: Okay, excellent, looking forward to it. And can you tell 263 00:15:24,520 --> 00:15:26,360 Speaker 1: me what I've seen a little bit of a video 264 00:15:26,400 --> 00:15:28,680 Speaker 1: on it where you've got a backpack. Maybe you could 265 00:15:28,680 --> 00:15:30,640 Speaker 1: just describe some of the main system elements. 266 00:15:31,280 --> 00:15:34,800 Speaker 3: Yeah, the physical system mainly consists of a backpack that 267 00:15:34,920 --> 00:15:38,640 Speaker 3: has Intel Look with a couple of new compute sticks 268 00:15:38,880 --> 00:15:42,040 Speaker 3: and this is the sort of the compute resource. And 269 00:15:42,160 --> 00:15:45,360 Speaker 3: at the front we have a camera. It's Obi camera 270 00:15:45,400 --> 00:15:47,800 Speaker 3: that is is put in the front and connect it 271 00:15:47,800 --> 00:15:52,720 Speaker 3: to the system behind and whatever this sensor collects the data. 272 00:15:53,200 --> 00:15:57,359 Speaker 3: We run some AA processing behind using deep learning techniques 273 00:15:57,800 --> 00:16:01,920 Speaker 3: and the system will infer useful information about the environment 274 00:16:02,000 --> 00:16:05,320 Speaker 3: and update the user such as where are the obstacles 275 00:16:05,680 --> 00:16:08,160 Speaker 3: and what are the common objects seen in the scenario, 276 00:16:08,400 --> 00:16:11,760 Speaker 3: what are the moving objects, what are the traffic conditions? 277 00:16:11,800 --> 00:16:16,400 Speaker 3: And more similar features for communicating there is audio interface 278 00:16:16,680 --> 00:16:21,680 Speaker 3: through bluetooth headphones, and we're also working on a haptic 279 00:16:21,760 --> 00:16:25,640 Speaker 3: band to communicate the same sort of information in form 280 00:16:25,640 --> 00:16:27,680 Speaker 3: of vibrations through tacktail information. 281 00:16:28,520 --> 00:16:31,160 Speaker 1: Lama was just wondering if you had any comments or 282 00:16:31,240 --> 00:16:33,440 Speaker 1: thoughts on this AI backpack. 283 00:16:34,120 --> 00:16:37,240 Speaker 2: I mean, it's a fantastic idea, and you know, if 284 00:16:37,280 --> 00:16:41,640 Speaker 2: you think about what is actually now possible with perception 285 00:16:42,040 --> 00:16:44,400 Speaker 2: and AI, I mean, it's it's just kind of like 286 00:16:44,440 --> 00:16:48,360 Speaker 2: the most natural thing to do to actually empower users 287 00:16:48,400 --> 00:16:52,280 Speaker 2: with such a capability that are vision impaired. I was 288 00:16:52,320 --> 00:16:56,760 Speaker 2: actually also quite intrigued by the haptic aspect of what 289 00:16:56,800 --> 00:16:59,800 Speaker 2: you mentioned, and I think it's something that tends to 290 00:16:59,800 --> 00:17:03,560 Speaker 2: be un they're utilized, but really kind of a natural 291 00:17:03,680 --> 00:17:06,720 Speaker 2: thing for this type of application, especially if you're trying 292 00:17:06,720 --> 00:17:08,959 Speaker 2: to kind of guide somebody in a direction. So I 293 00:17:08,960 --> 00:17:11,280 Speaker 2: was wondering maybe you can say a few words about that. 294 00:17:11,359 --> 00:17:12,760 Speaker 2: I was really intrigued by that. 295 00:17:13,560 --> 00:17:17,840 Speaker 3: Yeah, So if the first prototype contained mainly the audio interface, 296 00:17:18,440 --> 00:17:22,199 Speaker 3: all the information is actually shared via the wireless headphones 297 00:17:22,760 --> 00:17:26,679 Speaker 3: and not all the users prefer that main reason is 298 00:17:26,720 --> 00:17:30,320 Speaker 3: because they usually impair people rely on audio cues when 299 00:17:30,320 --> 00:17:34,800 Speaker 3: they're wearing earphones. We are sort of blocking a lot 300 00:17:34,840 --> 00:17:37,840 Speaker 3: of environmental cues, which is why we wanted to introduce 301 00:17:38,000 --> 00:17:43,560 Speaker 3: another modality for user interface, which is haptic bands. Basically, 302 00:17:43,680 --> 00:17:48,920 Speaker 3: using a combination of motors and vibration patterns, we can 303 00:17:49,000 --> 00:17:53,520 Speaker 3: communicate tons of information just using few motors, like even 304 00:17:53,600 --> 00:17:57,440 Speaker 3: less than ten morrors. And the current prototype