AI and the Economics of Doing Good with MIT’s Hala Hanna
Hala Hanna, executive director of MIT Solve, discusses why less than 1% of AI venture funding goes to social impact and what can be done about it. She explains how Solve finds and helps scale early-stage ventures that use technology to close gaps in health, wealth, learning and climate, and how it uses AI to review grant applications with humans making the final decisions. The hosts debate whether big tech can be expected to fund social good.
Transcript
AI is being sold as a technology that can transform everything. And we are not just talking about business, healthcare, education, climate resilience, food systems, and access to opportunity. But if you follow the money, a different story emerges. Hundreds of billions of dollars are flowing into AI investment, yet Hala Hanna, Executive Director of MIT Solve, points out that less than 1% of AI venture funding is going to social impact. If AI is so powerful, why is so little of that power being directed toward the people and the problems that need it the most?
Well, because big tech is not built to deliver social good. It's built to deliver shareholder value. So unless social impact AI becomes profitable, publicly funded, or required by regulation, why would the market prioritize it? Well, today we ask Hala Hanna, if AI really can be used for the public good, or if the promise of AI for good is just Silicon Valley lip service.
We're talking AI, social impact, and the economics of doing good with Hala Hanna, Executive Director of MIT Solve.
Alright, Stephen, from its inception, we have been hearing about how AI is going to transform the world, level the playing field, bring money to places that don't have any, yet that's not exactly what we're seeing. So what do you make of the disparity between all the money that is flowing into AI versus where the money's not going?
These AI hyperscalers are companies, right? And they have a legal obligation to maximize return to their shareholders. And it's as simple as that, right? And they're investing over a trillion dollars right now in that hyperscaling and a trillion dollars requires quite a return to the shareholders. So I believe Hala's right, that there is an imperative here for the social good. But I think we're actually going to have to do two things. First of all, we're going to have to do that consciously, because it's not just going to happen for itself. And second of all, we have to bear in mind the speed that things are happening right now. And we can't sit back and say, yeah, we'll come to that at some later point, because it's going to be too late by then. So how are we going to do it? And how quickly are we going to do it, Laila?
How are we going to do it? I think it's hard, because, like you said, the money that's going in needs to be made back first, as in that's the priority is and you can't just be distributing it here, there and everywhere when the amount of money that someone's put in, they haven't got back the two times return. I also think, right now, it's so hard, because we're not in a lovely era of globalization where, you know, the Melinda Gates Foundation is sending huge armies of American healthcare workers into various countries in Africa anymore. We're not. The priority in terms of social impact and social good is right at the bottom.
Very low.
Yeah, because there's urgency elsewhere, which is maximizing shareholder value or getting ahead in the global AI race and things like that. So it's very, very tough. I think it's the toughest time to pick to be in social impact.
It's tough. And when at the end of the day, these are private companies. And if you look at other private companies and the amount of money that they are or the amount of technology that they are either donating or providing for social change or for social good, is it somewhere in line with this 1% that, you know, the AI companies are doing?
Right. So we talked about this the other day, and we had a little bit of a tussle on it, where I said to Laila, I can't remember what it was we're talking about, but you can't leave this to the scientists. And you both had a bit of a giggle when you said, and you can't leave it to the CEOs either. And I think when it comes to AI for social good, you can't leave it to either. You can't leave it to the scientists, you can't leave it to the CEOs. So who else is going to get involved?
Well, I guess the government, historically.
Good luck with that.
Yes. Yes. You know, companies love being regulated by the government and being told how to spend their money. So that should go over really well.
We want to know what are your thoughts on this topic? We were promised that AI would help tackle access to health care and education, tackle climate issues and inequality. But the companies building the most powerful systems are under pressure to make money, as Stephen said, not solve the world's social problems. So do you think that AI can deliver social good at scale, or is the industry just simply not designed for that? Tell us in the comments and be sure to like and subscribe. Now, let's get straight to our conversation with the expert. Here's our conversation with Hala Hanna.
Hala, thank you so much for taking the time to speak with us today. It's such a pleasure to have you. You're celebrating 10 years of Solve at MIT. Can you kind of explain for us what exactly Solve is, and your role there?
Solve was created by the President of MIT the same year as the United Nations Sustainable Development Goals, a decade ago now, with the idea that the challenges of our time require us to stretch a hand to all problem solvers around the world. And so we believe that technology can and should serve humanity's most pressing needs, not just its most profitable markets. We think technology can close gaps in health, in wealth, in learning, and of course respond to climate change.
