1 00:00:00,040 --> 00:00:05,000 Speaker 1: Hey everyone, it's Robert and Joe here. Today we've got 2 00:00:05,040 --> 00:00:06,960 Speaker 1: something a little bit different to share with you. It 3 00:00:07,080 --> 00:00:10,680 Speaker 1: is a new edition of the Smart Talks podcast series, 4 00:00:10,720 --> 00:00:14,319 Speaker 1: which is produced in partnership with IBM. This season of 5 00:00:14,360 --> 00:00:18,640 Speaker 1: Smart Talks with IBM is all about new creators, the developers, 6 00:00:19,040 --> 00:00:22,600 Speaker 1: data scientists, c t o s, and other visionaries creatively 7 00:00:22,640 --> 00:00:27,120 Speaker 1: applying technology and business to drive change. They use their 8 00:00:27,160 --> 00:00:30,640 Speaker 1: knowledge and creativity to develop better ways of working, no 9 00:00:30,720 --> 00:00:34,840 Speaker 1: matter the industry. Join hosts from your favorite Pushkin Industries 10 00:00:34,920 --> 00:00:38,560 Speaker 1: podcast as they use their expertise to deepen these conversations. 11 00:00:39,040 --> 00:00:41,800 Speaker 1: Malcolm Gladwell will guide you through this season as your 12 00:00:41,840 --> 00:00:45,000 Speaker 1: host to provide his thoughts and analysis along the way. 13 00:00:45,320 --> 00:00:48,600 Speaker 1: Look out for new episodes of Smart Talks with IBM 14 00:00:48,680 --> 00:00:52,040 Speaker 1: every month on the I Heart Radio app, Apple Podcasts, 15 00:00:52,159 --> 00:00:55,480 Speaker 1: or wherever you get your podcasts. And learn more at 16 00:00:55,520 --> 00:01:04,480 Speaker 1: IBM dot com slash smart Talks. Hello, Hello, Welcome to 17 00:01:04,560 --> 00:01:08,400 Speaker 1: Smart Talks with IBM, a podcast from Pushkin Industries, I 18 00:01:08,600 --> 00:01:13,040 Speaker 1: Heart Radio and IBM. I'm Malcolm Babbo. This season, we're 19 00:01:13,040 --> 00:01:17,440 Speaker 1: talking to new creators, the developers, data scientists, c t 20 00:01:17,600 --> 00:01:21,360 Speaker 1: o s, and other visionaries who are creatively applying technology 21 00:01:21,400 --> 00:01:25,559 Speaker 1: and business to drive change. Channeling their knowledge and expertise, 22 00:01:25,760 --> 00:01:30,400 Speaker 1: they're developing more creative and effective solutions, no matter the industry. 23 00:01:31,319 --> 00:01:35,800 Speaker 1: Our guest today is Padre Bonadius, trust in AI Practice 24 00:01:35,880 --> 00:01:41,480 Speaker 1: leader within IBM Consulting. Advocating for artificial intelligence built and 25 00:01:41,560 --> 00:01:46,280 Speaker 1: deployed responsibly is no longer just a compliance issue, but 26 00:01:46,440 --> 00:01:50,440 Speaker 1: a business imperative. Part of Phaedre's job is to help 27 00:01:50,480 --> 00:01:55,280 Speaker 1: companies identify potential risks and pitfalls way before any code 28 00:01:55,320 --> 00:01:58,760 Speaker 1: is written. In today's show, you'll hear how Phaedre's team 29 00:01:58,760 --> 00:02:02,840 Speaker 1: and IBM is a pro genus challenge holistically and creatively. 30 00:02:03,560 --> 00:02:06,920 Speaker 1: Pedra spoke with Dr Lois Santos, host of the Pushkin 31 00:02:07,000 --> 00:02:11,160 Speaker 1: podcast The Happiness Lab. Laurie is a professor of psychology 32 00:02:11,200 --> 00:02:14,760 Speaker 1: at Yale University and an expert on human cognition and 33 00:02:14,800 --> 00:02:21,000 Speaker 1: the cognitive biases that impede better choices. Now let's get 34 00:02:21,040 --> 00:02:29,919 Speaker 1: to the interview, Pedro. I'm so excited that we get 35 00:02:29,919 --> 00:02:32,799 Speaker 1: a chance to chat today. You know, just to start off, 36 00:02:32,840 --> 00:02:35,560 Speaker 1: I'm wondering how did you get started in this role 37 00:02:35,560 --> 00:02:37,760 Speaker 1: at IBM, Like, what's the story to how you got 38 00:02:37,760 --> 00:02:41,160 Speaker 1: where you are? Today. Oh goodness. My background is actually 39 00:02:41,760 --> 00:02:45,239 Speaker 1: from the world of video games for entertainment, so AI 40 00:02:45,400 --> 00:02:48,280 Speaker 1: has always been very interesting to me, especially when you 41 00:02:48,280 --> 00:02:52,400 Speaker 1: intersect AI and play. But several years ago I began 42 00:02:52,440 --> 00:02:57,480 Speaker 1: to get very frustrated by what I was reading in 43 00:02:57,520 --> 00:03:03,799 Speaker 1: the news with respect to malintent through the use of AI. 44 00:03:04,000 --> 00:03:06,960 Speaker 1: And the more that I learned and the more that 45 00:03:07,040 --> 00:03:11,160 Speaker 1: I studied about this space of AI and ethics, the 46 00:03:11,200 --> 00:03:15,800 Speaker 1: more I recognized that even organizations that have the very 47 00:03:16,160 --> 00:03:23,160 Speaker 1: very best of intentions could inadvertently cause potential harm. And 48 00:03:23,280 --> 00:03:25,520 Speaker 1: so that's super cool. I love that your interest in 49 00:03:25,720 --> 00:03:29,280 Speaker 1: more responsible AI came from the gaming world. You have 50 00:03:29,360 --> 00:03:31,480 Speaker 1: to talk a little bit about your history with gaming 51 00:03:31,480 --> 00:03:34,880 Speaker 1: and that how that informed your interest and trustworthy AI. Well, 52 00:03:35,240 --> 00:03:40,400 Speaker 1: it wasn't as much necessarily the ethical components of AI 53 00:03:40,440 --> 00:03:43,680 Speaker 1: when I was working in games. It was more