Today’s guest speaker leads impact practices in the data for good movement at data client listen in as Jake Porway and I discuss how constructive uses of AI can create positive social outcomes. Why is it unfair to ask eight techs to solve community challenges? And, what impact practices are creating equitable outcomes in the USA? This is HumAIn.
Welcome to HumAIn, My name is David Yakobovitch, and I will be your host throughout this series together, we will explore AI through fireside conversations with industry experts, from business executives and AI researchers to leaders who advanced AI for all HumAIn is the channel to release new AI products to learn about industry trends and to bridge the gap between humans and machines in the fourth industrial revolution if you’d like this episode, remember to subscribe and leave a review.
Welcome back everyone to the HumAIn Podcast, today I have a very special guest who I had the opportunity to meet about four years ago at a hackathon for data for good in New York City. Today’s guest is Jake Porway. He’s an expert in the field of data and technology, and he’s the founder and executive director of Datacom. They run global, database events around data science and AI for good, for humanity and have been added since 2011. I had the great opportunity to meet Jake a few years ago and worked on one of these cool David James so Jake, thanks so much for being with us.
Thank you so much, Dave, it’s a pleasure to be here and you’re being a little Bit Humble because you were one of our data ambassadors who actually helped scope and set up a project with the L.A mayor’s office so that other technologists could work on it and that is no small role as huge so really glad that you got to join us that way.
One of the best ways for anyone, if you’re a new student in data science or you’re someone who’s a veteran to the field to constantly learn and to be part of teams, is to give back your time. I have to do that with data kind of volunteering and that was a great project the one thing I never mentioned I was really cool is then I was up as dinner a few weeks later and just so happened that someone from a mayor’s office in Los Angeles was there and I was talking about the topic and they said, this is so cool so I can only hope that great things follow through from these projects.
Here we are today closing out 2019 and has been so much going on in the data and AI world part of the reason I brought to life the HumAIn Podcast and we were talking just a few minutes earlier about the human center movement and human centered design is there’s so much parallels about what products can be created and what people can work on these products and what processes can be in place but often the case that people forget about is engaging community, engaging activism and getting everyone involved so I’d love to hear about how data kinds have been doing that recently with your mission and what use cases are evolving as a result.
And for folks who may not be familiar with data kind, we’ll say that we’re a nonprofit dedicated to using data science and AI explicitly in the service of humanity.
The idea was really born of this realization that there are as the data revolution occurred over the last 10 years, huge opportunities, not just for businesses to use these algorithms to increase profits or efficiency but also social change organizations. They find nonprofits and government agencies and civil society actors are now a wash in data from digital data. They collect mobile-based initiatives to satellite imagery and so the question was what would it take to bring the data science and algorithms responsible to their causes?
Because they’re already at the frontline of social change every day data is really kind of set up as a little bit like a doctors without borders for data geeks, where we would technologists from big tech companies or from academy volunteers, every time alongside social organizations, to co-create design solutions that really work in the benefit of humanity and of course, as you mentioned, you were a part of, one of the many events we do. We sometimes run weekend events that are like modified hackathons, where we work with organizations actually, as you saw for months ahead of time to understand where your data problem, where the data that might be possible, maybe more importantly where do data science and algorithms have no value to bring.
Let’s not even bother going there and then have a bunch of technologists work over the weekend on those challenges but we also do long-term projects where folks sign up with our data Corp to work for six to nine months on longer-term engagements and so I just to tell a quick story about what this looks like, there’s a group that we worked with called the Moulton Niguel water district, their whole mandate is to make sure that people in Southern California get water and especially as with the climate crisis, more droughts are occurring in that area.
One of the challenges they face is they really have to have accurate information about how much water people are going to consume, because if they run out or if they misestimate that really their only recourse is to take a dump truck drive it out to another state, fill it up with a hose, filled with water, drive it back and that’s an incredibly disruptive, costly maneuver it’s bad for the environment so you really want to limit the amount of times that happens so they were wondering could data and algorithms help us predict how much people are going to consume more accurately.
