WEBVTT - Misuse of Data is Solvable 0:00:15.076 --> 0:00:24.356 Pushkin. I'm Maybe Higgins, and this is solvable Interviews with 0:00:24.436 --> 0:00:27.876 the world's most innovative thinkers who are working to solve 0:00:27.956 --> 0:00:33.436 the world's biggest problems. My solvable is that every frontline 0:00:33.556 --> 0:00:36.796 social organization is the ability to use data and AI 0:00:37.196 --> 0:00:40.876 same way, same capacity that the big tech companies do today. 0:00:41.036 --> 0:00:43.196 I want to see a world where the same algorithms 0:00:43.196 --> 0:00:45.876 that are routing your packages to you your house coming 0:00:45.916 --> 0:00:48.076 so efficiently because an AI figured out the best way 0:00:48.116 --> 0:00:50.676 to avoid traffic and weather are just as equally being 0:00:50.716 --> 0:00:54.436 applied to delivering a vaccine through an area before it spoils. 0:00:55.036 --> 0:00:59.036 That's Jake poor Away, the founder and CEO of nonprofit 0:00:59.156 --> 0:01:02.796 Data Kind. He's talking to Jacob Weisberg about how he's 0:01:02.876 --> 0:01:07.636 working to make that world a reality. The Rockefeller Foundation 0:01:07.756 --> 0:01:10.716 has thought about this too. More than two point five 0:01:10.916 --> 0:01:15.196 quintillion bytes of data are produced every day. That's one 0:01:15.316 --> 0:01:20.196 hundred trillion bytes. This abundance of data, combined with rapidly 0:01:20.276 --> 0:01:25.396 advancing analytics capabilities, could really improve the lives of billions 0:01:25.436 --> 0:01:28.916 of people around the world, but it's only living up 0:01:28.996 --> 0:01:33.316 to a fraction of that potential. While private sector businesses 0:01:33.356 --> 0:01:36.916 have been building and deploying data science capabilities for many 0:01:36.996 --> 0:01:41.236 years now. Most organizations in the nonprofit and civic and 0:01:41.356 --> 0:01:45.716 public sectors are way behind. Of course, they want to 0:01:45.836 --> 0:01:48.436 use the applied data to make their work go farther 0:01:48.596 --> 0:01:51.956 and faster and to help more people, but they don't 0:01:52.036 --> 0:01:56.156 often have the resources. I mean, put yourself in the 0:01:56.196 --> 0:02:00.156 shoes of a newly minted graduate. They're probably wearing tivas. 0:02:00.836 --> 0:02:04.436 Last year, the San Francisco Chronicle analyze glass door data 0:02:04.476 --> 0:02:07.316 of the starting salaries of some of the biggest tech 0:02:07.356 --> 0:02:10.836 companies in the Bay Area. They found out that tech 0:02:11.036 --> 0:02:15.356 pays even for the young and inexperienced. The average starting 0:02:15.356 --> 0:02:19.556 salary for a software engineer was almost ninety two thousand dollars. 0:02:20.516 --> 0:02:23.956 So there's the workers and then there's the technology itself. 0:02:24.596 --> 0:02:27.356 We know the power data science can have for social 0:02:27.436 --> 0:02:30.836 good because we've seen it in action. When mission driven 0:02:30.916 --> 0:02:34.756 organizations have the right talent and tools and knowledge, data 0:02:34.756 --> 0:02:39.636 science can generate real human impact, helping vulnerable families access 0:02:39.676 --> 0:02:44.316 public benefits, saving water and money during droughts, and saving 0:02:44.356 --> 0:02:48.236 time in resettling refugees so that they can find homes 0:02:48.276 --> 0:02:53.516 and jobs faster. Jake Borway works on this stuff every day. 0:02:53.916 --> 0:02:57.356 He's a machine learning and technology enthusiast who loves nothing 0:02:57.396 --> 0:03:00.996 more than seeing good values in data. In twenty eleven, 0:03:01.036 --> 0:03:04.756 he found a data Kind, bringing together leading data scientists 0:03:04.756 --> 0:03:09.436 with high impact social organizations to better collect, analyze, and 0:03:09.596 --> 0:03:13.716 visualized data in the service of humanity. Jake works to 0:03:13.996 --> 0:03:18.436 ensure organizations like the Red Cross have access to AI 0:03:18.516 --> 0:03:21.796 and data science that's as good as the access enjoyed 0:03:21.836 --> 0:03:25.796 by huge companies like Facebook. Data Kind has twenty thousand 0:03:25.836 --> 0:03:29.196 volunteers around the world, who he likens to mets on 0:03:29.276 --> 0:03:33.476 San Frontier, the doctors without Borders, except their data scientists 0:03:33.636 --> 0:03:37.956 working pro bono with leading social change organizations on all 0:03:38.076 --> 0:03:41.996 kinds of projects, including one that has data scientists from 0:03:42.036 --> 0:03:47.196 Netflix predicting water usage in a California neighborhood. It's fascinating, 0:03:47.316 --> 0:03:50.156 So enjoy this conversation and I'll talk to you after. 