things like, 54 00:03:44,600 --> 00:03:48,160 Speaker 1: look at what non player characters can do, you know, 55 00:03:48,400 --> 00:03:50,320 Speaker 1: I mean, if you've got an AI acting as a 56 00:03:50,400 --> 00:03:53,240 Speaker 1: character within the game, and how is it that you 57 00:03:53,280 --> 00:03:55,640 Speaker 1: can use AI in order to make a game a 58 00:03:55,680 --> 00:04:00,160 Speaker 1: more interesting experience. Actually ended up joining IBM to be 59 00:04:00,200 --> 00:04:03,280 Speaker 1: our first global lead for something called serious games, which 60 00:04:03,280 --> 00:04:05,119 Speaker 1: is when you use video games to do something other 61 00:04:05,200 --> 00:04:08,560 Speaker 1: than just entertaining. And so the idea of integrating real 62 00:04:08,640 --> 00:04:12,440 Speaker 1: data and real processes within sophisticated games powered by AI 63 00:04:12,600 --> 00:04:16,960 Speaker 1: to solve complex problems. It wasn't until, as I mentioned, 64 00:04:17,080 --> 00:04:20,240 Speaker 1: like later, when we started to hear all of us 65 00:04:20,279 --> 00:04:25,440 Speaker 1: more and more news about just problems what could happen 66 00:04:25,480 --> 00:04:28,080 Speaker 1: with respect to rendering or putting out models that are 67 00:04:28,080 --> 00:04:31,680 Speaker 1: inaccurate or unfair. I know one of your inspirations for 68 00:04:31,720 --> 00:04:34,160 Speaker 1: hearing other interviews that you've done is sci Fi. I'm 69 00:04:34,200 --> 00:04:36,320 Speaker 1: also a sci Fi nerd, and I know sci Fi 70 00:04:36,400 --> 00:04:40,160 Speaker 1: has talked a lot about, you know, the trustworthiness issues 71 00:04:40,200 --> 00:04:42,360 Speaker 1: that come up when we're dealing with AI and so on, 72 00:04:42,680 --> 00:04:44,520 Speaker 1: and so talk a little bit about how you bring 73 00:04:44,560 --> 00:04:46,960 Speaker 1: that to your work in developing AI. That's a little 74 00:04:46,960 --> 00:04:51,000 Speaker 1: bit more ethical. A lovely question. So, my my parents 75 00:04:51,160 --> 00:04:54,640 Speaker 1: were major techno files. They both were immigrants to the 76 00:04:54,760 --> 00:04:57,839 Speaker 1: United States, came here to study engineering, and they met 77 00:04:58,600 --> 00:05:01,680 Speaker 1: UH in college. Growing up, my sister and I we 78 00:05:01,760 --> 00:05:08,400 Speaker 1: had Star Trek playing every night. My parents were both 79 00:05:08,560 --> 00:05:12,960 Speaker 1: big fans of Gene Roddenberry's vision of how technology could 80 00:05:12,960 --> 00:05:17,440 Speaker 1: really be used to help better humankind, and that was 81 00:05:17,600 --> 00:05:21,279 Speaker 1: the ethos that, of course we grew up in. The 82 00:05:21,360 --> 00:05:25,960 Speaker 1: wonderful thing about science fiction isn't that it predicts cars, 83 00:05:26,120 --> 00:05:29,840 Speaker 1: for example, but that it predicts traffic jams. You know. 84 00:05:30,480 --> 00:05:32,880 Speaker 1: And I think there's just so much we can learn 85 00:05:33,640 --> 00:05:36,840 Speaker 1: from science fiction, or in fact, like I said, play 86 00:05:36,880 --> 00:05:40,240 Speaker 1: as a mechanism to be able to teach science fiction 87 00:05:40,360 --> 00:05:44,760 Speaker 1: predicting traffic jams. I love it. But when we think 88 00:05:44,760 --> 00:05:48,040 Speaker 1: about AI and science fiction, we need to be careful. 89 00:05:48,680 --> 00:05:51,719 Speaker 1: We need to remember that AI is not something that's 90 00:05:51,720 --> 00:05:53,760 Speaker 1: going to enter our lives at some point in the 91 00:05:53,880 --> 00:05:59,200 Speaker 1: distant future. AI is something that's all around us today. 92 00:05:59,480 --> 00:06:03,200 Speaker 1: If you have a virtual assistant in your house, that's AI, 93 00:06:03,480 --> 00:06:07,480 Speaker 1: your phone app that predicts traffic AI. When a streaming 94 00:06:07,480 --> 00:06:12,240 Speaker 1: service recommends a movie, you've guessed it AI, Paeder says. 95 00:06:12,320 --> 00:06:16,400 Speaker 1: AI maybe behind the scenes determining the interest rate on 96 00:06:16,480 --> 00:06:19,400 Speaker 1: your loan, or even whether or not you're the right 97 00:06:19,440 --> 00:06:23,119 Speaker 1: candidate for that job you applied for. AI is both 98 00:06:23,320 --> 00:06:27,520 Speaker 1: ubiquitous and invisible, which is why it is so crucial 99 00:06:27,720 --> 00:06:32,080 Speaker 1: the companies learn how to build trustworthy AI. How do 100 00:06:32,160 --> 00:06:35,320 Speaker 1: we do that? When thinking about what does it take 101 00:06:35,440 --> 00:06:39,839 Speaker 1: to earn trust in something like an AI there are 102 00:06:39,920 --> 00:06:44,120 Speaker 1: fundamentally human centric questions to be asked, right, like, what 103 00:06:44,279 --> 00:06:47,720 Speaker 1: is the intent of this particular AI model? How accurate 104 00:06:47,880 --> 00:06:52,000 Speaker 1: is that model? How fair is it? Is it explainable 105 00:06:52,080 --> 00:06:55,400 Speaker 1: if it makes a decision that could directly affect my livelihood? 