that we're 305 00:17:57,440 --> 00:18:01,400 Speaker 3: working on is pretty simple version. It can be put 306 00:18:01,480 --> 00:18:06,480 Speaker 3: on the wrist and this can communicate potentially hundreds of 307 00:18:06,480 --> 00:18:10,880 Speaker 3: combinations of vibrations, and at some point we're really aiming 308 00:18:10,960 --> 00:18:14,240 Speaker 3: for a set of where we can communicate pretty much 309 00:18:14,280 --> 00:18:18,760 Speaker 3: everything the system sees through the happic vibrations. If a 310 00:18:18,880 --> 00:18:23,359 Speaker 3: user prefers completely one hundred percent haptic bands, that is 311 00:18:23,359 --> 00:18:26,080 Speaker 3: something that we are targeting for. At the same time, 312 00:18:26,119 --> 00:18:29,119 Speaker 3: some users might prefer, you know what, I want this 313 00:18:29,240 --> 00:18:32,200 Speaker 3: sort of information to be communicated via audio and some 314 00:18:32,240 --> 00:18:35,800 Speaker 3: sort of information with the haptics. We're also are working 315 00:18:35,840 --> 00:18:39,080 Speaker 3: with the combination of system as well, but having hapic 316 00:18:39,119 --> 00:18:42,919 Speaker 3: bands in a solution like this opens a sort of 317 00:18:42,920 --> 00:18:47,240 Speaker 3: a different dimension for the users here, especially we're visually impaired. 318 00:18:48,960 --> 00:18:52,399 Speaker 1: What Jagadish hints at with his explanation of haptic bands 319 00:18:52,480 --> 00:18:56,520 Speaker 1: versus audio interface is a very fascinating, multi pronged approach 320 00:18:56,560 --> 00:18:59,920 Speaker 1: to the solution. In technology terms, haptics is all about 321 00:19:00,040 --> 00:19:03,240 Speaker 1: how your device interacts with you through touch. Think of 322 00:19:03,320 --> 00:19:06,080 Speaker 1: the times when your phone vibrates in your pocket, or 323 00:19:06,359 --> 00:19:08,359 Speaker 1: when you play a video game and the controller shakes 324 00:19:08,359 --> 00:19:12,080 Speaker 1: in response to you taking damage from the endgame boss. Oftentimes, 325 00:19:12,080 --> 00:19:15,720 Speaker 1: in developing solutions for disabilities, there is a one size 326 00:19:15,720 --> 00:19:18,320 Speaker 1: fits all approach that seeks to do an adequate job 327 00:19:18,400 --> 00:19:21,359 Speaker 1: for the most number of people. This strategy fails to 328 00:19:21,400 --> 00:19:24,960 Speaker 1: take into account the nuances of the human experience in 329 00:19:25,000 --> 00:19:27,840 Speaker 1: the same way that some people are audio learners while 330 00:19:27,880 --> 00:19:30,679 Speaker 1: others are visual. When it comes to aiding someone with 331 00:19:30,720 --> 00:19:34,720 Speaker 1: a disability, it is important to consider what methods complement 332 00:19:34,760 --> 00:19:38,680 Speaker 1: their strengths and experiences. The beauty of how jagger Dish 333 00:19:38,960 --> 00:19:41,600 Speaker 1: seeks to develop this AI tool is that it is 334 00:19:41,720 --> 00:19:45,840 Speaker 1: constantly studying and creating more specialized options for the users, 335 00:19:46,880 --> 00:19:49,640 Speaker 1: from audio to haptics. It has the potential to grow 336 00:19:49,680 --> 00:19:52,240 Speaker 1: in a number of ways to accommodate the visually impaired 337 00:19:52,920 --> 00:19:55,800 Speaker 1: ways we never thought to supplement them, and maybe these 338 00:19:55,840 --> 00:19:58,880 Speaker 1: developments will even have an impact on those with perfect sight. 339 00:20:00,040 --> 00:20:03,479 Speaker 1: Leans into the human computer interface component that Lama mentioned earlier, 340 00:20:03,880 --> 00:20:06,840 Speaker 1: the constant study and assessment of how people will actually 341 00:20:06,920 --> 00:20:14,000 Speaker 1: use the tools. Given you're listening to technically Speaking an 342 00:20:14,000 --> 00:20:27,200 Speaker 1: Intel podcast will be right back. Welcome back to technically 343 00:20:27,240 --> 00:20:34,720 Speaker 1: speaking an Intel podcast. I'd just like to get more 344 00:20:34,760 --> 00:20:39,280 Speaker 1: broadly into Intel and it's AI efforts now, Jaged you 345 00:20:39,320 --> 00:20:42,879 Speaker 1: have a partnership with