WEBVTT - Smart Talks with IBM: Building Trustworthy AI: A Holistic Approach 0:00:04.400 --> 0:00:07.800 Welcome to tech Stuff, a production from I Heart Radio. 0:00:11.840 --> 0:00:14.480 This season of Smart Talks with IBM is all about 0:00:14.600 --> 0:00:18.480 new creators, the developers, data scientists, c t o s 0:00:18.560 --> 0:00:23.280 and other visionaries creatively applying technology in business to drive change. 0:00:23.800 --> 0:00:26.680 They use their knowledge and creativity to develop better ways 0:00:26.680 --> 0:00:30.280 of working, no matter the industry. Join hosts from your 0:00:30.320 --> 0:00:33.880 favorite pushkin industries podcasts as they use their expertise to 0:00:33.960 --> 0:00:37.480 deepen these conversations, and of course Malcolm Gladwell will guide 0:00:37.520 --> 0:00:39.839 you through the season as your host and provide his 0:00:39.920 --> 0:00:43.000 thoughts and analysis along the way. Look out for new 0:00:43.040 --> 0:00:45.480 episodes of Smart Talks with IBM on the I Heart 0:00:45.560 --> 0:00:49.159 Radio app, Apple Podcasts, or wherever you get your podcasts, 0:00:49.360 --> 0:00:56.520 and learn more at IBM dot com slash smart talks. Hello, Hello, 0:00:56.640 --> 0:01:00.760 Welcome to Smart Talks with IBM, a podcast from Industries 0:01:01.000 --> 0:01:05.680 Ighart Radio and IBM. I'm Malcolm Glabbo. This season, we're 0:01:05.680 --> 0:01:10.080 talking to new creators, the developers, data scientists, c t 0:01:10.240 --> 0:01:14.000 o s and other visionaries who are creatively applying technology 0:01:14.000 --> 0:01:18.199 and business to drive change. Channeling their knowledge and expertise, 0:01:18.400 --> 0:01:23.480 they're developing more creative and effective solutions, no matter the industry. 0:01:23.959 --> 0:01:28.440 Our guest today is Padre Bonadius, trust in AI practice 0:01:28.520 --> 0:01:34.120 leader within IBM Consulting. Advocating for artificial intelligence built and 0:01:34.200 --> 0:01:38.920 deployed responsibly is no longer just a compliance issue, but 0:01:39.080 --> 0:01:43.080 a business imperative. Part of Phedre's job is to help 0:01:43.120 --> 0:01:47.920 companies identify potential risks and pitfalls way before any code 0:01:47.960 --> 0:01:51.400 is written. In today's show, you'll hear how Phaeder's team 0:01:51.400 --> 0:01:56.560 and IBM is approaching this challenge holistically and creatively. Phedre 0:01:56.640 --> 0:02:00.200 spoke with Dr Loris Santos, host of the Pushkin podcast 0:02:00.480 --> 0:02:03.960 The Happiness Lab. Laurie is a professor of psychology at 0:02:04.000 --> 0:02:07.560 Yale University and an expert on human cognition and the 0:02:07.600 --> 0:02:13.760 cognitive biases that impede better choices. Now let's get to 0:02:13.760 --> 0:02:22.640 the interview, Padre. I'm so excited that we get a 0:02:22.680 --> 0:02:25.440 chance to chat today. You know, just to start off, 0:02:25.480 --> 0:02:28.160 I'm wondering how did you get started in this role 0:02:28.200 --> 0:02:30.400 at IBM, Like, what's the story to how you got 0:02:30.400 --> 0:02:33.760 where you are today? Oh goodness. My background is actually 0:02:34.400 --> 0:02:37.880 from the world of video games for entertainment, so AI 0:02:38.040 --> 0:02:40.880 has always been very interesting to me, especially when you 0:02:40.919 --> 0:02:45.040 intersect AI and play but several years ago, I began 0:02:45.080 --> 0:02:50.120 to get very frustrated by what I was reading in 0:02:50.160 --> 0:02:56.160 the news with respect to malintent through the use of AI. 0:02:56.639 --> 0:02:59.600 And the more that I learned and the more that 0:02:59.639 --> 0:03:03.800 I studied about this space of AI and ethics, the 0:03:03.840 --> 0:03:08.440 more I recognize that even organizations that have the very 0:03:08.760 --> 0:03:15.840 very best of intentions could inadvertently cause potential harm. And 0:03:15.880 --> 0:03:18.160 so that's super cool. I love that your interest in 0:03:18.360 --> 0:03:21.920 more responsible AI came from the gaming world. You have 0:03:22.000 --> 0:03:24.080 to talk a little bit about your history with gaming 0:03:24.120 --> 0:03:27.520 and that how that informed your interest in trustworthy AI. Well, 