106 00:06:56,120 --> 00:06:59,000 Speaker 1: Can I inquire what data did you use about me 107 00:06:59,279 --> 00:07:03,400 Speaker 1: to make this decision? Is it protecting my data? Is 108 00:07:03,440 --> 00:07:07,760 Speaker 1: it robust? Is it protected against people who could trick 109 00:07:07,839 --> 00:07:11,200 Speaker 1: it to disadvantage me over others? I mean, there's so 110 00:07:11,280 --> 00:07:15,680 Speaker 1: many questions to be asked. Earning trust in something like 111 00:07:15,800 --> 00:07:21,240 Speaker 1: AI is fundamentally not a technological challenge but a socio 112 00:07:21,240 --> 00:07:26,560 Speaker 1: technological challenge. It can't just be solved with a tool alone. 113 00:07:27,880 --> 00:07:29,880 Speaker 1: What are the kinds of risks that companies have to 114 00:07:29,920 --> 00:07:32,720 Speaker 1: think through? Is they're developing these technologies to make sure 115 00:07:32,720 --> 00:07:35,520 Speaker 1: they're as trustworthy as possible. Well, you know, they may 116 00:07:35,560 --> 00:07:39,280 Speaker 1: be putting a lot of money into investing in AI 117 00:07:39,720 --> 00:07:42,680 Speaker 1: that gets stuck in proof of concept planned like it's 118 00:07:42,720 --> 00:07:45,000 Speaker 1: get stuck in pilot. We we've done some research where 119 00:07:45,000 --> 00:07:48,440 Speaker 1: we have found about eight percent of investments in AI 120 00:07:48,600 --> 00:07:53,400 Speaker 1: get stuck, and sometimes it's because the investment isn't tied 121 00:07:53,480 --> 00:07:56,200 Speaker 1: directly to a business strategy, or more often than not, 122 00:07:56,360 --> 00:07:59,560 Speaker 1: people simply don't trust the results of the AI model. 123 00:08:00,640 --> 00:08:02,680 Speaker 1: As a company, who is of course thinking about this 124 00:08:02,760 --> 00:08:05,600 Speaker 1: so deeply. What do businesses need to consider when they're 125 00:08:05,640 --> 00:08:07,880 Speaker 1: trying to figure out, you know, how to solve this 126 00:08:07,960 --> 00:08:11,560 Speaker 1: big puzzle of AI ethics. It has to be approached holistically, 127 00:08:11,840 --> 00:08:15,240 Speaker 1: So you've got to be thinking about, for example, what 128 00:08:15,440 --> 00:08:19,840 Speaker 1: culture is required within your organization in order to really 129 00:08:19,880 --> 00:08:23,600 Speaker 1: be able to responsibly create AI, what processes are in 130 00:08:23,640 --> 00:08:26,240 Speaker 1: place to make sure that you're being compliant and that 131 00:08:26,280 --> 00:08:30,480 Speaker 1: your your practitioners know what to do. And then of 132 00:08:30,520 --> 00:08:34,640 Speaker 1: course AI engineering frameworks and tooling that can assist you 133 00:08:34,679 --> 00:08:38,680 Speaker 1: on this journey. There is so much fundamentally to do. 134 00:08:39,440 --> 00:08:44,200 Speaker 1: We found that actually those that were leading responsible AI 135 00:08:44,360 --> 00:08:47,960 Speaker 1: trust where the AI initiatives within their organization has switched 136 00:08:48,000 --> 00:08:51,560 Speaker 1: in the last three years. It used to be technical leaders, 137 00:08:52,120 --> 00:08:55,880 Speaker 1: for example, chief data officer or someone who is a 138 00:08:55,920 --> 00:08:59,720 Speaker 1: PhD in machine learning, and now it's switched to be 139 00:08:59,840 --> 00:09:03,280 Speaker 1: a the percent of those leaders are now non technical 140 00:09:03,800 --> 00:09:07,920 Speaker 1: business leaders, maybe you know, chief compliance officer, chief diversity 141 00:09:07,960 --> 00:09:12,320 Speaker 1: inclusivity officers, chief legal officer. So we're seeing a shift, 142 00:09:12,400 --> 00:09:17,760 Speaker 1: and I believe firmly. It's a recognition from organizations that 143 00:09:17,800 --> 00:09:21,360 Speaker 1: are seeing that in order to really pull this off well, 144 00:09:21,400 --> 00:09:25,560 Speaker 1: there has to be an investment than a focus in culture, 145 00:09:26,240 --> 00:09:30,080 Speaker 1: in people and getting people to understand why they should 146 00:09:30,080 --> 00:09:34,760 Speaker 1: care about this space. And so I see two challenges 147 00:09:34,840 --> 00:09:37,160 Speaker 1: with doing that right. One is, you know a lot 148 00:09:37,200 --> 00:09:40,640 Speaker 1: of these technology companies are really built to be tech companies, 149 00:09:40,640 --> 00:09:44,319 Speaker 1: not necessarily you know, social tech companies or having this 150 00:09:44,440 --> 00:09:47,800 Speaker 1: sort of training and ethics and beyond. Another issue seems 151 00:09:47,840 --> 00:09:51,520 Speaker 1: to be that you're really proposing a switch that's truly holistic, right, 152 00:09:51,600 --> 00:09:54,760 Speaker 1: that's like rethinking the way the company thinks about its 153 00:09:54,800 --> 00:09:57,800 Speaker 1: bottom line. And so as you think about working through 154 00:09:57,840 --> 00:10:00,600 Speaker 1: these kinds of challenges at IBM, how you tackled this, 155 00:10:00,679 --> 00:10:02,720 Speaker 1: like how have you brought new talent in? How have 156 00:10:02,760 --> 00:10:05,520 Speaker 1: you thought really carefully about this big holistic switch that 157 00:10:05,559 --> 00:10:08,520 Speaker 1: needs to come to make AI more trustworthy. Data is 158 00:10:08,559 --> 00:10:12,080 Speaker 1: an artifact of the human experience. And if you start 159 00:10:12,120 --> 00:10:15,480 Speaker 1: with that as your definition and then think about well 160 00:10:16,559 --> 00:10:19,840 Speaker 1: data is curated by data side this all data is 161 00:10:19,920 --> 00:10:25,600 Speaker 1: biased and so if you're not