0:03:55.716 --> 0:03:59.076 What's the problem. In a nutshell, the problem is that 0:03:59.316 --> 0:04:04.436 digital technology and artificial intelligence have exploded over the last 0:04:04.476 --> 0:04:08.516 ten or fifteen years, which have created huge opportunities in 0:04:08.676 --> 0:04:11.916 the corporate space or in building new apps for society, 0:04:12.356 --> 0:04:17.436 but there's very little application of that to social sector causes. 0:04:17.756 --> 0:04:20.636 So we have this huge opportunity to use a revolutionary 0:04:20.676 --> 0:04:24.796 technology to predict the future of things, to understand our 0:04:24.836 --> 0:04:27.396 society better, to automate things that we either don't want 0:04:27.396 --> 0:04:30.596 to or couldn't do, And yet there's a huge potential 0:04:30.676 --> 0:04:33.156 loss in that it's very difficult to get that applied 0:04:33.356 --> 0:04:36.436 to pro social causes that we need. Jake is a 0:04:36.556 --> 0:04:40.076 data scientist. When did you start to see some of 0:04:40.116 --> 0:04:44.476 the downsides around big data? Really? The article that I 0:04:44.556 --> 0:04:46.676 used to point to is like the beginnings of the 0:04:47.276 --> 0:04:50.716 tide turning to the negative. Was the article that was 0:04:50.756 --> 0:04:54.756 titled very salaciously, Target Knows You're pregnant, And if you 0:04:54.796 --> 0:04:57.636 remember this one from twenty thirteen, but the basic idea 0:04:57.756 --> 0:05:01.396 was that someone had their daughter, that maybe sixteen seventeen 0:05:01.436 --> 0:05:04.516 year old daughter was receiving mailers from Target that said, Hey, 0:05:04.596 --> 0:05:07.716 we think you need to buy kupons for baby diapers 0:05:07.996 --> 0:05:11.396 or formula, and the dad called up, you know Target, all, Matt, 0:05:11.436 --> 0:05:13.236 So what are you sending me all my daughter all 0:05:13.316 --> 0:05:16.116 these deals for having babies. She's not pregnant, Like, why 0:05:16.116 --> 0:05:18.956 are you trying to get her to become pregnant? And 0:05:19.236 --> 0:05:20.596 the person on the other end of the line, of 0:05:20.596 --> 0:05:22.876 course didn't know what was happening, because you know, the 0:05:22.916 --> 0:05:25.316 algorithms just send you what they think you're going to 0:05:25.396 --> 0:05:28.636 buy based on other stuff you've bought, and it's He 0:05:28.716 --> 0:05:30.636 called back later, kind of shame facedly and said, you know, 0:05:30.716 --> 0:05:32.956 I talked to my daughter and actually she is pregnant, 0:05:33.476 --> 0:05:35.156 and you know, the data had picked up on that 0:05:35.196 --> 0:05:36.996 simply because you know, it watched what she bought and 0:05:36.996 --> 0:05:40.396 she was probably buying you know, prenatal care, vitamins and stuff. 0:05:40.836 --> 0:05:44.596 But that article got shared around as the sign that 0:05:44.676 --> 0:05:48.436 big data was going to be negative. Target knows you're pregnant. 0:05:48.516 --> 0:05:50.996 What a horrible invasion of privacy. That title alone should, 0:05:51.036 --> 0:05:54.076 you know, make everyone's skin crawl. But that's the problem 0:05:54.196 --> 0:05:56.476 is that that shouldn't be the case. We think of 0:05:56.916 --> 0:05:59.876 there are so many opportunities to be using data and 0:05:59.956 --> 0:06:04.236 algorithms to see where disease outbreaks are going to occur 0:06:04.516 --> 0:06:06.476 or predict in the same way as what kind of 0:06:06.796 --> 0:06:09.276 conditions you might have so you can live a healthier life. 0:06:09.476 --> 0:06:12.516 And so I think it was then that we really thought, Okay, 0:06:12.636 --> 0:06:15.796 we need to come out and show the positive sides 0:06:15.796 --> 0:06:17.996 of this. Otherwise everyone's going to just run to the 0:06:18.036 --> 0:06:22.756 fear around what data science can do. We're interested on 0:06:22.796 --> 0:06:26.076 this podcast and people who've taken this leap to become 0:06:26.276 --> 0:06:29.876 problem solvers and to take on the biggest problems in 0:06:29.916 --> 0:06:34.636 the world. What made you take a leap to leave 0:06:34.716 --> 0:06:40.356 the private sector to start an organization with an ambitious goal. Well, 0:06:40.396 --> 0:06:41.876 I have to say it was a bit of an accident. 