0:03:27.880 --> 0:03:33.040 it wasn't as much necessarily the ethical components of AI 0:03:33.080 --> 0:03:36.320 when I was working in games. It was more things like, 0:03:37.240 --> 0:03:41.280 look at what nonplayer characters can do, you know, I mean, 0:03:41.360 --> 0:03:43.720 if you've got an AI acting as a character within 0:03:43.760 --> 0:03:46.320 the game, and how is it that you can use 0:03:46.360 --> 0:03:50.880 AI in order to make a game a more interesting experience. Actually, 0:03:50.880 --> 0:03:53.800 I ended up joining IBM to be our first global 0:03:53.880 --> 0:03:56.240 lead for something called serious games, which is when you 0:03:56.320 --> 0:03:59.119 use video games to do something other than just entertaining, 0:03:59.720 --> 0:04:01.839 and of the idea of integrating real data and real 0:04:01.840 --> 0:04:07.120 processes within sophisticated games powered by AI to solve complex problems. 0:04:07.720 --> 0:04:11.200 It wasn't until, as I mentioned, like later, when we 0:04:11.280 --> 0:04:14.400 started to hear all of us more and more news 0:04:14.440 --> 0:04:19.360 about just problems. What could happen with respect to rendering 0:04:19.440 --> 0:04:22.800 or putting out models that are inaccurate or unfair. I 0:04:23.080 --> 0:04:25.359 know one of your inspirations for hearing other interviews that 0:04:25.400 --> 0:04:27.960 you've done is sci Fi. I'm also a sci Fi nerd, 0:04:28.040 --> 0:04:30.680 and I know sci Fi has talked a lot about, 0:04:30.920 --> 0:04:33.599 you know, the trustworthiness issues that come up when we're 0:04:33.640 --> 0:04:35.840 dealing with AI and so on, and so talk a 0:04:35.920 --> 0:04:38.000 little bit about how you bring that to your work 0:04:38.080 --> 0:04:40.960 in developing AI. That's a little bit more ethical. A 0:04:41.000 --> 0:04:45.240 lovely question. So, my my parents were major techno files. 0:04:45.440 --> 0:04:48.560 They both were immigrants to the United States, came here 0:04:48.560 --> 0:04:53.320 to study engineering, and they met uh in college. Growing up, 0:04:53.360 --> 0:04:59.960 my sister and I, we had Star Trek playing every night. UH. 0:05:00.040 --> 0:05:03.880 My parents were both big fans of Gene Roddenberry's vision 0:05:04.160 --> 0:05:09.360 of how technology could really be used to help better humankind, 0:05:09.600 --> 0:05:12.360 and that was the ethos that, of course we grew 0:05:12.440 --> 0:05:16.800 up in. The wonderful thing about science fiction isn't that 0:05:16.920 --> 0:05:21.279 it predicts cars, for example, but that it predicts traffic jams, 0:05:22.200 --> 0:05:24.880 you know, and I think there's just so much we 0:05:24.960 --> 0:05:28.520 can learn from science fiction or in fact, like I said, 0:05:28.760 --> 0:05:32.520 play as a mechanism to be able to teach science 0:05:32.560 --> 0:05:37.159 fiction predicting traffic jams. I love it. But when we 0:05:37.240 --> 0:05:40.680 think about AI and science fiction, we need to be careful. 0:05:41.279 --> 0:05:44.360 We need to remember that AI is not something that's 0:05:44.360 --> 0:05:46.440 going to enter our lives at some point in the 0:05:46.520 --> 0:05:51.800 distant future. AI is something that's all around us today. 0:05:52.120 --> 0:05:55.839 If you have a virtual assistant in your house, that's AI, 0:05:56.120 --> 0:05:59.960 your phone app that predicts traffic AI. What a streaming 0:06:00.160 --> 0:06:04.880 service recommends a movie? You've guessed it AI, Paeder says. 0:06:04.960 --> 0:06:09.039 AI maybe behind the scenes determining the interest rate on 0:06:09.120 --> 0:06:12.040 your loan, or even whether or not you're the right 0:06:12.080 --> 0:06:15.760 candidate for that job you applied for. AI is both 0:06:15.920 --> 0:06:20.160 ubiquitous and invisible, which is why it is so crucial 0:06:20.360 --> 0:06:24.719 the companies learn how to build trustworthy AI. How do 0:06:24.800 --> 0:06:27.960 we do that? When thinking about what does it take 0:06:28.080 --> 0:06:32.479 to earn trust in something like an AI, there are 0:06:32.560 --> 0:06:36.760 fundamentally human centric questions to be asked, right like what 0:06:36.920 --> 0:06:40.320 is the intent of this particular AI model? How accurate 0:06:40.520 --> 0:06:44.640 is that model? How fair is it Is it explainable 0:06:44.720 --> 0:06:48.040 if it makes a decision that could directly affect my livelihood? 