recognizing bias with eyes 162 00:10:25,720 --> 00:10:31,760 Speaker 1: fully open, then ultimately you're calcifying systemic bias into systems 163 00:10:31,800 --> 00:10:35,000 Speaker 1: like AI. So some of the things that we've done 164 00:10:35,280 --> 00:10:40,080 Speaker 1: at IBM again recognizing this important need for culture is big, big, 165 00:10:40,120 --> 00:10:44,520 Speaker 1: big focus on diversity, not only looking at teams of 166 00:10:44,600 --> 00:10:47,520 Speaker 1: data scientists and saying how many women are on this team, 167 00:10:47,640 --> 00:10:52,439 Speaker 1: how many minorities are on this team, but also insisting 168 00:10:52,559 --> 00:10:56,040 Speaker 1: on recognizing that we need to bring in people with 169 00:10:56,120 --> 00:11:00,600 Speaker 1: different world views too, For example, what's your definition of fairness? 170 00:11:01,400 --> 00:11:05,240 Speaker 1: Is your definition equality is an equity? Also bringing people 171 00:11:05,280 --> 00:11:09,760 Speaker 1: with a wider variety of skill sets and roles, including 172 00:11:09,760 --> 00:11:18,640 Speaker 1: our social scientists, anthropologists, sociologists, psychologists like yourself, right, behavioral scientists, designers. 173 00:11:18,679 --> 00:11:23,160 Speaker 1: I mean we have one of the leading AI design 174 00:11:23,440 --> 00:11:27,040 Speaker 1: practices in the world. I mean the effort, the investments 175 00:11:27,040 --> 00:11:31,640 Speaker 1: we've been making in design thinking as a mechanism to 176 00:11:31,679 --> 00:11:36,240 Speaker 1: create frameworks for systemic empathy well before any code is written, 177 00:11:36,840 --> 00:11:41,040 Speaker 1: so people can think through, how would you design in 178 00:11:41,120 --> 00:11:45,000 Speaker 1: order to mitigate for any potential harm given not only 179 00:11:45,040 --> 00:11:47,520 Speaker 1: the values of your organization, but what are the rights 180 00:11:47,520 --> 00:11:52,760 Speaker 1: of individuals asking oneself? These kinds of questions reinforces than 181 00:11:52,960 --> 00:11:56,760 Speaker 1: the idea. The ethics doesn't come at the end like 182 00:11:56,880 --> 00:12:00,360 Speaker 1: it's some kind of quality assurance, like check I passed 183 00:12:00,360 --> 00:12:03,559 Speaker 1: the audit, I've got to go, you know. But instead, really, 184 00:12:03,600 --> 00:12:05,959 Speaker 1: you know, as soon as you're thinking about using an 185 00:12:05,960 --> 00:12:09,760 Speaker 1: AI for a particular use case, thinking about you know, 186 00:12:09,880 --> 00:12:13,640 Speaker 1: what is the intent of this model, what's the relationship 187 00:12:13,679 --> 00:12:17,400 Speaker 1: we ultimately want to have with AI? And again, these 188 00:12:17,440 --> 00:12:22,160 Speaker 1: are non technology questions. This is where social scientists. Having 189 00:12:22,160 --> 00:12:26,400 Speaker 1: a social scientist on your team helping think through these 190 00:12:26,480 --> 00:12:30,480 Speaker 1: kinds of questions is is critical. Let's pause here for 191 00:12:30,520 --> 00:12:34,319 Speaker 1: a second, because this is a really profound idea. Building 192 00:12:34,480 --> 00:12:38,199 Speaker 1: responsible AI does not mean that you create a system 193 00:12:38,440 --> 00:12:41,679 Speaker 1: then check in at the end and say is this okay? 194 00:12:41,720 --> 00:12:45,319 Speaker 1: Is this ethical? If you don't ask those questions until 195 00:12:45,360 --> 00:12:48,920 Speaker 1: the end of the process, you've already failed. You have 196 00:12:49,000 --> 00:12:52,559 Speaker 1: to think about ethics from the jump from the makeup 197 00:12:52,640 --> 00:12:54,800 Speaker 1: of the team to the data you're using to train 198 00:12:54,880 --> 00:12:57,880 Speaker 1: the model to the most basic question of all, is 199 00:12:57,920 --> 00:13:02,240 Speaker 1: this even the right use case artificial intelligence? The big 200 00:13:02,320 --> 00:13:07,360 Speaker 1: lesson from IBM is this responsible AI is something you 201 00:13:07,400 --> 00:13:12,200 Speaker 1: build at every step of the process. So this season 202 00:13:12,200 --> 00:13:15,200 Speaker 1: of smart Talk is all focused on creativity and business. 203 00:13:15,720 --> 00:13:18,360 Speaker 1: My guess is that thinking about trustworthy AI involves a 204 00:13:18,400 --> 00:13:20,800 Speaker 1: lot of creativity. But talk to me about some of 205 00:13:20,800 --> 00:13:23,320 Speaker 1: the spots where you see this work as being most creative. 206 00:13:24,160 --> 00:13:28,760 Speaker 1: Oh goodness, I would say incorporating design design thinking in 207 00:13:28,840 --> 00:13:33,320 Speaker 1: particular as well as straight up design in order to 208 00:13:33,720 --> 00:13:38,240 Speaker 1: craft AI responsibly. You've used this word design thinking, and 209 00:13:38,280 --> 00:13:40,320 Speaker 1: so I'm wondering exactly what you mean here? How do 210 00:13:40,360 --> 00:13:43,760 Speaker 1: you define this idea of design thinking. Design thinking is 211 00:13:43,800 --> 00:13:47,520 Speaker 1: a practice that we established here at IBM many years ago. 212 00:13:47,800 --> 00:13:51,559 Speaker 1: In essence, what it is, it's a way of working 213 00:13:51,920 --> 00:13:58,360 Speaker 1: with groups of people to co create a vision for something, 214 00:13:58,400 --> 00:14:01,280 Speaker 1: for a product or a sir risk or an outcome. 