0:06:41.956 --> 0:06:45.516 Actually it was maybe twenty ten or eleven, and I 0:06:45.556 --> 0:06:48.716 had just coincidentally come out of school with a computer 0:06:48.796 --> 0:06:51.476 science and a statistics degree, which little did I know 0:06:51.596 --> 0:06:53.636 was going to become what would lead to the title 0:06:53.716 --> 0:06:56.436 data scientist. And I was working at the New York 0:06:56.436 --> 0:07:00.116 Times R and D Lab, and really what seemed obvious 0:07:00.396 --> 0:07:02.316 was the fact that we had all of this new 0:07:02.356 --> 0:07:05.756 digital technology, from cell phones that people were carrying around 0:07:05.756 --> 0:07:09.396 with them, to satellites launching in the air, to sends 0:07:09.436 --> 0:07:12.156 being put around the world, that we were digitizing our 0:07:12.476 --> 0:07:15.996 very existence. We were becoming a digital species. There was 0:07:16.036 --> 0:07:18.796 almost like a central nervous system to the world, and 0:07:18.876 --> 0:07:21.716 that meant that were these huge opportunities to learn from 0:07:21.756 --> 0:07:25.556 that to you know, have algorithms drive maybe our greatest 0:07:25.636 --> 0:07:28.236 human values. But the folks who really knew how to 0:07:28.236 --> 0:07:32.396 convert data into those actions. The data scientists were largely 0:07:32.476 --> 0:07:35.716 locked up in tech companies, and you know, I would 0:07:35.836 --> 0:07:38.956 actually go to hackathons, which are you know, like weekend 0:07:38.956 --> 0:07:41.796 events where technologists would get together and just work on 0:07:41.836 --> 0:07:44.196 whatever they thought was cool. And I would sit there 0:07:44.196 --> 0:07:46.116 and think, this is so interesting because you know, we're 0:07:46.116 --> 0:07:48.556 not at a company, we're not at our jobs. We're 0:07:48.596 --> 0:07:50.516 here on the weekend. You know, I'm sitting next to 0:07:50.556 --> 0:07:53.676 some machine learning engineer from Google and NASA scientist, and 0:07:53.756 --> 0:07:56.556 I'm like, this is great. We can make whatever we want. 0:07:56.636 --> 0:07:59.956 Like the world has just become so ripe for what's possible. 0:08:00.556 --> 0:08:01.956 And at the end of the day, the stuff that 0:08:01.996 --> 0:08:05.956 people made was just so unfulfilling. You know that someone 0:08:05.996 --> 0:08:08.796 had made like Twitter for pets, or had improved how 0:08:08.836 --> 0:08:12.356 you'd find local deals in your neighborhood, and so I 0:08:12.396 --> 0:08:14.396 just said, man, there's got to be something more we 0:08:14.436 --> 0:08:17.516 can do for society, or something more fulfilling really than this, 0:08:17.676 --> 0:08:20.676 as opposed to solving the problems of very well paid 0:08:20.756 --> 0:08:23.796 twenty somethings in the Bay Area, right, which is the parody, 0:08:23.836 --> 0:08:26.236 but that is a lot of the new companies you 0:08:26.316 --> 0:08:29.476 hear about are solving problems like how do you get 0:08:29.516 --> 0:08:32.476 your food delivered or god knows how to get cannabis delivered? 0:08:32.596 --> 0:08:34.476 You know when you when you could already buy it 0:08:34.516 --> 0:08:37.236 by walking around the corner. You're exactly right. We solve 0:08:37.276 --> 0:08:39.796 the problems that we ourselves have. And as you've pointed out, 0:08:39.836 --> 0:08:45.196 the tech community for better for worse, excused young male US. So, yeah, 0:08:45.196 --> 0:08:46.396 I just thought, you know, what would it take for 0:08:46.516 --> 0:08:48.396 to be applied to the social sector. Where are the 0:08:48.436 --> 0:08:50.476 people who are on the front lines of getting people 0:08:50.516 --> 0:08:52.836 food or clean water? And how could you apply it there? 