0:06:48.760 --> 0:06:51.640 Can I inquire what data did you use about me 0:06:51.920 --> 0:06:56.040 to make this decision? Is it protecting my data? Is 0:06:56.080 --> 0:07:00.359 it robust? Is it protected against people who could trick 0:07:00.440 --> 0:07:03.839 it to disadvantage me over others? I mean, there's so 0:07:03.880 --> 0:07:08.360 many questions to be asked. Earning trust in something like 0:07:08.400 --> 0:07:13.880 AI is fundamentally not a technological challenge but a socio 0:07:13.880 --> 0:07:19.200 technological challenge. It can't just be solved with a tool alone. 0:07:20.520 --> 0:07:22.520 What are the kinds of risks that companies have to 0:07:22.560 --> 0:07:25.360 think through? Is they're developing these technologies to make sure 0:07:25.360 --> 0:07:28.160 they're as trustworthy as possible. Well, you know, they may 0:07:28.200 --> 0:07:31.920 be putting a lot of money into investing in AI 0:07:32.360 --> 0:07:35.320 that gets stuck in proof of concept planned like get's 0:07:35.320 --> 0:07:37.640 get stuck in pilot. We we've done some research where 0:07:37.640 --> 0:07:42.080 we have found about eight of investments in AI get stuck. 0:07:42.760 --> 0:07:46.640 And sometimes it's because the investment isn't tied directly to 0:07:46.680 --> 0:07:49.760 a business strategy, or more often than not, people simply 0:07:49.800 --> 0:07:53.960 don't trust the results of the AI model. As a company, 0:07:54.000 --> 0:07:56.360 who is of course thinking about this so deeply? What 0:07:56.480 --> 0:07:59.200 a businesses need to consider when they're trying to figure out, 0:07:59.400 --> 0:08:02.200 you know, how to solve this big puzzle of AI ethics. 0:08:02.480 --> 0:08:05.280 It has to be approached holistically, So you've got to 0:08:05.280 --> 0:08:10.280 be thinking about, for example, what culture is required within 0:08:10.320 --> 0:08:13.600 your organization in order to really be able to responsibly 0:08:13.680 --> 0:08:17.320 create AI, what processes are in place to make sure 0:08:17.400 --> 0:08:20.880 that you're being compliant and that your your practitioners know 0:08:21.040 --> 0:08:25.360 what to do. And then of course AI engineering frameworks 0:08:25.400 --> 0:08:29.080 and tooling that can assist you on this journey. There 0:08:29.160 --> 0:08:33.560 is so much fundamentally to do. We found that actually 0:08:33.720 --> 0:08:37.840 those that were leading responsible AI trust where the AI 0:08:37.920 --> 0:08:41.800 initiatives within their organization has switched in the last three years. 0:08:42.320 --> 0:08:46.120 It used to be technical leaders, for example, chief data 0:08:46.200 --> 0:08:50.120 officer or someone who is a PhD in machine learning, 0:08:50.640 --> 0:08:53.600 and now it's switched to be a d percent of 0:08:53.640 --> 0:08:58.079 those leaders are now non technical business leaders maybe you know, 0:08:58.200 --> 0:09:03.760 chief compliance officer, diversely inclusivity officers, chief legal officer. So 0:09:03.840 --> 0:09:07.240 we're seeing a shift, and I believe firmly it's a 0:09:07.280 --> 0:09:12.400 recognition from organizations that are seeing that in order to 0:09:12.480 --> 0:09:15.199 really pull this off well, there has to be an 0:09:15.240 --> 0:09:20.600 investment than a focus in culture in people and getting 0:09:20.600 --> 0:09:23.920 people to understand why they should care about this space. 0:09:25.720 --> 0:09:28.520 And so I see two challenges with doing that right. 