215 00:14:02,160 --> 00:14:07,200 Speaker 1: And typically it starts with things like, for example, empathy maps, 216 00:14:07,200 --> 00:14:10,440 Speaker 1: like if you're thinking about an end user, thinking through 217 00:14:10,559 --> 00:14:14,600 Speaker 1: what is this person thinking, seeing, hearing, feeling, like what 218 00:14:14,679 --> 00:14:19,400 Speaker 1: are they experiencing in order to ultimately craft and experience 219 00:14:19,440 --> 00:14:23,520 Speaker 1: for them that is targeted specifically for them. So we 220 00:14:23,640 --> 00:14:27,200 Speaker 1: use it in a really wide variety of different ways 221 00:14:27,720 --> 00:14:32,040 Speaker 1: with respect to trustworthy AI, even rendering an AI model 222 00:14:32,440 --> 00:14:35,320 Speaker 1: explainable to a subject. And I'll give you an example. 223 00:14:35,840 --> 00:14:39,720 Speaker 1: So we've got this wonderful program with an IBM caller, 224 00:14:39,760 --> 00:14:42,760 Speaker 1: our Academy of Technology, and we take on initiatives that 225 00:14:42,920 --> 00:14:47,480 Speaker 1: steer the company in innovative new directions. So we had 226 00:14:47,520 --> 00:14:51,480 Speaker 1: an initiative where it was titled What the Titanic taught 227 00:14:51,520 --> 00:14:58,920 Speaker 1: Us About Explainable AI, and the project was imagining if 228 00:14:58,960 --> 00:15:02,320 Speaker 1: there was an AI mo utle that could predict the 229 00:15:02,440 --> 00:15:06,479 Speaker 1: likelihood of a passenger getting a life raft on the Titanic. 230 00:15:07,000 --> 00:15:09,800 Speaker 1: And we broke up into two work streams. One was 231 00:15:09,840 --> 00:15:12,680 Speaker 1: the workstream full of the data scientists who were using 232 00:15:12,720 --> 00:15:15,320 Speaker 1: all the different explainers to come up with the predictions 233 00:15:15,320 --> 00:15:17,920 Speaker 1: and they would crank out the numbers. And the other 234 00:15:18,080 --> 00:15:22,160 Speaker 1: team here's where the social scientists lived and the designers 235 00:15:22,240 --> 00:15:25,400 Speaker 1: were right where we were thinking through how do we 236 00:15:25,440 --> 00:15:32,280 Speaker 1: empower people? Well, how do we explain this algorithm and 237 00:15:32,480 --> 00:15:36,440 Speaker 1: this predictor and the accuracy behind this prediction in such 238 00:15:36,480 --> 00:15:39,240 Speaker 1: a way as to ultimately empower an end users? They 239 00:15:39,280 --> 00:15:43,480 Speaker 1: could decide I'm not getting on that boat, or I 240 00:15:43,560 --> 00:15:47,680 Speaker 1: want to get a second opinion please, or I went 241 00:15:47,800 --> 00:15:52,480 Speaker 1: to contest the outputs of this model because I upgraded 242 00:15:52,880 --> 00:15:56,400 Speaker 1: to first class just yesterday. See what I'm saying. And 243 00:15:56,520 --> 00:16:00,680 Speaker 1: that takes a lot of creativity. How do you design 244 00:16:00,760 --> 00:16:04,760 Speaker 1: and experience for someone in order to ultimately empower them. 245 00:16:05,480 --> 00:16:10,400 Speaker 1: So design design design is critically critically important. And why 246 00:16:10,440 --> 00:16:12,480 Speaker 1: I mentioned you know, we we've got to open up 247 00:16:12,480 --> 00:16:15,040 Speaker 1: the aperture with respect to who we invite to the table, 248 00:16:15,080 --> 00:16:18,720 Speaker 1: and these kinds of conversations. Taking the time to really 249 00:16:18,800 --> 00:16:23,160 Speaker 1: understand other people's perspectives is so important when you're doing 250 00:16:23,200 --> 00:16:27,040 Speaker 1: anything creative, and it is fundamental to the way the 251 00:16:27,120 --> 00:16:31,080 Speaker 1: new creators work. The core question you should always be 252 00:16:31,160 --> 00:16:34,560 Speaker 1: asking is where will the user be meeting this product? 253 00:16:35,440 --> 00:16:39,640 Speaker 1: As Peder said, what will they be thinking, seeing, hearing, feeling. 254 00:16:40,360 --> 00:16:43,160 Speaker 1: If you can answer those questions the way IBM does 255 00:16:43,280 --> 00:16:46,600 Speaker 1: in its design thinking practice, you will be in great 256 00:16:46,640 --> 00:16:50,680 Speaker 1: shape to create almost anything. Really, let's hear how it 257 00:16:50,720 --> 00:16:54,520 Speaker 1: works in practice. And so we've been mostly talking kind 258 00:16:54,520 --> 00:16:56,360 Speaker 1: of at the metal level about, you know, how to 259 00:16:56,400 --> 00:16:59,800 Speaker 1: think about AI ethics generally. But of course the way 260 00:16:59,800 --> 00:17:02,800 Speaker 1: this probably occurs in the trenches as a client approach 261 00:17:02,840 --> 00:17:05,080 Speaker 1: as IBM, and they want to help with a specific 262 00:17:05,119 --> 00:17:07,879 Speaker 1: problem in AI. And so I'm wondering, from a client 263 00:17:07,920 --> 00:17:10,560 Speaker 1: based perspective, where do you start having some of these 264 00:17:10,560 --> 00:17:14,840 Speaker 1: tough conversations. It has varied, to tell you the truth, 265 00:17:15,160 --> 00:17:20,000 Speaker 1: we had one client that approached us to expand the 266 00:17:20,080 --> 00:17:24,439 Speaker 1: use of an AI model to infer skill sets of 267 00:17:24,480 --> 00:17:28,160 Speaker 1: their employees, but not just to infer their technical skills 268 00:17:28,200 --> 00:17:32,720 Speaker 1: but also their soft foundational skills, meaning, let me use 269 00:17:32,760 --> 00:17:35,520 Speaker 1: an AI to determine what kind of communicator you might 270 00:17:35,520 --> 00:17:41,280 Speaker 1: be Laurie right. Others might come to us with, Okay, 271 00:17:41,400 --> 00:17:44,400 Speaker 1: we recognize we need help setting an AI ethics board. 