0:08:53.276 --> 0:08:57.036 And so I just wanted that job myself. What didn't exist? 0:08:57.636 --> 0:08:59.516 So I just wrote to a couple of folks in 0:08:59.516 --> 0:09:02.196 the community here in New York and said, hey, you know, 0:09:02.476 --> 0:09:06.076 instead of going and building you know, a door dash competitor, 0:09:06.676 --> 0:09:09.116 could we, I don't know, work with the Red Cross 0:09:09.196 --> 0:09:12.996 US or Kiva who goes cash transfers to folks, and 0:09:13.036 --> 0:09:14.796 say what could we do with their data? What could 0:09:14.796 --> 0:09:16.636 we learn? What are the positive ways we could work 0:09:16.636 --> 0:09:19.516 together with them? And I thought people would just say, yeah, 0:09:19.596 --> 0:09:22.276 good idea, Jake, but no thanks. I kind of just 0:09:22.396 --> 0:09:26.676 buried the little sign up link for folks, and I 0:09:26.716 --> 0:09:29.236 was surprised to find that people started sharing around before 0:09:29.236 --> 0:09:31.356 I knew it. I came back to work the next time, 0:09:31.596 --> 0:09:33.996 hundreds of emails in my inbox from people not just 0:09:34.036 --> 0:09:36.316 in the city but around the world saying, though this 0:09:36.396 --> 0:09:38.556 is great, I want to get involved with data kind, 0:09:38.596 --> 0:09:41.516 I want to do data kind France. At one point, 0:09:41.556 --> 0:09:44.036 a few months into this, the White House called and said, hey, 0:09:44.036 --> 0:09:46.396 we're interested in big data initiatives. What's this thing? And 0:09:46.636 --> 0:09:50.556 you know, joke because I don't know, it's not really 0:09:50.596 --> 0:09:53.876 a thing, But to me it really tapped into an 0:09:53.996 --> 0:09:56.996 energy from both the technology side and the nonprofits and 0:09:57.076 --> 0:10:01.356 governments who are writing, who said, we're energetic to take 0:10:01.356 --> 0:10:03.796 on this new wave of this technology and figure out 0:10:03.796 --> 0:10:06.196 how could be applied. And so our job ever since 0:10:06.276 --> 0:10:09.556 has really just been trying to support that community, harness 0:10:09.556 --> 0:10:11.676 its energy, and be helpful in any way we can. 0:10:12.316 --> 0:10:15.516 Since you've been doing this, it's amazing how quickly attitudes 0:10:15.556 --> 0:10:19.956 have shifted around big data and algorithms. I mean, just 0:10:19.996 --> 0:10:23.236 think about Facebook, which even a few years ago was 0:10:23.396 --> 0:10:27.556 thought as a socially positive company. That was why part 0:10:27.596 --> 0:10:30.836 of why people went to work there, and in just 0:10:30.916 --> 0:10:34.956 a couple of years it's become something that people think 0:10:35.116 --> 0:10:38.956 is an overwhelmingly negative force. Are we're swinging too far 0:10:39.076 --> 0:10:43.396 in the other direction in our skepticism about what data 0:10:43.516 --> 0:10:45.996 is going to be used for? Well, I think there's 0:10:46.036 --> 0:10:49.076 a healthy reckoning on how we've been using data and 0:10:49.076 --> 0:10:52.316 technology in the past. You're right that in the last 0:10:52.356 --> 0:10:55.196 couple of years there was sort of unfettered techno optimism 0:10:55.196 --> 0:10:57.516 amongst a lot of the big companies and that this 0:10:57.556 --> 0:11:00.676 would just change everything and nothing could ever go wrong 0:11:00.716 --> 0:11:03.636 with social media and data. So I think there is 0:11:03.676 --> 0:11:05.796 an obviously very healthy reckoning of this, and we're starting 0:11:05.796 --> 0:11:09.556 to realize what the downsides could be. What your point 0:11:09.596 --> 0:11:11.676 I think is missing and we really need to get 0:11:11.956 --> 0:11:15.596 acclimated to, is where do we go from there? You know, 0:11:15.796 --> 0:11:17.876 is the idea that we're just going to put the 0:11:17.916 --> 0:11:20.916 genie back in the bottle, not use digital information in 0:11:20.956 --> 0:11:25.676 these ways, regulate all companies into existence. I'm in favor of, 0:11:25.836 --> 0:11:28.956 by the way, stronger regulation, for sure, But I think 