0:09:28.679 --> 0:09:31.280 One is, you know a lot of these technology companies 0:09:31.320 --> 0:09:34.560 are really built to be tech companies, not necessarily you know, 0:09:35.080 --> 0:09:37.880 social tech companies or having this sort of training and 0:09:38.040 --> 0:09:41.000 ethics and beyond. Another issue seems to be that you're 0:09:41.040 --> 0:09:44.520 really proposing a switch that's truly holistic, right, that's like 0:09:44.640 --> 0:09:48.120 rethinking the way the company thinks about its bottom line. 0:09:48.280 --> 0:09:50.880 And so as you think about working through these kinds 0:09:50.880 --> 0:09:53.400 of challenges at IBM, how have you tackled this, like 0:09:53.440 --> 0:09:55.480 how have you brought new talent in? How have you 0:09:55.559 --> 0:09:58.440 thought really carefully about this big holistic switch that needs 0:09:58.440 --> 0:10:01.160 to come to make AIM more trustworth be Data is 0:10:01.200 --> 0:10:04.680 an artifact of the human experience. And if you start 0:10:04.760 --> 0:10:08.120 with that as your definition and then think about, well 0:10:09.200 --> 0:10:13.079 data is curated by data sideists, all data is biased. 0:10:13.760 --> 0:10:19.200 And so if you're not recognizing bias with eyes fully open, 0:10:19.720 --> 0:10:25.640 then ultimately you're calcifying systemic bias into systems like AI. 0:10:26.080 --> 0:10:28.480 So some of the things that we've done at IBM, 0:10:28.520 --> 0:10:32.720 again recognizing this important need for culture is big, big, 0:10:32.760 --> 0:10:37.160 big focus on diversity, not only looking at teams of 0:10:37.240 --> 0:10:40.160 data scientists and saying how many women are on this team, 0:10:40.280 --> 0:10:45.080 how many minorities are on this team, but also insisting 0:10:45.200 --> 0:10:48.680 on recognizing that we need to bring in people with 0:10:48.760 --> 0:10:53.240 different world views too, for example, what's your definition of fairness? 0:10:54.040 --> 0:10:57.480 Is your definition equality or is it equity? Also bringing 0:10:57.559 --> 0:11:01.560 people with a wider variety of skill sets and roles, 0:11:01.920 --> 0:11:08.640 including our social scientists, anthropologists, sociologist, psychologists like yourself, right, 0:11:09.360 --> 0:11:13.160 behavioral scientists, designers. I mean we have one of the 0:11:13.360 --> 0:11:18.800 leading AI design practices in the world. I mean the effort, 0:11:18.880 --> 0:11:22.560 the investments we've been making in design thinking as a 0:11:22.679 --> 0:11:27.960 mechanism to create frameworks for systemic empathy well before any 0:11:28.040 --> 0:11:32.280 code is written, so people can think through how would 0:11:32.320 --> 0:11:36.200 you design in order to mitigate for any potential harm 0:11:36.240 --> 0:11:39.480 given not only the values of your organization, but what 0:11:39.520 --> 0:11:43.720 are the rights of individuals asking oneself? These kinds of 0:11:43.800 --> 0:11:48.400 questions reinforces then the idea the ethics doesn't come at 0:11:48.400 --> 0:11:51.640 the end, like it's some kind of quality assurance, like 0:11:51.840 --> 0:11:54.600 check I passed the audit, I've got to go, you know, 0:11:55.120 --> 0:11:57.360 But instead, really, you know, as soon as you're thinking 0:11:57.400 --> 0:12:01.440 about using an AI for a particular use case thinking 0:12:01.480 --> 0:12:05.160 about you know, what is the intent of this model, 0:12:05.320 --> 0:12:08.560 what's the relationship we ultimately want to have with AI? 0:12:09.240 --> 0:12:13.400 And again, these are non technology questions. This is where 0:12:13.440 --> 0:12:18.160 social scientists. Having a social scientist on your team helping 0:12:18.280 --> 0:12:22.520 think through these kinds of questions is is critical. Let's 0:12:22.520 --> 0:12:25.160 pause here for a second, because this is a really 0:12:25.160 --> 0:12:29.920 profound idea. Building responsible AI does not mean that you 0:12:30.000 --> 0:12:32.240 create a system then check in at the end and 0:12:32.320 --> 0:12:36.480 say is this okay? Is this ethical? If you don't 0:12:36.520 --> 0:12:39.679 ask those questions until the end of the process, you've 0:12:39.720 --> 0:12:43.480 already failed. You have to think about ethics from the 0:12:43.559 --> 0:12:46.600 jump from the makeup of the team to the data 0:12:46.640 --> 0:12:49.120 you're using to train the model to the most basic 0:12:49.240 --> 0:12:52.079 question of all, is this even the right use case 0:12:52.520 --> 0:12:56.960 for artificial intelligence? The big lesson from IBM is this 0:12:57.640 --> 0:13:02.280 responsible AI is something you build at every step of 0:13:02.320 --> 0:13:05.839 the process. So this season of smart Talk is all 0:13:05.920 --> 0:13:09.280 focused on creativity and