Intel. You've mentioned it before, how 346 00:20:43,200 --> 00:20:45,800 Speaker 1: you're working with Intel and how they've come to the party, 347 00:20:45,840 --> 00:20:49,160 Speaker 1: so to speak. What's it like working with their team 348 00:20:49,640 --> 00:20:52,159 Speaker 1: in terms of their support and assistance they've given you. 349 00:20:52,600 --> 00:20:55,960 Speaker 1: It's fantastic. The amount of exposure and the support that 350 00:20:56,000 --> 00:20:59,600 Speaker 1: we've received from Intel is really amazing. What we admire 351 00:20:59,600 --> 00:21:03,560 Speaker 1: about Intel is how open they are in developing the 352 00:21:03,800 --> 00:21:08,440 Speaker 1: solutions for accessibility, and they have a dedicated team who's 353 00:21:08,440 --> 00:21:10,920 Speaker 1: purely working on solutions like this. And we also got 354 00:21:10,960 --> 00:21:14,840 Speaker 1: opportunity to look into the projects that Lama's team has 355 00:21:14,840 --> 00:21:18,439 Speaker 1: been working on. They're simply superb. I think these solutions 356 00:21:18,480 --> 00:21:22,520 Speaker 1: like this are transformative and it's going to change lives 357 00:21:22,520 --> 00:21:26,040 Speaker 1: for people. And in terms of the support that we've received, 358 00:21:26,400 --> 00:21:30,760 Speaker 1: they have been helping us on many aspects all the 359 00:21:30,760 --> 00:21:35,280 Speaker 1: way from helping with putting up a process training the 360 00:21:35,320 --> 00:21:38,600 Speaker 1: model assets, you know, creating a platform for training models, 361 00:21:38,680 --> 00:21:42,920 Speaker 1: and also sharing the connections, and also with funding. So 362 00:21:43,160 --> 00:21:45,520 Speaker 1: a lot of the features that is going to come 363 00:21:45,560 --> 00:21:49,000 Speaker 1: as part of Phoenix is coming out of the IrDA project, 364 00:21:49,440 --> 00:21:53,520 Speaker 1: and the sort of feedbacks that would risk in improving 365 00:21:53,560 --> 00:21:57,720 Speaker 1: the solution is something that we don't get outset easily. 366 00:21:59,080 --> 00:22:01,800 Speaker 2: Yeah, and it's one of the things that I believe 367 00:22:02,280 --> 00:22:04,440 Speaker 2: we're really all about at Intel, right If you look 368 00:22:04,480 --> 00:22:07,440 Speaker 2: at our mission, it's really enriched the life of every 369 00:22:07,560 --> 00:22:10,280 Speaker 2: person on the planet, and every person, right, not just 370 00:22:10,359 --> 00:22:14,000 Speaker 2: able people. So it's really wonderful that you're seeing that 371 00:22:14,119 --> 00:22:17,479 Speaker 2: support and the diversity of the type of platforms and 372 00:22:17,480 --> 00:22:20,000 Speaker 2: solutions that we have, right. So I'm really just very 373 00:22:20,040 --> 00:22:23,240 Speaker 2: heartened by what you said, Drakhandesh. One of the things 374 00:22:23,280 --> 00:22:27,320 Speaker 2: maybe that is a top of mind project something called Omnibridge, 375 00:22:27,960 --> 00:22:32,080 Speaker 2: and Omnibridge is essentially a software that is meant to 376 00:22:32,520 --> 00:22:36,440 Speaker 2: bridge again the silence gap, but for people who are 377 00:22:36,960 --> 00:22:40,439 Speaker 2: hearing impaired, so that you know, essentially you're translating in 378 00:22:40,480 --> 00:22:43,720 Speaker 2: and out of like sign language, so you know, people 379 00:22:43,800 --> 00:22:47,400 Speaker 2: can sign into their PC and then the PC can 380 00:22:47,440 --> 00:22:51,959 Speaker 2: actually translate that into language on the other end, and 381 00:22:52,000 --> 00:22:54,679 Speaker 2: then vice versa. Right, So it's like, you know, what 382 00:22:54,720 --> 00:22:59,000 Speaker 2: you're really enabling again is to enable people in their 383 00:22:59,320 --> 00:23:01,720 Speaker 2: everyday life life to actually be able to do that. 384 00:23:01,840 --> 00:23:04,280 Speaker 2: And to be able to do that, you need a 385 00:23:04,320 --> 00:23:07,439 Speaker 2: lot of the AI support and AI compute on these platforms. 