0:11:28.996 --> 0:11:32.316 what we need now is more examples and more of 0:11:32.356 --> 0:11:34.716 a community of practice around what it looks like to 0:11:34.796 --> 0:11:38.276 use these technologies ethically. That's a big conversation obviously, that's 0:11:38.276 --> 0:11:39.916 in the space right now. You hear a lot about 0:11:39.916 --> 0:11:42.996 the ethics of data use, ethics of AI, but even 0:11:42.996 --> 0:11:45.876 then I find those conversations fairly academic. I think what 0:11:45.916 --> 0:11:48.036 we need are some more positive examples of how it 0:11:48.076 --> 0:11:51.076 can be applied and positive principles that we all agree 0:11:51.076 --> 0:11:53.916 to adhere to. And so the data kind that's something 0:11:53.916 --> 0:11:58.156 we're really working to try to demonstrate, is to say, yes, 0:11:58.276 --> 0:12:02.956 we need to protect ourselves, uphold our civil liberties through data. 0:12:03.076 --> 0:12:05.716 Make sure that we're not degrading human life with what's 0:12:05.716 --> 0:12:09.316 going on with data in the business world? And what 0:12:09.356 --> 0:12:12.476 does it look like when you want to use data 0:12:12.516 --> 0:12:16.596 and algorithms to predict, say, inclement weather that could wipe 0:12:16.596 --> 0:12:20.236 out a crop and that's critical to someone's sustenance in 0:12:20.356 --> 0:12:22.836 another part of the world. What's the good version of this? 0:12:22.956 --> 0:12:25.036 You know? How do you make sure that it's accountable 0:12:25.076 --> 0:12:27.276 to those folks? How do we make sure that everyone 0:12:27.356 --> 0:12:30.196 involved has some sense of what the algorithm is doing 0:12:30.196 --> 0:12:32.516 and how their data is being used. And I don't 0:12:32.556 --> 0:12:34.716 think we can move past that point just by talking 0:12:34.716 --> 0:12:37.956 about it. I think we need real concrete examples of 0:12:38.356 --> 0:12:43.076 data scientists, nonprofits, social organizations, constituents getting together to say, 0:12:43.116 --> 0:12:45.276 what does the good version of this look like a 0:12:45.356 --> 0:12:48.116 better version. I should say there was a positive example 0:12:48.156 --> 0:12:52.076 in the news recently with the prediction of the cyclone 0:12:52.396 --> 0:12:57.156 in South Asia that killed very few people, and in 0:12:57.196 --> 0:13:00.956 the world before big data, that same storm might have 0:13:01.036 --> 0:13:04.836 killed a lot of people through panic, through all sorts 0:13:04.876 --> 0:13:07.516 of consequences because people wouldn't have known it was coming. 0:13:07.676 --> 0:13:09.796 I mean, is that the kind of example we're talking 0:13:09.836 --> 0:13:13.156 about here? Something positive? I think that's exactly right. So 0:13:13.756 --> 0:13:16.796 at data Kind we team technologists like data scientists who 0:13:16.836 --> 0:13:20.396 want to volunteer their time alongside social change organizations, be 0:13:20.476 --> 0:13:24.276 they government agencies or nonprofits who have a pro social mission, 0:13:24.596 --> 0:13:26.436 might be able to use data and algorithms to do 0:13:26.516 --> 0:13:29.956 even more, and we together they collaborate and kind of 0:13:29.996 --> 0:13:33.716 codesign the solutions that they might foster a better world. 0:13:34.116 --> 0:13:36.116 So some examples that we've seen are exactly what you're 0:13:36.156 --> 0:13:38.836 talking about. There was a project that a group did 0:13:38.836 --> 0:13:41.636 as a water district in California, and the problem they 0:13:41.676 --> 0:13:44.476 faced was when drought season comes, you know, it's really 0:13:44.476 --> 0:13:46.556 hard to get water to folks. People don't have water. 