business. My guess is that thinking 0:13:09.280 --> 0:13:12.560 about trustworthy AI involves a lot of creativity. But talk 0:13:12.640 --> 0:13:14.400 to me about some of the spots where you see 0:13:14.400 --> 0:13:18.120 this work as being most creative. Oh goodness, I would 0:13:18.120 --> 0:13:24.280 say incorporating design design thinking in particular as well as 0:13:24.280 --> 0:13:29.079 straight up design in order to craft AI responsibly. You've 0:13:29.160 --> 0:13:32.080 used this word design thinking, and so I'm wondering exactly 0:13:32.080 --> 0:13:33.920 what you mean here. How do you define this idea 0:13:33.960 --> 0:13:37.800 of design thinking. Design thinking is a practice that we 0:13:37.920 --> 0:13:41.120 established here at IBM many years ago. In essence, what 0:13:41.240 --> 0:13:45.560 it is, it's a way of working with groups of 0:13:45.600 --> 0:13:51.440 people to co create a vision for something, for a 0:13:51.480 --> 0:13:56.600 product or a service or an outcome. And typically it 0:13:56.760 --> 0:14:00.319 starts with things like, for example, empathy maps. If you're 0:14:00.320 --> 0:14:03.640 thinking about an end user, thinking through what is this 0:14:03.720 --> 0:14:08.760 person thinking, seeing, hearing, feeling like, what are the experiencing 0:14:09.400 --> 0:14:12.600 in order to ultimately craft and experience for them that 0:14:12.840 --> 0:14:16.800 is targeted specifically for them. So we use it in 0:14:16.920 --> 0:14:21.040 a really wide variety of different ways with respect to 0:14:21.040 --> 0:14:26.680 trustworthy AI, even rendering an AI model explainable to a subject. 0:14:26.680 --> 0:14:29.200 And I'll give you an example. So we've got this 0:14:29.280 --> 0:14:33.840 wonderful program with an IBM caller, our Academy of Technology, 0:14:33.840 --> 0:14:37.760 and we take on initiatives that steer the company in 0:14:37.760 --> 0:14:41.720 innovative new directions. So we had an initiative where it 0:14:41.800 --> 0:14:46.000 was titled what the Titanic taught Us About explainable AI, 0:14:46.920 --> 0:14:52.360 and the project was imagining if there was an AI 0:14:52.480 --> 0:14:57.400 model that could predict the likelihood of a passenger getting 0:14:57.440 --> 0:15:00.400 a life raft on the Titanic. And we broke up 0:15:00.440 --> 0:15:03.360 into two work streams. One was the workstream full of 0:15:03.400 --> 0:15:06.720 the data scientists who were using all the different explainers 0:15:06.760 --> 0:15:08.600 to come up with the predictions and they would crank 0:15:08.600 --> 0:15:12.280 out the numbers. And the other team here's where the 0:15:12.360 --> 0:15:16.000 social scientists lived and the designers were right where we 0:15:16.000 --> 0:15:20.480 were thinking through how do we empower people? How do 0:15:20.560 --> 0:15:27.000 we explain this algorithm and this predictor and the accuracy 0:15:27.080 --> 0:15:30.320 behind this prediction in such a way as to ultimately 0:15:30.440 --> 0:15:33.760 empower and end users. They could decide I'm not getting 0:15:33.840 --> 0:15:37.960 on that boat, or I want to get a second 0:15:37.960 --> 0:15:43.080 opinion please, or I want to contest the outputs of 0:15:43.080 --> 0:15:47.320 this model because I upgraded to first class just yesterday. 0:15:47.880 --> 0:15:51.160 See what I'm saying. And that takes a lot of creativity. 0:15:52.040 --> 0:15:55.480 How do you design an experience for someone in order 0:15:55.520 --> 0:16:01.440 to ultimately empower them? So design designed as is critically 0:16:01.560 --> 0:16:04.200 critically important. And why I mentioned you know, we we've 0:16:04.240 --> 0:16:06.640 got to open up the aperture with respect to who 0:16:06.680 --> 0:16:09.160 we invite to the table and these kinds of conversations. 