386 00:23:07,440 --> 00:23:09,119 Speaker 2: So one of the reasons again what I was saying, 387 00:23:09,400 --> 00:23:12,280 Speaker 2: it's really nice to see it at these platforms and 388 00:23:12,359 --> 00:23:14,840 Speaker 2: at the lowest cost that you can actually bring it. 389 00:23:14,920 --> 00:23:17,520 Speaker 2: You start to really democratize AI in ways that really 390 00:23:17,560 --> 00:23:18,800 Speaker 2: improve people's lives. 391 00:23:19,000 --> 00:23:21,200 Speaker 1: Yeah, for me, I mean one of the key things 392 00:23:21,200 --> 00:23:25,640 Speaker 1: you've just said is democratizing technology, and I think that's 393 00:23:25,680 --> 00:23:28,440 Speaker 1: the real power of it is. Yes, we can have 394 00:23:28,480 --> 00:23:33,320 Speaker 1: those really fancy solutions that Professor Hawking has, but for me, 395 00:23:33,480 --> 00:23:37,000 Speaker 1: it's about trying to get that cost down so that 396 00:23:37,080 --> 00:23:40,880 Speaker 1: it makes it so much easier for people to use. 397 00:23:41,320 --> 00:23:44,600 Speaker 2: So actually, just as a correction, so Professor Hawking didn't 398 00:23:44,640 --> 00:23:47,199 Speaker 2: have a fancy system, I was actually at PC with 399 00:23:47,280 --> 00:23:50,840 Speaker 2: a very lightweight sensor and in fact, like a big 400 00:23:50,880 --> 00:23:52,800 Speaker 2: part of what we've been really trying to do with 401 00:23:52,880 --> 00:23:55,920 Speaker 2: BCI is also democratize that because the problem with BCI 402 00:23:56,240 --> 00:23:58,840 Speaker 2: is if you want something with really high fidelity, you're 403 00:23:58,880 --> 00:24:02,320 Speaker 2: paying fifteen to two thousand dollars on a headset versus 404 00:24:02,359 --> 00:24:04,960 Speaker 2: what we're really trying to do is like use OpenBCI, 405 00:24:05,160 --> 00:24:08,159 Speaker 2: so it's you know, a really low cost, but you know, 406 00:24:08,280 --> 00:24:12,399 Speaker 2: compensate for the fidelity constraints with a lot of machine learning. 407 00:24:12,720 --> 00:24:16,679 Speaker 1: Okay, great, and jacket ashly did say it was relatively 408 00:24:17,320 --> 00:24:21,520 Speaker 1: low cost. Is that one of the primary motivating factors 409 00:24:21,520 --> 00:24:24,960 Speaker 1: for you? And how do you go about designing systems 410 00:24:24,960 --> 00:24:27,600 Speaker 1: to try and get that cost down. 411 00:24:28,240 --> 00:24:31,879 Speaker 3: Absolutely, it's a major restricting factor. Just a bit of context, 412 00:24:32,560 --> 00:24:36,320 Speaker 3: The unemployment rate in the visually impaired people community is 413 00:24:36,359 --> 00:24:39,880 Speaker 3: extremely high. I think it's more than sixty percent, so 414 00:24:40,000 --> 00:24:44,679 Speaker 3: it's hard for them to afford any product that is expensive. 415 00:24:45,160 --> 00:24:47,600 Speaker 3: And this is something that we want to change by 416 00:24:47,920 --> 00:24:51,439 Speaker 3: one making it completely open source, so that anybody in 417 00:24:51,480 --> 00:24:54,760 Speaker 3: the world they can just if they have the technical skills, 418 00:24:54,800 --> 00:24:57,240 Speaker 3: they can just assemble the system and they can get 419 00:24:57,240 --> 00:25:00,000 Speaker 3: the system. If not, we can help them assemble the system. 420 00:25:00,520 --> 00:25:03,400 Speaker 3: The complete solution is going to be open source. Two 421 00:25:03,640 --> 00:25:09,600 Speaker 3: is building the product using the hardware systems that are cheap. 422 00:25:10,280 --> 00:25:13,160 Speaker 3: At the same time, that are efficient, and that's where 423 00:25:13,240 --> 00:25:17,200 Speaker 3: products like Intelook stands apart one because it has very 424 00:25:17,200 --> 00:25:19,719 Speaker 3: good capability for running a lot of models in bartle 425 00:25:20,359 --> 00:25:24,600 Speaker 3: and also using accelerators like neural computer stick. So things 426 00:25:24,680 --> 00:25:27,359 Speaker 3: like that help us in shrinking the form factor and 427 00:25:27,400 --> 00:25:30,760 Speaker 3: also the cost quite a bit. And at the same time, 428 00:25:31,680 --> 00:25:34,919 Speaker 3: at the software design level, if you are putting in 429 00:25:34,960 --> 00:25:39,040 Speaker 3: a modular based design, where if somebody wants to use 430 00:25:39,080 --> 00:25:42,280 Speaker 3: a cheaper sensor, they could plug in a different sensor. 