So we found some folks who were environmental conservationists and even the chief data scientist at Netflix, I’ll be able to volunteer on this project to see what they could do with data that the district had as well as digital data that was available out in the public space and after about nine months, they came up with an algorithm that was accurate at predicting down to the block level and about 90% accuracy, how much people are going to consume and this would update over the season and in the first season of using this, the water district estimated to save them about $25 million and allowing them to deliver water more effectively to folks.
So this was sort of a little sample of what I am a project that is a very typical data kind of project and just to name, what I really love about it is that this is not the technologist coming in and saying we know what you need, or going off in a room on their own to build something that would be, it’d be forced it on the social sector, maybe core to our work is co-creation these are folks that really kind of try to paint a picture of saying, what if this small Indigo water district had Google’s engineering team? And they were all kind of embedded with the project? What could they do? And so it’s been very exciting over the last eight years to see this grow to about 20,000 volunteers around the world represented in most countries. We’ve got chapters in cities, Bangalore, Singapore, the UK, San Francisco in DC, and done over 300 projects, helping social change organizations get a little further on their own community-based goals with technology.
This use case you just described was a phenomenal project because it does hit so much to home for everyone in the United States today, we see in the news every other week about the wildfires that have been involved in California since going back to 2017 and I recall actually, when we worked on one of those David jam projects, Even in 2015, wildfires and forest fires.
This is one of the topics that are being explored so it’s amazing to see that regardless of unfortunately climate change and everything that’s going on, that you’re able to bring people together for a common good and to create results that are saving, municipalities and saving lives and that’s amazing for water that’s amazing for clean drinking and that’s for everyone of every socioeconomic status. One of the biggest takeaways I just took away from your story, which is coming off the heels of another conference I attended in New York City about open source in the past few weeks is that open source has created so much access and the exact words you just used before was that you can have the quality of Google engineers, even at your NGO or nonprofit because of open source because of all these projects, and that’s a huge part of what the other kinds of been working with.
When you talk about humans in the loop, this is really our goal is to help humans on both sides and in a way, actually be a little bit for the really good data kind out of the way to be honest, we just see ourselves as empowering those who would otherwise work together if there just weren’t so many blockers in the way so of course there are social change organizations that are working on everything from reducing homelessness to making sure that there’s more racial equity in our country, too stopping infant mortality that are cared deeply about these human causes, could be boosted by technology and then on the other side, tons of compassionate technologists who realized they’ve got skills, whether it’s coding or an analytics or machine learning who could be using those skills for those problems.
But of course you can engage on that on their own and so really what kind of fills me with joy really is a being able to help connect these folks and not just connect it someone once said you’re kind of like a matchmaker and a relationship counselor, public facilitate how folks can work together and make this a space where the focus is not like so much of our technology lives are about building it, building the app or making the cool thing, but actually started from the principle of what does a community want.
What does a community need to see done? And then how can technology support that or not? That’s often sometimes the most useful learning lesson is there’s not a place for data here or the risk is too high and so it’s maybe better to do something else, but that’s what I really like is that it’s all about folks who share a vision of the world being better and technology having a role in it working together.
And that’s brilliant because I was looking at some of these top surveys on, how are projects being completed at both startups and NGOs and multinational companies? And what it said is back in 2015? This is just as the data surge was starting to occur somewhere about, 30 to 50, the one for having data or designer individuals compared to software engineers at companies and now we fast forward to 2019 and it’s somewhere from a five to one to nine to one ratio now of having data scientists and Schumann sensor designers and design thinking at organizations so it’s no longer just, you got to build the product, but is the product going to be used? And are we thinking about what is the best way to use these tools?
I as you’re looking at kind of the AI debate today, there’s a lot of very necessary conversation about the proper use of AI, the ethical use of AI in our society and what you’re seeing is folks are realizing that as this technology is being rolled out and applications and being released into society, we may not have done enough of that human centered design upfront and really reigned in the limitations of this work.