272 00:17:44,640 --> 00:17:47,919 Speaker 1: Is this something you can assist us with? Or we 273 00:17:48,000 --> 00:17:52,560 Speaker 1: have these values, we need to establish AI ethics principles 274 00:17:52,720 --> 00:17:56,639 Speaker 1: and processes to help us ensure that we're compliant given 275 00:17:56,680 --> 00:18:00,600 Speaker 1: regulations coming down the pike. Or we've had clients come 276 00:18:00,640 --> 00:18:03,119 Speaker 1: to us saying, please train our people how to assess 277 00:18:03,880 --> 00:18:08,399 Speaker 1: for unexpected patterns in an AI model, but then also 278 00:18:09,160 --> 00:18:15,240 Speaker 1: how to holistically mitigate to prevent any potential harm. And 279 00:18:15,520 --> 00:18:21,560 Speaker 1: those have been phenomenal engagements. They're huge learning moments. And 280 00:18:21,600 --> 00:18:24,320 Speaker 1: so it seems like the real additional value that IBM 281 00:18:24,400 --> 00:18:27,480 Speaker 1: is bringing through this process isn't necessarily just providing an 282 00:18:27,480 --> 00:18:30,600 Speaker 1: AI algorithm or consulting on sam AI algorithm. It seems 283 00:18:30,640 --> 00:18:34,439 Speaker 1: like the real value added is explaining how this design 284 00:18:34,480 --> 00:18:37,679 Speaker 1: thinking works. You're almost like this therapist or like a 285 00:18:37,720 --> 00:18:40,400 Speaker 1: really good bartender who talks to people, who talks whole 286 00:18:40,440 --> 00:18:43,040 Speaker 1: companies through some of their problems to try to figure 287 00:18:43,080 --> 00:18:46,080 Speaker 1: out where they're going astray before they start implementing these things. 288 00:18:46,960 --> 00:18:52,720 Speaker 1: Can I put Chief Bartender Office on my metaphor, I'll 289 00:18:52,720 --> 00:18:56,239 Speaker 1: tell you some of our our most valuable people on 290 00:18:56,280 --> 00:19:01,080 Speaker 1: the team for that engagement. We had an industrial organization psychologist, 291 00:19:01,480 --> 00:19:06,000 Speaker 1: we had an anthropologist. That's why I'm saying it's important 292 00:19:06,080 --> 00:19:09,280 Speaker 1: we bring in the social scientists because you're exactly right, 293 00:19:09,960 --> 00:19:15,480 Speaker 1: it's more than just scrutinizing the algorithm in its state. 294 00:19:15,720 --> 00:19:18,040 Speaker 1: You have to be thinking about how is it being 295 00:19:18,160 --> 00:19:21,399 Speaker 1: used holistically? And so if I was a business that 296 00:19:21,520 --> 00:19:24,080 Speaker 1: was trying to think about how a company like IBM 297 00:19:24,080 --> 00:19:26,880 Speaker 1: could come in and help out with more trustworthy AI, 298 00:19:27,040 --> 00:19:30,320 Speaker 1: what would this process really look like. Well, what we're 299 00:19:30,359 --> 00:19:33,840 Speaker 1: finding more often than not is that there'll be smaller 300 00:19:33,920 --> 00:19:38,960 Speaker 1: teams within broader organizations that either have the responsibility of 301 00:19:39,119 --> 00:19:43,200 Speaker 1: compliance and see the writing on the wall, or they've 302 00:19:43,200 --> 00:19:46,919 Speaker 1: been the ones investing in AI and are trying to 303 00:19:46,960 --> 00:19:50,040 Speaker 1: figure out how to get the rest of the organization 304 00:19:50,359 --> 00:19:53,600 Speaker 1: on board with respect to things like setting up an 305 00:19:53,600 --> 00:19:57,600 Speaker 1: ethics board or establishing principles or things like that. So 306 00:19:58,440 --> 00:20:01,679 Speaker 1: some things that we've done help companies do this is 307 00:20:01,720 --> 00:20:05,840 Speaker 1: we kick off engagements with what we called our our 308 00:20:05,880 --> 00:20:10,520 Speaker 1: AI for leaders workshops. On the one hand, it's teaching 309 00:20:10,640 --> 00:20:13,600 Speaker 1: why you should care, but on the other hand, it's 310 00:20:13,640 --> 00:20:16,439 Speaker 1: meant to get people so excited across the organization that 311 00:20:16,480 --> 00:20:18,280 Speaker 1: they want to raise their hand and say, I want 312 00:20:18,280 --> 00:20:20,920 Speaker 1: to represent this part, like, for example, I want to 313 00:20:20,960 --> 00:20:22,960 Speaker 1: be part of the ethics board as it is being 314 00:20:23,000 --> 00:20:26,520 Speaker 1: stood up. The heart parts, not the tech. The hard 315 00:20:26,600 --> 00:20:28,560 Speaker 1: part is human behavior. And I know I'm preaching to 316 00:20:28,600 --> 00:20:31,520 Speaker 1: the choir given your background, it's so nice as a 317 00:20:31,560 --> 00:20:34,240 Speaker 1: psychologist to hear this. I'm like snapping my fingers like 318 00:20:34,320 --> 00:20:38,640 Speaker 1: peach exactly. The hard part is human behavior. So it's 319 00:20:38,680 --> 00:20:42,199 Speaker 1: been like drinking from a fire hose. I mean in 320 00:20:42,280 --> 00:20:45,200 Speaker 1: terms of the kinds of things that that we've all 321 00:20:45,280 --> 00:20:48,480 Speaker 1: been