0:16:10.000 --> 0:16:14.000 Taking the time to really understand other people's perspectives is 0:16:14.040 --> 0:16:17.800 so important when you're doing anything creative, and it is 0:16:18.120 --> 0:16:22.120 fundamental to the way the new creators work. The core 0:16:22.240 --> 0:16:25.360 question you should always be asking is where will the 0:16:25.480 --> 0:16:29.680 user be meeting this product? As Peder said, what will 0:16:29.760 --> 0:16:33.680 they be thinking, seeing, hearing, feeling. If you can answer 0:16:33.760 --> 0:16:37.640 those questions the way IBM does in its design thinking practice, 0:16:38.080 --> 0:16:41.880 you will be in great shape to create almost anything. Really, 0:16:42.600 --> 0:16:46.080 let's hear how it works in practice. And so we've 0:16:46.080 --> 0:16:48.480 been mostly talking kind of at the metal level about 0:16:48.520 --> 0:16:51.720 you know, how to think about AI ethics generally, but 0:16:51.800 --> 0:16:54.240 of course the way this probably occurs in the trenches 0:16:54.320 --> 0:16:56.920 as a client approach as IBM and they want help 0:16:56.920 --> 0:16:59.800 with the specific problem in AI. And so I'm wondering, 0:16:59.800 --> 0:17:02.760 from a client based perspective, where do you start having 0:17:02.800 --> 0:17:06.880 some of these tough conversations. It has varied, to tell 0:17:06.880 --> 0:17:11.320 you the truth, We had one client that approached us 0:17:11.359 --> 0:17:15.800 to expand the use of an AI model to infer 0:17:16.040 --> 0:17:19.200 skill sets of their employees, but not just to infer 0:17:19.280 --> 0:17:24.159 their technical skills, but also their soft foundational skills, meaning 0:17:25.000 --> 0:17:26.760 let me use an AI to determine what kind of 0:17:26.760 --> 0:17:31.359 communicator you might be A Laurie right. Others might come 0:17:31.400 --> 0:17:35.919 to us with, Okay, we recognize we need help setting 0:17:35.920 --> 0:17:38.640 an AI ethics board. Is this something you can assist 0:17:38.760 --> 0:17:42.720 us with? Or we have these values, we need to 0:17:42.880 --> 0:17:48.000 establish AI ethics principles and processes to help us ensure 0:17:48.040 --> 0:17:51.919 that we're compliant given regulations coming down the pike. Or 0:17:52.280 --> 0:17:54.439 we've had clients come to us saying, please train our 0:17:54.480 --> 0:17:59.919 people how to assess for unexpected patterns in an aim 0:18:00.000 --> 0:18:04.800 at all, but then also how to holistically mitigate to 0:18:04.920 --> 0:18:11.480 prevent any potential harm. And those have been phenomenal engagements. 0:18:12.119 --> 0:18:15.000 They're huge learning moments. And so it seems like the 0:18:15.119 --> 0:18:18.440 real additional value that IBM is bringing through this process 0:18:18.520 --> 0:18:21.639 isn't necessarily just providing an AI algorithm or consulting on 0:18:21.680 --> 0:18:24.840 sam AI algorithm. It seems like the real value added 0:18:25.200 --> 0:18:28.520 is explaining how this design thinking works. You're almost like 0:18:28.600 --> 0:18:31.919 this therapist or like a really good bartender who talks 0:18:31.920 --> 0:18:34.240 to people, who talks whole companies through some of their 0:18:34.240 --> 0:18:36.879 problems to try to figure out where they're going astray 0:18:36.960 --> 0:18:40.600 before they start implementing. These things can I put Chief 0:18:40.680 --> 0:18:45.560 Bartender Office on my I like the metaphor. I'll tell 0:18:45.600 --> 0:18:49.040 you some of our our most valuable people on the 0:18:49.080 --> 0:18:53.720 team for that engagement. We had an industrial organizational psychologist, 0:18:54.119 --> 0:18:58.680 we had an anthropologist. That's why I'm saying it's important 0:18:58.520 --> 0:19:01.920 that we bring in the social scientists because you're exactly right, 0:19:02.600 --> 0:19:08.119 it's more than just scrutinizing the algorithm in its state. 0:19:08.359 --> 0:19:10.640 You have to be thinking about how is it being 0:19:10.800 --> 0:19:14.040 used holistically? And so if I was a business that 0:19:14.119 --> 0:19:16.680 was trying to think about how a company like IBM 0:19:16.720 --> 0:19:19.520 could come in and help out with more trustworthy AI, 0:19:19.680 --> 0:19:22.960 what would this process really look like. Well, what we're 0:19:22.960 --> 0:19:26.480 finding more often than not is that there'll be smaller 0:19:26.520 --> 0:19:31.600 teams within broader organizations that either have the responsibility of 0:19:31.760 --> 0:19:35.679 compliance and see the writing on the wall, or they've 0:19:35.840 --> 0:19:39.520 been the ones investing in AI and are trying to 0:19:39.600 --> 