There’s outcries against Uber putting self-driving cars in the street and at least in one case, it actually killed someone potentially because of improper testing I said, that’s improper, but it’s up for debate with what is a how that, why that was rolled out the way it was. The questions about facial surveillance, facial recognition and what that means for surveillance privacy are huge what is when we look at what people are asking for the St Louis, what I hear in the conversation is like, we want to live by our societal values that we believe in, and that some of this technology might be encroaching on it and so there’s a very strong kind of pushback say we need to ban this technology we need to stop what’s happening with it’s creating, I don’t even any of your listeners probably are aware of all the headlines about, reinforcing racial bias, increasing systemic inequity real true problems and so I will say, I want that work to continue because I happen to agree.
This technology needs more guard rails and possibly regulation. I’ll save my personal I don’t know why I kept why I qualified that it needs some regulation, but what is missing from this conversation is the constructive path forward that framing sort of sets up as the Global North and large tech companies are going to visit AI upon us and it’s the Citizen’s job to push back and there is that dynamic, but where’s the space for a conversation around and if we could build, take advantage of the opportunities of this technology for what we as humans want, what would that look like?
And that’s where I’m like, very excited to see the conversation of all of the stuff that we’re trying to push here at data guys to say, well, everything we dislike about what we call by implication unethical AI, how would we need to organize ourselves to build the alternative? And so I will say at Datacom, one thing that we’re working on is trying to look at what that kind of next level up of a solution could look like beyond just working with one organization.
So like the example I told you about Moulton Niguel like that solves one problem for wanting to go water district what if you want to take a step back and say, while Niguel is trying to tackle a water availability in drought climate that’s a problem that affects lots of groups outside of mountain decals, water district it affects it depends on actors, even within that water district that are not just Moulton Niguel itself depends on consumers, depends a lot, etcetera if you want to step back and say, not just how do we help them be more efficient, but how might we actually use AI or machine learning I’m sure you’ve caveated many times on your podcast about that the meaning of that term but I’ll just say, if you were to say how to use this digital technology for that challenge.
That’s a very interesting and really complex problem and so you could imagine there could be things like maybe this water demand predictor could work for many organizations, but to even figure that out, you’d have to work with, so many different groups understand their workflows you have to get so many different data sources from them cause they all have different data and then at the end, it wouldn’t be enough just to build the water need predictor if you really want to build this kind of ethical AI in the way that communities define what it does, you need to also make sure that community members and social activists are involved in the process from design all the way to the oversight of the system to say, we’re not done until it, until it achieved what we wanted to achieve.
And so that’s a very different model than saying we’ll work with an organization one-on-one and so data kind of we’re trying to do here is something we call impact practices, where we find areas like issue areas that we think seem that there are a lot of folks who are organized around them that have enough data and funding and interest to then say, great let’s do portfolios of projects with all of these partners, help them individually.
But also learn from that about where there might be these prototypes or opportunities or datasets that could actually grow to help folks at scale help many people. Maybe we’re certainly not going to go so far to say we could solve a social problem and I’m making the big air quotes with my fingers but until we get technologists and communities and social organizations working together, we may not even know the answer to whether there’s even something we could do and what it would look like without that so I was just pause there to say, in this whole ethical AI conversation, we need a constructive voice and getting to do that beyond just the one-off projects and towards issuers is that is it a fundamentally different human centered design problem than we’re often faced with in the field.
Another research that was just announced in November, 2019 was that GitHub, this big open source platform where a lot of developers put their code, said that the state of the Octoverse came out, this is their annual report on the state of developers and they said, we now have over 40 million developers on the platform and everyone says that’s amazing. So many developers when they said off of those 40 million, 10 million join in 2018.
So that is the growth that we’re seeing in the developer community and all these developers do have the opportunity you put it to have that voice right to speak about not just the ethical implications we see a lot of packages coming out and all these programming languages like Python that say, this is the metric, this is the tool to make sure you’re being ethical, responsible HumAIn but the phrase that you just shared, I’ve heard a couple times this year unethical AI that is a new emerging discipline that when you think of human centered design, we do have to think about what is not just ethical, but what’s unethical. It’s almost like back in school with our synonyms and antonyms it’s this, or it’s not that.