learning, and there's still so much to learn. It 322 00:20:49,240 --> 00:20:53,439 Speaker 1: really bugs me that those who are lucky enough to 323 00:20:53,480 --> 00:20:56,159 Speaker 1: be able to take classes and things like data ethics 324 00:20:56,280 --> 00:21:00,439 Speaker 1: or AI ethics self categorized as coders machine learning dissert 325 00:21:00,520 --> 00:21:03,920 Speaker 1: data scientists. If we're living in a world where AI 326 00:21:04,119 --> 00:21:07,359 Speaker 1: is fundamentally being used to make decisions that could directly 327 00:21:07,400 --> 00:21:11,360 Speaker 1: affect our livelihoods. We need to know more, We need 328 00:21:11,440 --> 00:21:16,040 Speaker 1: to have more literacy, and also make sure that there 329 00:21:16,160 --> 00:21:21,359 Speaker 1: is a consistent message of accessibility such that we are 330 00:21:21,400 --> 00:21:24,400 Speaker 1: saying you don't just have to be interested in coding, 331 00:21:24,680 --> 00:21:28,240 Speaker 1: like you're interested in social justice or psychology or anthropologies. 332 00:21:28,720 --> 00:21:31,480 Speaker 1: There's a seat at the table for you here because 333 00:21:31,520 --> 00:21:35,000 Speaker 1: we desperately need you. We desperately need that kind of 334 00:21:35,040 --> 00:21:39,840 Speaker 1: skill set. Just getting people to think about how do 335 00:21:39,880 --> 00:21:44,879 Speaker 1: you design something given an empathy lens to protect people? 336 00:21:44,920 --> 00:21:47,200 Speaker 1: I mean that, I think is such a crucial skill 337 00:21:47,240 --> 00:21:50,000 Speaker 1: to learn. You know, one thing I love about your 338 00:21:50,040 --> 00:21:52,560 Speaker 1: approaches that when you're talking to clients, you're almost doing 339 00:21:52,600 --> 00:21:54,840 Speaker 1: what I'm doing is a professor, where you're kind of 340 00:21:54,920 --> 00:21:57,919 Speaker 1: instructing students, getting them to think in different ways. But 341 00:21:58,000 --> 00:22:00,000 Speaker 1: I know from my field that I wind up learning 342 00:22:00,320 --> 00:22:02,920 Speaker 1: as much from students as I think sometimes they learned 343 00:22:03,000 --> 00:22:05,920 Speaker 1: from me. And so I'm wondering what what you've learned 344 00:22:05,920 --> 00:22:08,879 Speaker 1: in the process of helping so many businesses approach AI 345 00:22:08,960 --> 00:22:11,399 Speaker 1: a little bit more ethically, Like, have there been insights 346 00:22:11,400 --> 00:22:13,760 Speaker 1: that you've gotten through your interaction with clients and the 347 00:22:13,840 --> 00:22:20,080 Speaker 1: challenges they've been facing. I'm learning with every single interaction. 348 00:22:20,320 --> 00:22:26,840 Speaker 1: For example, in my mind, given the experiences that IBM 349 00:22:26,960 --> 00:22:31,000 Speaker 1: has had with respect to setting up our principles are 350 00:22:31,160 --> 00:22:36,240 Speaker 1: pillars ARII, ethics board. There's a process to follow, right 351 00:22:36,480 --> 00:22:38,080 Speaker 1: if you're thinking about it like a book, these are 352 00:22:38,080 --> 00:22:43,320 Speaker 1: the chapters in order to to optimize the approach, let's say, 353 00:22:43,359 --> 00:22:45,879 Speaker 1: but sometimes we work with clients that say, I'm going 354 00:22:45,920 --> 00:22:48,119 Speaker 1: to install this tool and I want to jump to 355 00:22:48,200 --> 00:22:52,320 Speaker 1: chapter seven, and it's like, okay, you know, how how 356 00:22:52,400 --> 00:22:55,760 Speaker 1: do we help navigate clients that want to skip over 357 00:22:56,960 --> 00:23:00,679 Speaker 1: steps that we think are important. Another on is again 358 00:23:01,240 --> 00:23:05,879 Speaker 1: the social scientists and bringing them into really push hard 359 00:23:06,040 --> 00:23:08,919 Speaker 1: on what is the right context of where this data 360 00:23:09,000 --> 00:23:12,680 Speaker 1: tell me the origin story? Again like really pushing us 361 00:23:12,680 --> 00:23:17,520 Speaker 1: to think hard and with their perspective, you don't know, 362 00:23:17,640 --> 00:23:21,359 Speaker 1: just constant, constant learning. Which is why one of the 363 00:23:21,400 --> 00:23:24,400 Speaker 1: things we did at IBM is we've established something called 364 00:23:24,440 --> 00:23:27,200 Speaker 1: our Center of Excellence where we said, you know what 365 00:23:27,280 --> 00:23:30,000 Speaker 1: IBM or we don't care what your background is, we 366 00:23:30,040 --> 00:23:32,640 Speaker 1: don't care who you are. If you're interested in this space, 367 00:23:33,119 --> 00:23:36,359 Speaker 1: you can become a member. The Center of Excellence is 368 00:23:36,400 --> 00:23:40,119 Speaker 1: a way in which we have not only projects people 369 00:23:40,119 --> 00:23:42,679 Speaker 1: can join in order to get real life experience, but 370 00:23:42,720 --> 00:23:45,880 Speaker 1: then also share back. Here's what we learned. We did 371 00:23:45,920 --> 00:23:48,879 Speaker 1: this with this particular I had. Here was our epiphany, 372 00:23:48,920 --> 00:23:53,400 Speaker 1: because if we're not sharing back and we're not constantly educating, 373 00:23:54,200 --> 00:23:58,440 Speaker 1: then we're missing the opportunity to establish the right culture. 