0:13:47.316 --> 0:13:51.556 That's obviously problematic. You need drinking water and water to bathe, etc. 0:13:52.356 --> 0:13:55.556 But more than that, the cost of not getting them 0:13:55.556 --> 0:13:58.356 water is really high because the only way that they 0:13:58.356 --> 0:13:59.996 can get water to the places they don't have it 0:14:00.076 --> 0:14:03.276 is to actually take a dump truck, drive it up 0:14:03.316 --> 0:14:06.516 to some other reservoir, maybe over to Nevada, literally fill 0:14:06.556 --> 0:14:09.556 it by hand and drive it back. So you're also 0:14:09.556 --> 0:14:13.436 facing like huge environmental costs, huge energy costs. So they 0:14:13.436 --> 0:14:15.716 ask the question, you know, could we figure out a 0:14:15.756 --> 0:14:18.196 way to predict how much water demand there's going to 0:14:18.276 --> 0:14:20.516 be at a more granular level so we can really 0:14:20.596 --> 0:14:23.956 understand and ration more effectively. And so we team them 0:14:23.996 --> 0:14:27.036 up with some data scientists that come from everywhere from 0:14:27.036 --> 0:14:31.596 Netflix to environmental science organizations, and together they collected the 0:14:31.676 --> 0:14:34.316 data at almost a block by block level, and they 0:14:34.396 --> 0:14:36.716 built an algorithm that sort of takes that data in 0:14:36.756 --> 0:14:39.916 and continually gives updates. Does water district to say, hey, 0:14:39.916 --> 0:14:41.596 this is how much we think people are going to use. 0:14:41.636 --> 0:14:43.836 Here's how much they've already used. Tomorrow, you're probably going 0:14:43.876 --> 0:14:45.756 to see this, And they said, in the first year 0:14:45.756 --> 0:14:48.236 of using this, they saved over twenty five million dollars 0:14:48.476 --> 0:14:50.876 in addition to getting water to people much more effectively. 0:14:51.516 --> 0:14:53.276 So I think when you hear about cases like that 0:14:53.716 --> 0:14:55.876 those are the kinds of examples that we want to 0:14:56.596 --> 0:14:58.996 kind of platform and see more even the world where 0:14:59.436 --> 0:15:03.556 within the confines of social organization these data and algorithms 0:15:03.556 --> 0:15:07.356 that can really drive real effectiveness. Now your people are 0:15:07.356 --> 0:15:11.116 all doing this for good. We've all heard about the 0:15:11.236 --> 0:15:14.676 kinds of bias issues that have started to turn up 0:15:14.716 --> 0:15:18.796 with predictive algorithms of different kinds, and they seem to 0:15:18.796 --> 0:15:24.236 get embedded just because of the inherited unconscious biases of 0:15:24.276 --> 0:15:27.076 the people who write the algorithm. Absolutely, how do you 0:15:27.116 --> 0:15:32.556 avoid recapitulating that problem again with the projects you're working on? 0:15:32.956 --> 0:15:35.796 Such an awesome question, and I think just to comment 0:15:35.836 --> 0:15:39.836 on the challenge generally, I think you really nailed it there. 0:15:39.876 --> 0:15:43.436 That the challenge that we face is that humans have 0:15:43.476 --> 0:15:48.236 been collecting data from our activities that incorporate unconscious bias, 0:15:48.316 --> 0:15:50.196 and so if you then have a machine learn from 0:15:50.196 --> 0:15:53.836 it or you analyze it, you write replicating that. So, 0:15:54.596 --> 0:15:56.636 while I will not admit that we have a perfect solution, 0:15:56.636 --> 0:15:58.796 because I mean we're sort of talking about the challenge 0:15:58.796 --> 0:16:02.436 of bias and humanity, some of the things that we 0:16:02.476 --> 0:16:05.676 really focus on is the technology to us that we're 0:16:05.676 --> 0:16:09.996 building is secondary to the outcome for people. So, for example, 0:16:10.076 --> 0:16:13.516 it's not exciting to us to build an algorithm that 0:16:13.956 --> 0:16:18.076 helps a homeless shelter triage people to the right homeless 0:16:18.116 --> 0:16:21.556 shelters correctly just because it's a cool algorithm. We only 0:16:21.596 --> 0:16:23.556 care if at the end of the day, the ultimate 0:16:23.596 --> 0:16:27.196 success metric that you know, a wide range of inclusive 0:16:27.236 --> 0:16:31.196 folks are getting housing is achieved. So I want to 0:16:31.196 --> 0:16:33.036 say that first because I think one of the reasons 0:16:33.036 --> 0:16:36.116 we see some of these biased challenges rise up is 0:16:36.116 --> 0:16:39.116 that folks say, hey, the algorithm is doing something. It's 0:16:39.116 --> 0:16:42.236 doing a thing I want, like giving out sentences in 0:16:42.396 --> 0:16:45.476 courts or you know, policing folks, but without a question 0:16:45.476 --> 0:16:48.676 of and how is it biased? Towards the end, you know, 0:16:48.716 --> 0:16:50.836 what's it achieving. But the other thing we do is 0:16:50.876 --> 0:16:54.356 we work extremely closely with our NGEO partners who are 0:16:54.436 --> 0:16:58.116 on the ground and who understand a lot of those challenges. 