So that’s important because when we think of these impact practices, all of these areas are critical to society if they don’t receive the attention of racial bias of system and the quality of clean drinking water are not something that we focus on then they’re going to degrade over time and we’ll have situations like, shall I say, the area of Venice, Italy that is underwater because of the amount of corruption and political infighting that they’ve spent now almost 20 years, maybe even say 40 years to get these water floodgates systems up for the city, they haven’t been able to so we are able to in the United States, these impact practices are areas that we’re drawing the attention to, as you say with the right data, the right funding, the right interest, it’s possible and we have to keep humans in the loop just recently on one of the political debates for the democratic party as mentioned, one of the candidates.
This year Andrew Yang from New York had spoken about data. We all need data AI. We have to think about AI. He’s the only whether wherever that will take the conversation is the most important part of that theme to what he’s brought up in the debates is where as a society are, we thinking about being data first AI first in the service of community
And you bring up a lot of really interesting points there and the to your point about so many GitHub developers joining there are it’s exciting to see that more and more folks are gaining, not just technology skills, but being involved in open source communities and that the to me, that’s also the lesson of data can I just that there are so many opportunities outside of work if you happen to be privileged enough to have a good job and have some time outside of this where you can you in command what you want to use this tool for, and this technology for whether that’s joining an open-source community to build a game that you like, or joining sort of this open source movement of working with social organizations to see some social change.
So the opportunities are more greater than ever I do think finding and matching that opportunity, like if free developer time to aid and meaningful problem it is where the really hard part comes in where the human centered design piece is so critical because you can’t just knock on the door of say the red cross and say, Hey, I’m a coder what do you need coded and say well processes you’ve experienced with us to really getting to something folks can work on so that’s sort of one of the next frontiers that we need to work on is to start saying, what are the problems that we can go through a discovery process with? How do we scope out problems for folks can help with?
So that’s critical I want to comment on one other thing that you mentioned there though in terms of being a bringing up, but you don’t have to see the news about Venice being good being underwater, not being able to resolve the flood situation and thinking about that what’s unethical AI, ethical AI I kind of want to say, like to me AI is honestly a huge red herring in this whole conversation the way I look at it, AI is just a big accelerant it’s just like software it is just software it’s lighter, fluid on whatever goal you set it towards whatever system you put it on and so I hesitated a little bit, even hearing it said back to me, it was like unethical AI is ethical in the end it’s there are different systems that are designed to do different things and they will use AI for the goals they have so companies, and I don’t mean that disparagingly are by law designed to grow and get big to make profits.
That’s why you incorporate a company and so understandably AI, when those situations built by and for companies, it’s going to be used for growth and where we’re seeing tension is of course, some of that growth comes at the cost of other social elements that we’ve come to rely on, hence the tension but what’s also important is even going to the social sector and saying, Hey, there’s good stuff.
We want to fight on like, we don’t want Venice to be under water does not by nature imply that just that AI will be able to apply there and I do argue that of course, AI built by and for folks within the social sector, at least is more incentivized to reach the outcome, the social outcomes we want over anything else, but let’s face it a lot of our social problems are human problems? And so AI will make a racist criminal justice system more racist it’ll accelerate those racial outcomes, but it’s not going to solve racism, it’s not going to actually make us more racially equitable. If he can believe someone, tell me about that project.
But similarly we have a lot of squats as humans constantly debating policy, what we want to see in the world as someone with bad faith using AI will use that to advance their, bad faith policy but I don’t know that there’s an AI solution to getting folks to CII or a reason about the kind of world that we want to live in, because that is an inherently messy human process with no, no clear objective function so just want to say, the kind of humble stance I would take on all of this as like, as an accelerant and there are some systems and working social elements that AI could help with but they are select and the trick is finding them and really promoting them as opposed to thinking that it could then it’s naturally ethical if you’re doing it for a quote for a good cause or that it can solve all of the social human challenges, because unfortunately those surf really the toughest ones to solve.
What I loved in that, what you just shared Jake, is I heard so many great things about what’s needed for community building and for open source two of the big phrases that you just used for policy and process and that’s usually where everything gets started. We talk about what’s needed to create the change and then you have projects and these are all the projects that David kinds working on in the short term and the medium and the long-term and then ultimately the next part is standards, creating open standards, which can be available for many organizations and some of that is these software packages that we’re seeing online on platforms like GitHub for different code bases.