Now, tech doesn't change the world, people do. So we see Solve as this marketplace for tech for social impact. We find incredible early-stage ventures that are using technology to close those equity gaps, and we help them scale.
We have a network of over 460 solutions globally that are reaching 430 million lives, and we facilitate to these organizations almost $90 million direct funding, that is, and they have raised over $1.4 billion. So that's real economic value that they're creating in their community. So we connect them with those who want to invest in making the world the most sustainable place for everybody.
We've been talking about the fact that when it comes to venture funding, AI and AI companies saw $290 billion in investment, but only 1% of that just under 3 billion, went to social causes and change for good. I guess my question to you is, obviously, you would like to see a lot more of that money being utilized for some of the causes that you champion. Are you wanting to see that money, though, in the form of money that you can then take and implement and try to work within the initiatives and the programs that you want to see come to fruition? Or are you actually wanting to see AI infrastructure in some of these underserved communities and underdeveloped areas?
That's a good question. You're right that it's a rounding error of what's directed to AI applications on solving these social and environmental challenges. And, you know, sometimes in my comfortable Massachusetts bubble, I forget that 4.6 billion people still lack basic healthcare services, that 4 billion live in places that are already impacted by climate change, and half of all children in low- and middle-income countries can't read and write by 10 today.
So, the opportunity is huge because, at its best, AI can help a rural nurse diagnose disease earlier, a teacher personalize learning for three times as many students, a farmer predict crop failure, a government actually deliver much-needed services. These are all really examples from our portfolio.
I think the biggest point is that these innovators exist, they're out there. These are founders that are building with a fraction of the resources that we see are being spent today in conditions that are designed to stop them, right? And just in our portfolio, the ones that are enabled by AI reach five times as many people as those that are not. So, again, just to say how big the opportunity and the problems are.
And the mismatch with the current AI economy is that it's after more compute, bigger data centers, higher valuations. You know, they're optimizing, again, for the most profitable markets. And I admit that social impact AI is harder to invest in. It's harder to measure the returns. It's fragmented markets. It's longer timelines. But it's a rare opportunity to distribute capability more cheaply and broadly than we ever have before.
And to your point about venture capital, it's not so much one type of capital. I think there are concessionary capital that are better kind of designed to absorb some of the risk that private capital doesn't necessarily. But harder doesn't mean impossible and it means that the first bet is that much more consequential.
But I think it's great that you have your Massachusetts bubble in the first place. And I think, the work that you're talking about is amazing. But is a trillion dollars being spent by the hyperscalers? So my question really is, how do we have this conversation? Because the conversation that the hyperscalers are having is, as you say, it's about profit. You know, these are private companies that generate profit. And the whole conversation seems to be around them and the stock market. And have we got a stock market bubble? I don't hear the conversation that you're talking about anywhere. So outside of the Massachusetts bubble, how do we as a society or how do we as a group of societies focus on that conversation?
I'm so glad you're bringing the broligarchy up. I mean, companies do respond to incentives. So the question is not so much to your point, will big technology voluntarily save the world out of the goodness of their non-existing heart? It's more under what position will they serve public interest? And right now, as you said, these companies are operating in a calculus that is fully limited to valuations and stock prices. It's completely disassociated from the real life of actual people.
It's also the regulatory Wild West. Today, a deli sandwich is more regulated than AI. You know, an MIT professor yesterday was saying this on stage.
And at Solve, our call is we're not going to wait for big tech's conscience to kick in. I think it's either that we structure the incentives, which is mostly government and advocacy work, or we build the alternative model. And we're building the latter because I think there is a danger to a single story, about who gets to be an innovator, what progress is.
And just so we're on the same page, the vision we're being sold is, we're going to cover the entire world with data centers, and space too, by the way, to summon a demigod that will leave us probably unemployed, may or may not precipitate our extinction. And then every time I press enter on a query, it runs on the entire body of data ever created online. And we're back to prioritizing now cheap energy over clean energy, basically burning barrels of oil over micro doses of dopamine. So sign me up.
I love the way she explains all this. But on that note, because we are not talking about what is actually happening in developing countries in the global south, and they've kind of just been completely left out of the discourse and what you're talking about, what is happening there. And also in terms of what could we be doing much better there in terms of investing in founders and the consequences of what's going on with this huge trillion dollar investment in like data centers on the moon, because they've just completely been left out of the equation.