431 00:25:42,880 --> 00:25:46,600 Speaker 3: The rest of the robotics or stack will remain intact 432 00:25:46,880 --> 00:25:49,360 Speaker 3: as far as they take care of the sensor obstruction layer. 433 00:25:49,440 --> 00:25:53,080 Speaker 3: And same thing goes for probably for haptic interface, probably 434 00:25:53,080 --> 00:25:56,680 Speaker 3: audio interface, and potentially for computer interface. So we want 435 00:25:56,720 --> 00:25:59,159 Speaker 3: to modularize it as much as possible and shrink the 436 00:25:59,240 --> 00:26:01,400 Speaker 3: cost as much as possible. 437 00:26:02,960 --> 00:26:06,679 Speaker 1: Jagadish mentions something I had never really considered, which is 438 00:26:06,680 --> 00:26:10,679 Speaker 1: the difficulty in finding gainful employment for those with visual impairments. 439 00:26:11,760 --> 00:26:14,560 Speaker 1: In the US and other developed nations, there are protocols 440 00:26:14,560 --> 00:26:19,200 Speaker 1: to provide reasonable accommodations to workers with disabilities, but globally 441 00:26:19,320 --> 00:26:22,359 Speaker 1: that has yet to become a common practice. With an 442 00:26:22,359 --> 00:26:26,240 Speaker 1: AI tool such as Jagadees being open source. It really 443 00:26:26,280 --> 00:26:28,199 Speaker 1: helps move the needle in terms of what those with 444 00:26:28,320 --> 00:26:32,920 Speaker 1: visual impairments can do for themselves. Lama also mentions BCI, 445 00:26:33,359 --> 00:26:37,760 Speaker 1: or brain computer interfaces. Most brain computer interfaces use electrical 446 00:26:37,880 --> 00:26:41,240 Speaker 1: energy of the brain to directly interface with computers or machines. 447 00:26:42,080 --> 00:26:44,760 Speaker 1: The best way to imagine BCI is the character Cyborg 448 00:26:45,200 --> 00:26:48,040 Speaker 1: from the Teen Titan series, where he developed superpowers from 449 00:26:48,040 --> 00:26:52,800 Speaker 1: interfacing computing technology with his biological self. I'm wondering what 450 00:26:52,840 --> 00:26:56,280 Speaker 1: accommodations are considered when the user has ADHD or some 451 00:26:56,400 --> 00:26:58,440 Speaker 1: other form of cognitive processing disorder. 452 00:27:01,800 --> 00:27:07,359 Speaker 2: So we've been looking specifically at utilizing BCI for communication 453 00:27:07,920 --> 00:27:11,240 Speaker 2: for locked in patients, right, And really, you don't want 454 00:27:11,240 --> 00:27:14,240 Speaker 2: to use BCI for communication unless you have to, because 455 00:27:14,280 --> 00:27:18,360 Speaker 2: it's not i mean, unless you're actually have something that's 456 00:27:18,400 --> 00:27:21,639 Speaker 2: implanted in your brain. If you're going outside of the skull, 457 00:27:21,840 --> 00:27:25,040 Speaker 2: you have a very very noisy signal. So that's in 458 00:27:25,080 --> 00:27:27,360 Speaker 2: some sense you can think of it as a last resort. However, 459 00:27:27,680 --> 00:27:30,200 Speaker 2: what you just mentioned is something very different, right, which 460 00:27:30,280 --> 00:27:35,160 Speaker 2: is utilizing BCI as another sensing modality for all sorts 461 00:27:35,240 --> 00:27:39,879 Speaker 2: of other inferences, not to communicate your intention, but to 462 00:27:40,000 --> 00:27:43,960 Speaker 2: actually understand your state, and that is something that is 463 00:27:44,240 --> 00:27:48,680 Speaker 2: you know, yes, can be totally utilized for understanding, for example, 464 00:27:49,200 --> 00:27:54,080 Speaker 2: things like emotional state and concentration and focus and all 465 00:27:54,119 --> 00:27:57,400 Speaker 2: sorts of things like that that can help in cases 466 00:27:57,440 --> 00:28:00,240 Speaker 2: where you have people with autism, for example, and they're 467 00:28:00,280 --> 00:28:04,480 Speaker 2: having a hard time expressing emotional state as it's actually 468 00:28:04,480 --> 00:28:05,119 Speaker 2: getting worse. 469 00:28:05,240 --> 00:28:05,440 Speaker 1: Right. 470 00:28:05,520 --> 00:28:07,920 Speaker 2: There has been actually quite a lot of interesting research 471 00:28:08,080 --> 00:28:11,399 Speaker 2: out of Jojia Tech, for example, specifically looking at that 472 00:28:11,560 --> 00:28:15,000 Speaker 2: as an interesting modality for these type of settings. 