But we’re just getting started with standards going to talk about human centered design ethics there isn’t very much standards today everyone has an opinion, but the European Union came out with theirs whole report on what they think the standards for data and AI is they came out with a 67 page report it just a few months ago and that’s amazing, but what action can we take? What strategy, what cultural changes will happen from these standards? So I’m just curious, what your take is on open standards and how that might be defined or reshaped over the next year
Like you said, very naturally come out of community processes and say, Hey, this is a policy by which we can all work together. I don’t know that I have a lot of kind of profound feelings about standards in this space, except that we are struggling right now with setting standards for HumAIn or ethical AI because we’re sort of focused on the AI as opposed to the goals and constraints of the system it’s being designed in there’s been a large push for ethical AI standards, for computer scientists and AI engineers, machine learning folks to adhere to and that is a very natural step towards standardizing our practice, the problem found is one, there is no standard set of ethical principle at this point, there’s gotta be 50 to 100 that I’ve seen now, other than they’re all great.
Like they’re not like a bad one or one that’s particularly better or worse it’s just that everyone seems to have wanted to create their own but more than that standards are only as good as your ability to enforce them and one of the things I found really interesting about the ethical AI standards debate is there seems to be this belief that the engineers, or at least there’s one school of thought that if engineers were trained in ethics or had more ethical frameworks, maybe we wouldn’t have some of the outcomes we have in companies today.
But about what it would mean for an engineer to stand by their code of ethics in a corporation the one is like one challenge to the accountability of the standard is there’s a chance unless you have a particularly a company with a strong whistleblower policy qnd obviously everyone has to have one legally, but strong record of standing up for its people.
You have to risk at least your reputation, if not your job to speak up. She said, Hey, I don’t think what we’re building here is good. I don’t want to build whatever the facial recognition algorithm that I feel is racially biased so that’s one it’s like there’s a real cost to doing that and we’ve seen some incredibly disheartening examples of Google actually removing folks and letting them go, who spoke up about ethical practices at AI, at Google so that’s one is you’ve got to take huge personal risks that’d beyond that.
It’s challenging to actually know what is an ethical risk. Some things are sort of low hanging fruit if you testify actual recognition on a different complexion, different races if it’s not performing accurately on all of them well then maybe that’s not a very good algorithm.
That’s pretty low hanging fruit but what about stuff like coming up with a better routing algorithm for Google maps, if there maybe systems level effects you haven’t anticipated, like for the fact, people realize and the early deployments of that algorithm cars are being rerouted off of crowded highways and into, side streets that were basically congesting, currently poor, usually low economic status communities that live by the highway so we created like a social and unintended social consequence so you probably aren’t able to actually estimate that at the time that you’re get the assignment from your boss upgrade the algorithm, decrease people’s route times.
And then like lastly, I was like I don’t think you want the engineer making that decision if the engineer were aware of that, like do you want the Google engineer in isolation making a decision about whether routing through a small side street in these cases, 70% of the time, this is good or bad for our society?
Like, no, that needs much more oversight from the rest of the world. So I’ll just say, like we’re in a little bit of frontier land with any of these standards would be these ethical codes be they, how AI should or shouldn’t be used be they just no standards for what’s sort of proper labeling of data sets such that you’ll have even just racially equitable and gender equitable outcomes from the algorithm it’s tricky and what we really need more than standards right now, or actually just standard accountability methods we need to have. Ways for citizens and folks who and consumers to have oversight and agency to speak up when these algorithms don’t do what we want them to do. that’s ultimately the kind of solution at this point before we even get to standards.
I had the opportunity just a few days ago in New York to go for the first time to an Amazon Go store and these are these new stores that Amazon launched, where there are no cashiers I took the Amazon app I scan the QR code and they went into the store and then it was like where’s that technology, how is there no cashiers? I looked up to the ceiling there were, I don’t know, at least 500 cameras connected to a mesh of so many wires and signals just to be able to capture my every movement in the store to make sure which item I was grabbing it I just got like a $2 vitamin water, just one item, picked up a Coca-Cola, put it back, grab the vitamin water, just to see if it could recognize it and it did and I was so blown away by the technology, but of course that’s for me as a white caucasian male.