0:19:42.680 figure out how to get the rest of the organization 0:19:43.000 --> 0:19:46.240 on board with respect to things like setting up an 0:19:46.240 --> 0:19:50.240 ethics board or establishing principles or things like that. So 0:19:51.080 --> 0:19:54.040 some things that we've done to help companies do this 0:19:54.200 --> 0:19:57.959 is we kick off engagements with what we called our 0:19:58.160 --> 0:20:02.400 our AI for leaders workshops On the one hand, it's 0:20:02.680 --> 0:20:05.960 teaching why you should care, But on the other hand, 0:20:06.040 --> 0:20:08.960 it's meant to get people so excited across the organization 0:20:09.000 --> 0:20:10.679 that they want to raise their hand and say, I 0:20:10.720 --> 0:20:13.480 want to represent this part, like, for example, I want 0:20:13.520 --> 0:20:15.320 to be part of the ethics board as it is 0:20:15.359 --> 0:20:18.919 being stood up. The heart parts, not the tech. The 0:20:19.000 --> 0:20:21.119 hard part is human behavior. And I know I'm preaching 0:20:21.119 --> 0:20:24.080 to the choir given your background, it's so nice as 0:20:24.080 --> 0:20:26.760 a psychologist to hear this. I'm like snapping my fingers, 0:20:26.800 --> 0:20:30.360 like preach exactly. The hard part is human behavior. So 0:20:31.040 --> 0:20:34.720 it's been like drinking from a fire hose. I mean 0:20:34.760 --> 0:20:37.640 in terms of the kinds of things that that we've 0:20:37.680 --> 0:20:40.280 all been learning, and there's still so much to learn. 0:20:41.040 --> 0:20:45.359 It really bugs me that those who are lucky enough 0:20:46.000 --> 0:20:48.440 to be able to take classes and things like data 0:20:48.480 --> 0:20:52.960 ethics or AI ethics self categorized as coders, machine learning scientists, 0:20:53.000 --> 0:20:55.640 or data scientists. If we're living in a world where 0:20:55.640 --> 0:20:59.640 AI is fundamentally being used to make decisions that could 0:20:59.680 --> 0:21:03.760 directly affect our livelihoods, we need to know more. We 0:21:03.840 --> 0:21:08.280 need to have more literacy and also make sure that 0:21:08.440 --> 0:21:13.560 there is a consistent message of accessibility such that we 0:21:13.640 --> 0:21:17.040 are saying you don't just have to be interested in coding, 0:21:17.320 --> 0:21:20.840 like you're interested in social justice or psychology or anthropologies. 0:21:21.359 --> 0:21:24.119 There's a seat at the table for you here because 0:21:24.160 --> 0:21:27.639 we desperately need you. We desperately need that kind of 0:21:27.680 --> 0:21:32.480 skill set. Just getting people to think about how do 0:21:32.520 --> 0:21:37.520 you design something given an empathy lens to protect people? 0:21:37.560 --> 0:21:39.840 I mean that, I think is such a crucial skill 0:21:39.880 --> 0:21:42.640 to learn. You know, one thing I love about your 0:21:42.640 --> 0:21:45.200 approaches that when you're talking to clients, you're almost doing 0:21:45.240 --> 0:21:47.520 what I'm doing is a professor, where you're kind of 0:21:47.560 --> 0:21:50.560 instructing students, getting them to think in different ways. But 0:21:50.640 --> 0:21:52.840 I know from my field that I wind up learning 0:21:52.960 --> 0:21:55.560 as much from students as I think sometimes they learned 0:21:55.640 --> 0:21:58.560 from me. And so I'm wondering what what you've learned 0:21:58.560 --> 0:22:01.520 in the process of helping so many businesses approach AI 0:22:01.600 --> 0:22:04.040 a little bit more ethically, Like, have there been insights 0:22:04.040 --> 0:22:06.399 that you've gotten through your interaction with clients and the 0:22:06.480 --> 0:22:12.640 challenges they've been facing. I'm learning with every single interaction. 