473 00:28:15,440 --> 00:28:20,040 Speaker 1: In terms of other individuals and organizations contributing, you've mentioned 474 00:28:20,240 --> 00:28:24,480 Speaker 1: both of you mentioned the open source initiative that Intel's pushing. 475 00:28:25,160 --> 00:28:29,639 Speaker 1: If individuals and organizations want to be involved, what's the 476 00:28:29,640 --> 00:28:33,160 Speaker 1: best way for them to get in and start contributing. 477 00:28:33,920 --> 00:28:37,920 Speaker 2: So basically with ACAT, we have essentially it's an open 478 00:28:37,960 --> 00:28:41,239 Speaker 2: source project, right, and it's open to developers. We have 479 00:28:41,280 --> 00:28:44,800 Speaker 2: different people contributing all sorts of different things, right. I mean, 480 00:28:44,840 --> 00:28:48,400 Speaker 2: for example, we've seen a lot of interest in having 481 00:28:48,560 --> 00:28:52,120 Speaker 2: ACAT beyond in different languages for people around the world, Right, 482 00:28:52,160 --> 00:28:54,960 Speaker 2: So we have a way for having people easily contribute 483 00:28:55,000 --> 00:28:58,560 Speaker 2: to extend it to other languages. As an example, extending 484 00:28:58,640 --> 00:29:00,600 Speaker 2: it to other sensing modalities and so on, so you 485 00:29:00,600 --> 00:29:03,560 Speaker 2: can go through that project and then just kind of 486 00:29:03,560 --> 00:29:06,959 Speaker 2: communicate and submit what you want and communicate with us 487 00:29:06,840 --> 00:29:09,560 Speaker 2: as the people who are still kind of overseeing the project. 488 00:29:10,040 --> 00:29:13,120 Speaker 2: There are also like specific groups that we work with 489 00:29:13,160 --> 00:29:15,760 Speaker 2: because we're trying to also kind of get access to 490 00:29:16,600 --> 00:29:19,600 Speaker 2: users that we can test that technology with. So for example, 491 00:29:19,640 --> 00:29:21,760 Speaker 2: you know the M and D or the ALS groups 492 00:29:21,800 --> 00:29:25,200 Speaker 2: and things like that. So depending on usually some of 493 00:29:25,240 --> 00:29:27,760 Speaker 2: these groups have access to a lot of the different 494 00:29:27,800 --> 00:29:30,719 Speaker 2: solutions and open source systems that exist there. So that 495 00:29:30,880 --> 00:29:33,120 Speaker 2: also is a way I mean not necessarily just for ACAT, 496 00:29:33,160 --> 00:29:34,720 Speaker 2: but more broadly. 497 00:29:35,000 --> 00:29:37,840 Speaker 3: We are seeing the very strong trend of a lot 498 00:29:37,840 --> 00:29:43,120 Speaker 3: of projects being open sourced, and because of this trend, 499 00:29:43,160 --> 00:29:46,640 Speaker 3: we're seeing a lot of powerful projects being democratized and 500 00:29:46,800 --> 00:29:51,280 Speaker 3: reaching people much easily than before. In fact, a lot 501 00:29:51,320 --> 00:29:54,320 Speaker 3: of companies are actually following this model, starting to switch 502 00:29:54,360 --> 00:29:56,680 Speaker 3: from a different model to open sourcing model, which is 503 00:29:56,720 --> 00:30:01,440 Speaker 3: fantastic for the community. It's just fantastic for the world. However, 504 00:30:01,480 --> 00:30:04,680 Speaker 3: there are certain things needs to be considered when developing 505 00:30:04,680 --> 00:30:07,719 Speaker 3: open source solution. One of the most important things is 506 00:30:08,720 --> 00:30:11,920 Speaker 3: how an open source project is defined, how can it 507 00:30:12,000 --> 00:30:15,280 Speaker 3: evolve by itself at some point. Initiatively there is going 508 00:30:15,320 --> 00:30:17,280 Speaker 3: to be primary contributors, but at some point there is 509 00:30:17,320 --> 00:30:18,560 Speaker 3: going to be a lot of people. You're going to 510 00:30:18,560 --> 00:30:22,160 Speaker 3: get contribution from all over the world, and this can 511 00:30:22,200 --> 00:30:26,400 Speaker 3: be both good and bad. If the response is very high, 512 00:30:26,520 --> 00:30:29,480 Speaker 3: then the initial contributors cannot handle it, right, it might 513 00:30:29,560 --> 00:30:32,240 Speaker 3: end up pretty damaging, right, But at the same time, 514 00:30:32,320 --> 00:30:35,640 Speaker 3: you need those responses, So it's important to know that 515 00:30:35,800 --> 00:30:39,160 Speaker 3: balance and also come up with how do we address 516 00:30:39,200 --> 00:30:41,000 Speaker 3: this as a process in general? 