So it doesn’t mean that would be for everyone and so I said: here’s a case that’s maybe working and, and perhaps with more labeled data, we can get there but then the other day I had a case that for me actually was almost upsetting, but was, it was comical is that I love the Google Photos app so Google has their Photos app where you can, back up and they’re free cloud storage and then have you ever tried to recall your photos? What was the picture of my dog? You could type the phrase dog and it comes up and I see my beautiful scottish Boston terrier and then I said, what about my old dog? The one that died a couple of years ago and so I type Sheltie shut-in sheep, dog and it comes up and then it shows and then I had the name of the dog.
So I typed the name of the dog in and then not only did that dog come up but the thought a different Shetland sheepdog that took a picture of in central park was the same dog as my dog that died two years ago so for me it was comical in this sense, but it was showing that facial recognition technology does have a long way to go because. How is the AI going to know to label those dogs as different animals? The things it’s still the same species of Sheltie there’s still a white person or a black person so there’s a lot of work to be done there and some of the work we’re going to see, I had a few months ago on the humane podcast, I had the CEO of cloud factory Mark Sears on, and I know they just recently raised a $65 million round and a few weeks ago to expand data labeling because you can have it, the AI machine learning, you can have the feature engineering cleaning, but if you don’t have the label data, well where’s it going to leave yet?
That’s totally and you’re right it’s low stakes when it’s you and me buying a vitamin water or looking at our photos, but when these are being used for say, predicting recidivism and being used in criminal sentencing, or I there’s so many horror stories out right now like the different allowances given by Apple card the different genders like then that actually has real implications on people’s lives and there’s a great, actually talk just came a few days ago. I hope he forgives him if he hears this, I’m not sure how to pronounce his name, but it’s Arvind Narayanan who’s a professor of computer science at Princeton who read it gave a talk called how to recognize AI snake oil and one of the most present slides, and that was really great was sort of talking about, he says this kind of breakdown of one of the things that, where we were AI is a pretty proven, like machine learning has worked pretty well in terms of like medical diagnosis from scans, or like we’re making genuine rapid progress on reverse image search, audio search.
He’s got another section that around automating judgment that’s like, it’s imperfect right now, but it’s improving like spam detection and copyright violation and then there’s a whole category that he labels fundamentally dubious not to say it, these don’t necessarily work, but she’d already sort of have your awareness up and these are particularly around predicting social outcomes predicting at-risk children, predictive policing and that’s maybe the frame that we need to be thinking about with all this stuff is you brought in these imperfect systems what are the outcomes for which we’ve may be able to say, well I’m taking on less risk in this system, or I can understand why machine learning would be better at that task then another task is right now it’s all again, it’s one of those gigantic and too expansive umbrella of AI digital tag, which people use to mean anything from generalized AI team, just a computer did something so maybe one path through this, as things get risky, that’s constructive as get a little more literacy on just what segments we’re doing pretty well on versus where we might want to be a little more skeptical and then also I’m a big fan of the about ML project that the partnership on AI is doing and I feel so bad. I can’t remember what the acronym stands for, but the point of the working group is to come up with sort of an explainable version of what algorithms are doing that sits between the kind of marketing hype of like how much progress is AI is going to make, and how much is gonna save you money.
It doesn’t really tell you much about the algorithm and the alternative, which seems to be like spec sheets and like, , the code documentation, which is way too technical for most audiences, it’s kind of a middle ground Bettina number of companies saying we’ve got to have some way for folks like users or even just the public to understand it enough about what this thing is doing or not to know how we should use or not so one did pick some constructive steps forward because it’s too easy to point to all the negative articles about this stuff these days but some of those kind of projects to increase literacy can be really helpful for understanding how and why these systems work the way they do.