0:16:58.476 --> 0:17:00.516 And so we'll actually do what we call a pre 0:17:00.596 --> 0:17:03.036 mortem some other companies do, which is before we even 0:17:03.036 --> 0:17:05.116 start a project, we'll say, okay, let's pretend we jump 0:17:05.196 --> 0:17:08.716 to the end. Well, you know, basic questions like how 0:17:08.716 --> 0:17:11.276 will this be maintained, who's actually going to use this 0:17:11.276 --> 0:17:12.956 tool at the end of the day. But then we'll 0:17:12.996 --> 0:17:15.836 also ask two questions, which is one, what's the worst 0:17:15.876 --> 0:17:18.236 that happens if we fail? So if you're relying on 0:17:18.356 --> 0:17:21.436 us to build, this is not something we would necessarily build. 0:17:21.436 --> 0:17:23.676 But let's say someone said, hey, we want a tool 0:17:23.716 --> 0:17:26.396 that predicts whether you have cancer or not. Okay, well 0:17:26.436 --> 0:17:29.236 that's pretty serious. And if we don't succeed, are you 0:17:29.996 --> 0:17:32.316 stuck because you really needed that and now your organization 0:17:32.356 --> 0:17:34.756 can't proceed. That's important to know. But then we also 0:17:34.796 --> 0:17:37.516 ask what's the worst that happens if we succeed? So 0:17:38.196 --> 0:17:39.956 who is this going to affect? How would you know 0:17:40.036 --> 0:17:42.236 that it's wrong? Right? Like, how would you know just 0:17:42.236 --> 0:17:45.036 because it's chugging away making predictions? Is it doing the 0:17:45.116 --> 0:17:48.636 right thing? Is it disenfranchising certain groups? Could somebody use 0:17:48.636 --> 0:17:51.956 it to intentionally target people who have cancer? We ask 0:17:51.956 --> 0:17:54.356 a lot of those questions, and what's really important us 0:17:54.356 --> 0:17:58.156 in that questioning is who has the power and agency 0:17:58.236 --> 0:18:02.116 to both understand the algorithm and change the algorithm Because 0:18:02.156 --> 0:18:05.036 in the current landscape, when tech companies build algorithms, it's 0:18:05.076 --> 0:18:06.516 not much you can do. But you know, I don't 0:18:06.516 --> 0:18:10.116 have enough agency to know how Facebook's news feed algorithm works, 0:18:10.116 --> 0:18:13.156 nor can I really affect it much? But that's not 0:18:13.196 --> 0:18:16.116 acceptable to me when you're bringing algorithms into the public 0:18:16.396 --> 0:18:19.116 good space and this is actually affecting folks lives. So 0:18:19.156 --> 0:18:20.716 those are some of the questions we ask up front 0:18:20.756 --> 0:18:22.596 and really try to be rigorous with our partners around 0:18:22.676 --> 0:18:25.316 oversight of and oftentimes that's enough for us to not 0:18:25.396 --> 0:18:28.356 take on a project. It's great that you're thinking steps 0:18:28.356 --> 0:18:33.716 ahead about these projects, and your own solvable is, ironically, 0:18:33.756 --> 0:18:35.876 to put yourself out of business is to create a 0:18:35.916 --> 0:18:38.436 world in which you don't need a data kind to 0:18:38.476 --> 0:18:41.716 point people towards positive uses of data. That's right, What 0:18:41.716 --> 0:18:44.556 would it take to make that happen? And I guess 0:18:44.716 --> 0:18:48.556 playing your chess game. What happens when that happens. The 0:18:48.636 --> 0:18:51.116 day we close our doors is the data. Every frontline 0:18:51.236 --> 0:18:54.356 social change organization has the capabilities to use data and 0:18:54.396 --> 0:18:57.156 AI the same way the big tech companies do ethically 0:18:57.236 --> 0:19:00.636 and capably. And so you know, our little slice of 0:19:00.676 --> 0:19:03.796 that today is to bridge the gap in getting the 0:19:03.876 --> 0:19:06.836 human capital, the talent, the data scientists AI engineers to 0:19:07.036 --> 0:19:10.436 social organizations. That sort of step one is to show 0:19:10.436 --> 0:19:12.876 people the art of the possible and really get some 0:19:12.876 --> 0:19:15.076 of those challenges solved. But what do it take