517 00:30:41,080 --> 00:30:41,360 Speaker 1: Right? 518 00:30:41,520 --> 00:30:43,600 Speaker 3: How can somebody contribue, but how can somebody create a 519 00:30:43,600 --> 00:30:46,160 Speaker 3: pr It's going to be completely democratized and there will 520 00:30:46,200 --> 00:30:49,040 Speaker 3: be more reviewers distributed throughout the world. 521 00:30:49,720 --> 00:30:53,040 Speaker 2: One of the things that I'm really happy to see 522 00:30:53,440 --> 00:30:57,680 Speaker 2: is really the amount of contribution in the open source 523 00:30:58,240 --> 00:31:02,720 Speaker 2: on all sorts of AI capabilities and language models, and 524 00:31:02,880 --> 00:31:07,760 Speaker 2: which really I think is enabling a lot of democratization 525 00:31:07,920 --> 00:31:12,160 Speaker 2: of AI, specifically to all of these different usages. Right, 526 00:31:12,200 --> 00:31:14,600 Speaker 2: because if you think about a sist of computing in 527 00:31:14,680 --> 00:31:17,760 Speaker 2: some sense, in many cases you're trying to compensate for 528 00:31:17,880 --> 00:31:21,840 Speaker 2: some sort of a sense impairment. Right, So if you're 529 00:31:21,880 --> 00:31:25,280 Speaker 2: able to actually use AI to help extract that sense 530 00:31:25,320 --> 00:31:29,840 Speaker 2: automatically from the world. Having access to that democratization in 531 00:31:29,920 --> 00:31:33,800 Speaker 2: AI models and algorithms is something that is really transformational 532 00:31:33,920 --> 00:31:36,440 Speaker 2: for this space. And if I remember, for example, like 533 00:31:36,480 --> 00:31:39,480 Speaker 2: in the past, right even getting something access to something 534 00:31:39,520 --> 00:31:42,320 Speaker 2: like an ASR was really hard to do right in 535 00:31:42,640 --> 00:31:44,640 Speaker 2: the open source, at least the level of quality that 536 00:31:44,680 --> 00:31:48,520 Speaker 2: you would see. But now lately, because of that quick movement, 537 00:31:49,080 --> 00:31:51,640 Speaker 2: you're seeing a lot of capability in the open source 538 00:31:51,680 --> 00:31:55,320 Speaker 2: that actually rivals that of the really you know, big companies, 539 00:31:55,320 --> 00:31:57,800 Speaker 2: which is I think is absolutely transformational. 540 00:31:58,280 --> 00:32:01,240 Speaker 1: Yeah, that's great, no question for both of you. Are 541 00:32:01,320 --> 00:32:04,360 Speaker 1: start with jaggedesh. You know, we're seeing AI being used 542 00:32:04,400 --> 00:32:08,640 Speaker 1: for accessibility efforts. Looking forward ten years, what's the number 543 00:32:08,640 --> 00:32:12,680 Speaker 1: one area which you would want AI to help in 544 00:32:12,720 --> 00:32:13,760 Speaker 1: this industry. 545 00:32:14,400 --> 00:32:17,400 Speaker 3: I'd be really pleased to see a system that is 546 00:32:17,480 --> 00:32:20,720 Speaker 3: really small that somebody can put in like a glass 547 00:32:20,840 --> 00:32:24,720 Speaker 3: or any firm, that goes totally unnoticed and it provides 548 00:32:24,840 --> 00:32:27,959 Speaker 3: all the capabilities of human eye. I think that'll be fantastic, 549 00:32:28,360 --> 00:32:31,480 Speaker 3: And same thing goes for other forms of disabilities. I 550 00:32:31,480 --> 00:32:35,920 Speaker 3: think that will be fantastic to see and in tenuous timeline, 551 00:32:35,960 --> 00:32:37,320 Speaker 3: I think it might be possible. 552 00:32:38,040 --> 00:32:40,480 Speaker 2: My number one area that I want to see solved, 553 00:32:40,680 --> 00:32:43,920 Speaker 2: not necessarily in a sist of computing, but actually climate change. 