You said it best is that we’re often very prone to blame, but that’s because we want to have a better life, better quality of living and this technology that we’re starting to see is so new this AI implementation, whether we call it AI, machine learning, deep learning data science, bundle it altogether it’s only been out the past few years and it will get better over time, but we as citizens, we should have the advocacy and we should have the agency to participate in this dialogue to make sure systems are implemented properly and as you mentioned from the professor Princeton, whether it’s policing job success, terrorist risk, at-risk kids, recidivism if all these different topics and they all have different outcomes.
Some that are more urgent and some that are more important we look at the classic case terrorists: you go to an airport, we have facial surveillance attached to our passport yo know if we can board a flight I don’t think many people are going to oppose asking the government no, I don’t want you to have this, they want to make sure terrorist and on boarding planes so I’m like, sense, but in other cases, that’s where the challenge is like job success, job hiring gender bias for credit cards with Apple are you kidding me? We’re in 2019 where I had this conversation a few weeks ago with one of my colleagues who said, have you noticed how in the rest of the world, there’s not really a big conversation on gender and computer science.
You’re a man, you’re a woman and you’re smart, you’re my equal, you’re my peer it doesn’t matter. We’re all working together, but in the U.S it’s something has changed I don’t know if it’s culturally, socially, wherever that divergence happened and there’s a lot of research that says maybe it’s somewhere around sixth and seventh grade are entering high school somewhere there, we have a lot less focused on STEM and, making sure that the girls up today are willing to and capable and allowed the access to be on the same playing field as men? In computer science and in the fields that we’re growing with and so when you share that comment earlier in the Apple card that just triggers me to say like, are we not beyond this?
I do not think Apple engineers are beyond this, that they would exclude that parameter from the machine learning, but perhaps not yet and you wonder where the breakdown in communication and human centered design goes wrong, but like you said, we should be constructive to look towards the future to see what’s next and we’re going to be moving there predicting social outcomes is the next step.
We’ve heard about them all throughout this whole journey for election cycle 2020, whether those social outcomes are taxes or healthcare is affordable housing there’s a lot more work to be done there. And I’m always really excited to see the use cases that you’re describing data kinds working on and anyway, I just want to tie that all together to hear about snacks. What are some of the next challenges and projects are exciting, things that are yourself and data kind of working on.
You said it best at the front. It’s still incredibly new and so it’s 2019, but 2019 in an age of the technology that we’ve really barely gotten our hands around and so one of the things that we are really committed to seeing is a world like you described where we may not have cases of things like gender bias in these technologies, if perhaps more folks who were affected by the technology were involved in the design and oversight of the process and also outside of what are the kind of traditional business models for creating technology? If you want to create machine learning or AI right now, chances are the only place you could do that has the resources as a university or a company and as I was sort of saying before, AI is an accelerant for the goals of those, those institutions so you’re mostly going to work on tools that if you’re a company you can sell to people or a university may be advanced or research.
But we want to create a space where communities can actually build the AI technologies they want for the social outcomes they need and so as I was saying a little bit alluded to this impact practice approach we’re really transforming data kind now we’re taking the network and trying to move from just doing individual projects to saying for significant social challenges, how can we support the ecosystem of players trying to tackle them? So for example, one of the first issue areas is impact practices we’re working on is in the community health worker space.
The challenge here is that the world health organization has said look we were far behind where it needed to be for going to hit our health goals for 2030. I whole set of goals the U.N but alcohol the sustainable development goals to touch on various health outcomes ranging from infant mortality to a reduction of disease we believe if we hit we’ll live a better life and really far behind that, especially because a lot of areas just don’t have good health infrastructure and just are not hospitals you and about going to a clinic or a hospital here in the U.S. as if it’s an annoyance but available, but imagine in parts of East Africa, there’s just none of that infrastructure you just you’d have to walk miles to get that care.
And so there’s been these innovations in something called community health workers, folks in the community who actually go deliver health outcomes by going door to door, kind of like an on-call doctor and so it’s been identified as this intervention that might actually bridge this gap people believe if you could have more community health workers, all trained well-resourced might be able to bring down this gap between where we are and where we need to be with healthcare the challenge is there that even though of digitization has actually happened in that space, a lot of health workers, even in these kind of low resource areas, have mobile phones and tablets to do there.