374 00:24:00,200 --> 00:24:04,560 Speaker 1: Establishing the right culture to share what we're learning is 375 00:24:04,600 --> 00:24:08,080 Speaker 1: so important. And so I wanted to end. But going 376 00:24:08,119 --> 00:24:10,800 Speaker 1: back to where we started, you with your technofile family 377 00:24:10,960 --> 00:24:13,320 Speaker 1: watching Star Trek. I think if we were to fast 378 00:24:13,359 --> 00:24:16,200 Speaker 1: forward a couple of decades, we probably couldn't have imagined 379 00:24:16,240 --> 00:24:18,960 Speaker 1: that we'd be in the place with AI generally where 380 00:24:19,000 --> 00:24:20,920 Speaker 1: we are now, and especially as we think through more 381 00:24:20,960 --> 00:24:25,320 Speaker 1: trustworthy AI. And so you know, with such change happening 382 00:24:25,480 --> 00:24:27,639 Speaker 1: right now, with the fact that it's a fire hose 383 00:24:27,720 --> 00:24:30,760 Speaker 1: that's gonna just get even more powerful over time, what 384 00:24:30,800 --> 00:24:32,679 Speaker 1: do you think is next in this world of thinking 385 00:24:32,680 --> 00:24:37,040 Speaker 1: through more trustworthy AI. I would say next is far 386 00:24:37,200 --> 00:24:41,800 Speaker 1: more education, far more understanding, and we're starting to see 387 00:24:41,840 --> 00:24:45,879 Speaker 1: that shift far more CEO saying, yeah, ethics has to 388 00:24:45,920 --> 00:24:48,399 Speaker 1: be corridor our business. There's that, but there's a shift 389 00:24:48,680 --> 00:24:52,800 Speaker 1: barely half of the CEO is in we're saying that 390 00:24:53,280 --> 00:24:56,879 Speaker 1: a ethics was key or important to their business, and 391 00:24:56,920 --> 00:25:02,680 Speaker 1: now you're saying the great majority so education, education, education, 392 00:25:02,880 --> 00:25:06,360 Speaker 1: And again I would underscore making it far more accessible 393 00:25:06,400 --> 00:25:10,600 Speaker 1: to far more people, which means it's not just our 394 00:25:10,720 --> 00:25:16,000 Speaker 1: classes and higher ed institutions, it's our conferences, it's anytime 395 00:25:16,040 --> 00:25:19,960 Speaker 1: we write white papers, anytime we publish articles, anytime we 396 00:25:20,040 --> 00:25:24,199 Speaker 1: do podcasts like this. Right, the way we talk about 397 00:25:24,240 --> 00:25:27,320 Speaker 1: this space has to be far more accessible and open 398 00:25:27,400 --> 00:25:31,840 Speaker 1: and inviting two people with different roles, different skill sets, 399 00:25:31,880 --> 00:25:37,520 Speaker 1: different worldviews, because else again we're just codifying our own bias. Well, Fature, 400 00:25:37,560 --> 00:25:40,280 Speaker 1: I want to express my gratitude today for making AI 401 00:25:40,320 --> 00:25:43,120 Speaker 1: a little bit more accessible to everyone. This has been 402 00:25:43,119 --> 00:25:45,679 Speaker 1: such a delightful conversation. Thank you so much for joining 403 00:25:45,680 --> 00:25:48,320 Speaker 1: me for it. The pleasure was mine. Looie, thank you 404 00:25:48,359 --> 00:25:56,960 Speaker 1: for being the consummate host. I want to close by 405 00:25:57,000 --> 00:25:59,480 Speaker 1: going back to that moment when Lorie suggested that Phedra 406 00:25:59,840 --> 00:26:04,600 Speaker 1: was actually IBM's Chief Bartender Officer, not just because that's 407 00:26:04,640 --> 00:26:07,680 Speaker 1: the best C suite title ever, but because it gets 408 00:26:07,680 --> 00:26:10,960 Speaker 1: at what I think is the biggest, most important idea 409 00:26:11,280 --> 00:26:14,520 Speaker 1: in today's episode, pedro Boiled it down into a single 410 00:26:14,600 --> 00:26:17,760 Speaker 1: line when she said, the hard part is not the tech, 411 00:26:18,200 --> 00:26:22,679 Speaker 1: the hard part is human behavior. Why is building AI 412 00:26:22,840 --> 00:26:27,840 Speaker 1: so complicated? Because people are complicated. IBM believes that building 413 00:26:27,880 --> 00:26:32,480 Speaker 1: trust into AI from the start can lead to better outcomes, 414 00:26:32,520 --> 00:26:35,679 Speaker 1: and that to build trustworthy AI, you don't just need 415 00:26:35,760 --> 00:26:38,480 Speaker 1: to think like a computer scientist. You need to think 416 00:26:38,680 --> 00:26:44,879 Speaker 1: like a psychologist, like an anthropologist, You need to understand people. 417 00:26:47,920 --> 00:26:51,800 Speaker 1: Smart Talks of IBM is produced by Molly Sosha, Alexandra Garratton, 418 00:26:52,200 --> 00:26:57,119 Speaker 1: Royston Reserve, and Edith Russolo with Jacob Goldstein. We're edited 419 00:26:57,240 --> 00:27:01,159 Speaker 1: by Jen Guerra. Our engineers are Jason gam Brell, Sarah 420 00:27:01,160 --> 00:27:06,880 Speaker 1: Brugre and Ben Holliday. Theme song by Grandmascope. Special thanks 421 00:27:06,920 --> 00:27:10,920 Speaker 1: to Carli Migliore, Andy Kelly, Kathy Callaghan and the eight 422 00:27:11,000 --> 00:27:15,960 Speaker 1: Bar and IBM teams, as well as the Pushkin marketing team. 423 00:27:16,119 --> 00:27:18,960 Speaker 1: Smart Talks with IBM is a production of Pushkin Industries 424 00:27:19,160 --> 00:27:23,280 Speaker 1: and I Heart Media. To find more Pushkin podcasts, listen 425 00:27:23,359 --> 00:27:27,240 Speaker 1: on the I Heart Radio app, Apple Podcasts, or wherever 426 00:27:27,760 --> 00:27:32,359 Speaker 1: you listen to podcasts. I'm Malcolm Gladwell. This is a 427 00:27:32,400 --> 00:27:40,840 Speaker 1: paid advertisement from IBM.