554 00:32:44,000 --> 00:32:46,800 Speaker 2: That's where I think, like we all need this otherwise 555 00:32:46,800 --> 00:32:48,040 Speaker 2: I'm not sure we're going to have a world to 556 00:32:48,080 --> 00:32:50,640 Speaker 2: actually do anything else. And in the area of asis 557 00:32:50,640 --> 00:32:53,160 Speaker 2: of computing, it's really what I was saying earlier, which 558 00:32:53,200 --> 00:32:58,080 Speaker 2: is I envision being able to compensate for every single 559 00:32:58,120 --> 00:33:03,240 Speaker 2: sense that the human is missing, and that, to Jaggediesh's point, 560 00:33:03,760 --> 00:33:06,240 Speaker 2: is only going to be possible if that is meeting 561 00:33:06,280 --> 00:33:08,880 Speaker 2: people where there are in the world, which means they 562 00:33:08,880 --> 00:33:11,680 Speaker 2: have to be sustainable, they have to be extremely power efficient, 563 00:33:11,720 --> 00:33:13,640 Speaker 2: they need to be robust enough to everything that it 564 00:33:13,680 --> 00:33:16,200 Speaker 2: hasn't seen in the world right, so, which is really 565 00:33:16,240 --> 00:33:19,480 Speaker 2: not necessarily where things are today, but you know, given 566 00:33:19,520 --> 00:33:22,360 Speaker 2: that appid improvement, I would really hope that that's where 567 00:33:22,360 --> 00:33:23,680 Speaker 2: we would be in ten years from now. 568 00:33:24,000 --> 00:33:27,479 Speaker 1: Excellent. Okay, thank you very much, thank you, thank you. 569 00:33:30,160 --> 00:33:32,760 Speaker 1: I would like to thank my guests Jaggedish Mahindra and 570 00:33:32,840 --> 00:33:36,400 Speaker 1: Lama Nachman for joining me on this episode of Technically Speaking, 571 00:33:36,600 --> 00:33:41,200 Speaker 1: an Intel podcast. I really enjoyed this conversation with Jagged 572 00:33:41,240 --> 00:33:43,480 Speaker 1: Ash and Lama. I love being able to delve into 573 00:33:43,480 --> 00:33:46,760 Speaker 1: the motivations of the why, but also the how. You 574 00:33:46,800 --> 00:33:49,200 Speaker 1: heard from Jaggedesh and the story of his visually impaired 575 00:33:49,240 --> 00:33:51,800 Speaker 1: friend being struck by a tree branch, and that was 576 00:33:51,800 --> 00:33:54,880 Speaker 1: the seed for his idea for an AI assistant backpack. 577 00:33:55,560 --> 00:33:59,240 Speaker 1: For me, this is the true technological empowerment, the ability 578 00:33:59,280 --> 00:34:01,680 Speaker 1: for individuals to use their skills and talent to make 579 00:34:01,720 --> 00:34:05,240 Speaker 1: a difference, taking action rather than just talking about it. 580 00:34:05,880 --> 00:34:08,880 Speaker 1: These are the true innovators. It was great to hear 581 00:34:08,920 --> 00:34:11,880 Speaker 1: of Lama's work with Professor Stephen Hawking and the context 582 00:34:11,920 --> 00:34:14,800 Speaker 1: of where system her team developed. What is so pleasing 583 00:34:14,800 --> 00:34:17,080 Speaker 1: to me was that it wasn't a Rolls Royce design, 584 00:34:17,440 --> 00:34:21,440 Speaker 1: but rather an elegant yet simple system of sensors connected 585 00:34:21,440 --> 00:34:24,160 Speaker 1: to a PC to allow Professor Hawking to interact and 586 00:34:24,160 --> 00:34:29,080 Speaker 1: communicate with others. Because of this relatively inexpensive solution, it 587 00:34:29,120 --> 00:34:31,920 Speaker 1: can be used by a wider range of people. This 588 00:34:32,080 --> 00:34:36,000 Speaker 1: is what democratization of technology does for the world. I 589 00:34:36,040 --> 00:34:38,839 Speaker 1: hope that Lama and Jaggedish's stories inspire you to take 590 00:34:38,840 --> 00:34:41,600 Speaker 1: the leap and contribute to improving the lives of people, 591 00:34:41,640 --> 00:34:46,400 Speaker 1: regardless of their background. Please join us on Tuesday, November 592 00:34:46,400 --> 00:34:50,160 Speaker 1: fourteenth for the next episode of Technically Speaking, an Intel podcast. 593 00:34:54,480 --> 00:34:57,960 Speaker 1: Technically Speaking was produced by Ruby Studios from iHeartRadio in 594 00:34:58,000 --> 00:35:01,879 Speaker 1: partnership with Intel, and hosted by me Graham Class. Our 595 00:35:01,920 --> 00:35:05,360 Speaker 1: executive producer is Molly Sosher, our EP of Post Production 596 00:35:05,480 --> 00:35:09,040 Speaker 1: is James Foster, and our Supervising producer is Nikair Swinton. 597 00:35:09,880 --> 00:35:13,080 Speaker 1: This episode was edited by Sierra Spreen and written and 598 00:35:13,120 --> 00:35:23,719 Speaker 1: produced by Tyree Rush.