0:22:12.960 --> 0:22:19.480 For example, in my mind, given the experiences that IBM 0:22:19.560 --> 0:22:23.640 has had with respect to setting up our principles are 0:22:23.800 --> 0:22:28.880 pillars Arii Ethics Board. There's a process to follow, right 0:22:29.119 --> 0:22:30.720 if you're thinking about it like a book, these are 0:22:30.720 --> 0:22:35.960 the chapters in order to to optimize the approach. Let's say, 0:22:36.000 --> 0:22:38.640 but sometimes we work with clients that say I'm gonna 0:22:38.640 --> 0:22:41.600 install this tool and I want to jump to chapter seven, 0:22:42.800 --> 0:22:45.320 and it's like, okay, you know, how how do we 0:22:45.400 --> 0:22:50.119 help navigate clients that want to skip over steps that 0:22:50.200 --> 0:22:54.480 we think are important. Another one is again the social 0:22:54.600 --> 0:22:59.159 scientists and bringing them into really push hard on what 0:22:59.359 --> 0:23:02.119 is the right context where this data tell me the 0:23:02.119 --> 0:23:06.000 origin story? Again like really pushing us to think hard 0:23:06.040 --> 0:23:12.399 and with their perspective, you don't know, just constant, constant learning, 0:23:12.440 --> 0:23:14.720 which is why one of the things we did at 0:23:14.720 --> 0:23:18.359 IBM is we've established something called our Center of Excellence, 0:23:18.800 --> 0:23:21.280 where we said, you know what IBM or we don't 0:23:21.280 --> 0:23:23.560 care what your background is, we don't care who you are. 0:23:23.680 --> 0:23:27.000 If you're interested in this space, you can become a member. 0:23:27.720 --> 0:23:30.199 The Center of Excellence is a way in which we 0:23:30.320 --> 0:23:33.720 have not only projects people can join in order to 0:23:33.760 --> 0:23:37.080 get real life experience, but then also share back here's 0:23:37.119 --> 0:23:39.760 what we learned. We did this with this particular and 0:23:39.880 --> 0:23:43.400 here was our epiphany, because if we're not sharing back 0:23:43.440 --> 0:23:49.200 and we're not constantly educating, then we're missing the opportunity 0:23:49.280 --> 0:23:54.679 to establish the right culture. Establishing the right culture to 0:23:54.960 --> 0:23:59.760 share what we're learning is so important. So I wanted 0:23:59.760 --> 0:24:02.119 to But going back to where we started, you with 0:24:02.200 --> 0:24:05.359 your technofile family watching Star Trek, I think if we 0:24:05.359 --> 0:24:07.920 were to fast forward a couple of decades, we probably 0:24:07.920 --> 0:24:10.080 couldn't have imagined that we'd be in the place with 0:24:10.119 --> 0:24:13.000 AI generally where we are now, and especially as we 0:24:13.000 --> 0:24:16.200 think through more trustworthy AI. And so you know, with 0:24:16.440 --> 0:24:19.520 such change happening right now, with the fact that it's 0:24:19.560 --> 0:24:22.320 a fire hose that's gonna just get even more powerful 0:24:22.359 --> 0:24:24.439 over time, what do you think is next in this 0:24:24.480 --> 0:24:28.120 world of thinking through more trustworthy AI. I would say 0:24:28.240 --> 0:24:33.360 next is far more education, far more understanding. And we're 0:24:33.440 --> 0:24:37.600 starting to see that shift, far more CEO saying yeah, 0:24:37.720 --> 0:24:40.320 ethics has to be corrid or business. There's that, but 0:24:40.359 --> 0:24:44.720 there's a shift. Barely half of the CEO is we're 0:24:44.760 --> 0:24:49.120 saying that a ethics was key or important to their business. 0:24:49.400 --> 0:24:55.320 And now you're saying the great majority so education, education, education, 0:24:55.520 --> 0:24:59.000 And again I would underscore making it far more accessible 0:24:59.040 --> 0:25:03.200 to far more people, which means it's not just our 0:25:03.359 --> 0:25:08.639 classes in higher ed institutions, it's our conferences, it's anytime 0:25:08.640 --> 0:25:12.600 we write white papers, anytime we publish articles, anytime we 0:25:12.640 --> 0:25:16.840 do podcasts like this. Right the way we talk about 0:25:16.880 --> 0:25:19.960 this space has to be far more accessible and open 0:25:20.040 --> 0:25:24.480 and inviting two people with different roles, different skill sets, 0:25:24.520 --> 0:25:30.159 different worldviews, because else again, we're just codifying our own bias. Well, feature, 0:25:30.200 --> 0:25:32.880 I want to express my gratitude today for making AI 0:25:32.960 --> 0:25:35.760 a little bit more accessible to everyone. This has been 0:25:35.760 --> 0:25:38.320 such a delightful conversation. Thank you so much for joining 0:25:38.320 --> 0:25:40.960 me for it. The pleasure was mine. Looie, thank you 0:25:41.000 --> 0:25:49.119 for being the consummate host. Thank you. I want to 0:25:49.119 --> 0:25:51.600