Their rounds it’s just not a lot of data scientists there’s not do we all know technologists data scientists are incredibly expensive they’re often here in the west. It’s just, there’s a total resource inequity there. I’d say even with the technologists in the east like Indies, Africa it’s hard for these organizations to afford to work with them and so we’ve been sort of approaching a number of collectors are working with collectives there that are trying to help community health workers to become more effective to say we’d be willing to work with you on understanding if and how data science algorithms could help with any of this work.
And what I really like about this model is in the past, we might’ve said we’re just going to work with a group, try to help them with their community health workers, but in this approach, because of course there are thousands of technologists in the data of network we’ve talked to about 30 organizations.
Now think about 15 of which are going forward with projects that we can put teams on sort of all in parallel and across them, we think we have, if we were to work with all of these groups have pretty good coverage of most of the community health worker, digital platforms to then to answer questions, like what would it take to kind of automatically check data quality? One of the big challenges in these systems is that if there are data anomalies, they can really erode trust in the system the governments won’t trust the data, they won’t adopt these programs, but now as community health workers using digital platforms like mobile to input their data, it’s too much for just a human operator to kind of scan all of that.
So the question now is how might you use anomaly detection on these systems to help human oversight and again where we might’ve worked with one group to do that before now we can work with say the five major players on this and say if we can solve that, and what does it take to really solve this challenge across all these organizations? And more importantly, again we’re not beholden to having to build a product for our shareholders.
I say we, again data is more of a facilitator it’s not us we don’t really, we’re a public good so this can take some time community members are, could be involved all through the process community, be health workers, weighing in on both the design and the oversight of these systems and the hope is that by doing this two things will happen one we’ll find these opportunities, like I said, probably won’t solve it, but it’s more likely to say: this prototype worked for these three groups and therefore this could be developed across the space more than just one fun one-off project.
It could actually help it be taken forward at scale and also we have an opportunity to demonstrate how to build this kind of technology in a way that meets the criteria we talk about when we talk about ethical, AI has no motive, but the positive health outcomes that is what’s designed for the people overseeing it are both the administrators, the NGOs, and the community health workers, and that we keep working with them until it gets the outcomes at once can make it transparent that’s really the promise of this approach so that’s what comes next we’re going to be going into these issue areas junior health work is going to be number one. We’re going to be also looking at inclusive growth and kind of financial inclusion second, but the goal is over the next five years to do 10 of these.
So we’re always open to those ideas because of that we’re always expanding the network as well so we’re looking this year to open up new chapters of folks in their own cities who want to run data kinds of programs, again with a strong belief that the best change, the most important change happens with local communities are driving these, these decisions themselves and no matter what, whether that’s the kind of thing that excites you, or you just want to be a part of the data for good movement, come void now, do you have join firstname.lastname@example.org get involved, or many of the others data for good African groups that exist out there? What’s most important is that folks take part in some way in this, because we are all part of this system and the more that we can use our skills to do good at it, the better it’s going to be.
I could not have said that better and we’re moving into a world where everything’s being defined by data. It starts with collecting. It starts with the digital trust that you’re talking about with the community health work in parts of Africa and globally it moves some data labeling into considering the machine learning and AI so this whole workflow is becoming more mature.
It’s exciting to see the social good, these predictive outcomes that are only becoming possible now and positive social outcomes is what we have to focus on and if we think of these positive social outcomes, then ethical AI just becomes part of our workflow it becomes an everyday occurrence that we’re always thinking, how do we help? How do we become good stewards of society? And by doing that, we’re only being humane and it’s data kind for all.
I love it that’s perfect. I couldn’t have said it better.
Excellent. Well, Jake thanks so much for being with us on the HumAIn Podcast. They really appreciate your time.
My pleasure. Thanks so much.
Hey humans thanks for listening to this episode of HumAIn. My name is David and if you like humane, remember to click subscribe Apple Podcast, Spotify or Luminary. Thanks for tuning in and join us for our next episode, new releases are every Tuesday.