to 0:19:15.076 --> 0:19:16.636 do that? Long runs to think about what are the 0:19:16.676 --> 0:19:19.356 problems and hurdles we're trying to overcome with that model today, 0:19:19.756 --> 0:19:23.076 and they are that in the social sector there isn't 0:19:23.156 --> 0:19:25.516 enough awareness about what the technology could do or where 0:19:25.516 --> 0:19:27.476 it would be applied. So we have to start with that, 0:19:27.556 --> 0:19:30.956 and I think now increasingly you're seeing more of more 0:19:30.996 --> 0:19:35.276 folks understanding that, more companies talking about doing data and 0:19:35.396 --> 0:19:38.636 AI for good. So I feel like there's some progress there, 0:19:38.876 --> 0:19:40.796 But if you go further, you have to think, well, 0:19:40.796 --> 0:19:44.036 how would a government or nonprofit get access to these 0:19:44.076 --> 0:19:47.316 resources in the long term, And there I think there's 0:19:47.356 --> 0:19:50.036 going to be a long term shift in getting funding 0:19:50.316 --> 0:19:54.876 to move towards nonprofits for things like data science and AI. 0:19:55.036 --> 0:19:59.196 You're going to need maybe consultancies that actually provide this 0:19:59.316 --> 0:20:02.436 service in the social sector. There's lots of different models 0:20:02.436 --> 0:20:04.836 for where that capacity could come from, but I think 0:20:04.836 --> 0:20:06.836 the biggest things that we need right now are that 0:20:06.876 --> 0:20:09.476 awareness of how could be used and then the I say, 0:20:09.476 --> 0:20:12.276 the funding for ngox to be able to hire a 0:20:12.356 --> 0:20:15.916 data sciences and incorporate them into the work they do. Now. 0:20:16.356 --> 0:20:19.476 When that happens, what happens. Oh, I mean, I'd love 0:20:19.476 --> 0:20:23.676 to say that all challenges that are stymied by not 0:20:23.756 --> 0:20:27.196 having data science and AI are solved live apply ever after. 0:20:27.796 --> 0:20:31.116 But actually, what I think my most ambitious hope for 0:20:31.156 --> 0:20:34.516 the world is that we could actually tip the balance 0:20:34.556 --> 0:20:38.716 a little bit to where the social sector is paving 0:20:38.756 --> 0:20:42.036 the path for how machine learning and AI could be used. 0:20:42.516 --> 0:20:44.876 I think we're so built into this default model that 0:20:45.356 --> 0:20:48.996 business and wealthy countries set the agenda and everyone else 0:20:49.076 --> 0:20:52.676 kind of struggles to catch up and imitate. We're talking 0:20:52.796 --> 0:20:55.956 about a technology that is so fundamental to humanity because 0:20:55.996 --> 0:20:59.036 it relies on data about us. When we talk about AI, 0:20:59.156 --> 0:21:01.916 it is like automating human processes that I don't think 0:21:01.916 --> 0:21:04.276 that's something that should be just a business application that 0:21:04.396 --> 0:21:07.276 is ported to the world. There should be a place 0:21:07.636 --> 0:21:10.036 for us to say, what does it look like when 0:21:10.036 --> 0:21:12.516 we apply the technology to the better angels of our nature? 0:21:12.756 --> 0:21:16.036 What is human based AI? What are the things we 0:21:16.076 --> 0:21:18.316 care about? And I can't think of any other place 0:21:18.316 --> 0:21:21.356 besides the social sector whose sole mandate is to look 0:21:21.356 --> 0:21:24.756 out for humanity. So my dream is when you bridge 0:21:24.836 --> 0:21:27.476 that gap, when that's there. You could actually have this 0:21:27.596 --> 0:21:30.836 voice from the social sector itself saying what it looks 0:21:30.876 --> 0:21:33.236 like to have human based ai Jick. Do you think 0:21:33.276 --> 0:21:37.716 about the training of data scientists. I sometimes think we're 0:21:37.796 --> 0:21:43.996 just missing the intersection between moral philosophy and computer science. 0:21:44.076 --> 0:21:46.476 You know, the people who are majoring in college and 0:21:46.676 --> 0:21:50.396 electronic engineering aren't reading much Kant, and the people who 0:21:50.396 --> 0:21:54.756 are reading Kant don't understand much about computer programming, you know, 0:21:54.796 --> 0:21:57.476 And in a way, the problem is that the people 0:21:57.516 --> 0:22:00.716