That's right. And just because I talked about big tech, I do want to say that we work with plenty of big tech that actually do want to do more. And to your point, they might not make the first page, but we work with HP, with Amazon, with GM. We partner with them because they want to find innovations that help them reach their own environmental and social goals, which they've set. For Amazon it's carbon neutrality by 2040. They're still going for that. For HP, it's getting 150 million people included in the digital economy and that's worldwide. It's not just in US bubbles and incentives can align. There is a world out there where this work is actually happening.
One thing that I always try to remind myself is that if we don't imagine the world that we want, we can't make it happen. And I remember Paul Polman, every time he had a board meeting at Unilever, he used to leave an empty chair. And what if today we left three empty chairs at every boardroom, one for the customer, one for earth and one for future generations? We can't create the world we don't imagine. And we're certainly forgetting billions of people behind us.
Alright, we're going to pause just for a second. But coming up after the break, the use of AI when it comes to reviewing grant requests for social change initiatives. We ask Hala whether AI can help or hurt the approval process for those looking to make a positive difference in the world.
We're back with Hala Hanna from MIT Solve. I want to talk about AI funding decisions and how AI can actually be utilized to help you specifically. It takes approximately four minutes for an average human to review one of the grant applications. But obviously, with the use of AI, that can be expedited, and they can turn over hundreds of grants per minute. How beneficial is that? And do you worry at all about some inherent biases that we know these LLMs can unfortunately have?
Just to take grant applicants hours and hours to write the applications that funders would put forward, and now it takes minutes too. So everybody is grappling with this increase in the number of applications. Last year, Solve received 3,000 applications from 190 countries. I think that's fantastic. That's 3,000 dots of hope and people all over saying, we don't take the status quo for granted. We want to change things. So that's, I think, good news. But it's, of course, a lot for a small team.
So we built an AI tool to create a more standardized and consistent first-pass review with researchers from Harvard, from MIT, and from the University of Washington. It basically assists our judges in expediting decisions while actually minimizing bias because humans remain accountable for all decisions. So the goal is to reduce noise and give better feedback, and only remove the ones that are really obviously either not matching the standards and the requirements, or the ones that are so clearly great that they should be reviewed by a human.
So really, the AI helps us narrow the time we spend on the applications where human judgment is most needed. And AI gets it right more often, actually, when you've trained it on diverse data, of course. We're very deliberate about that, because if the training data reflects the same old biases, we've just automated faster injustice.
Now, over 10 years, we have a body of 28,000 applications, again, from all over. So that's a very rich data set. Now, we put two guardrails in place to ensure that those biases don't creep in. First of all, you can't ever be sure that you fully solve it. So you have to commit to iterating. We audit actively. We test continuously for those recommendations, how they correlate with geography, with language, with credential proxies, so that when we find something fishy, we correct for it right away.
And then we never let AI make the final call. So AI flags and summarizes and suggests, humans decide every time. I actually love this saying by Audrey Tang that humans in the loop is the wrong metaphor, because we don't want the humans to just become a hamster on a wheel without knowing where the wheel is going. We want AI in the loop of humanity instead. So humans are setting the destination, and AI is helping us navigate.
Can I take you back to the point you were making before the break, which I found really fascinating. You're talking about the importance of imagining the world that we want, because if we don't imagine the world that we want then... Now, I come from the north of England, and I often talk about what's going on now with what I call the agentic revolution, with what happened in the industrial revolution, where I came from. And the Victorians did that, they did imagine the world that they wanted with parks and with parliament and new people voting. But it took them 100 years to imagine the world that they want. And I don't think we have that amount of time this time. So how, as a society, do we participate in that imagined process so that we don't leave it to the Cambridge bubble, not even Massachusetts bubble?
I wish I had all the answers, Stephen. But maybe I can tell you what we do. And I would actually love to hear what you think, if that's okay. Because that's something that I think about all the time. I do think there is less... People are angry and quite frightened. And I think that's the starting point of something. And I think we need to kind of come together and think about how do we use that to create something productive.
Because we know the world is not short on innovation, and we have this decade of data on what works. We're looking at three things typically, and helping scale those solutions in wherever they are in the communities that they're working in, because they're close to the problem, they understand what it is. So it's about the solution, the type of capital and the ecosystem.
So when it comes to the solution, first of all, we really believe that experience is expertise, that proximity to the lived problem matters, because it changes the nature of problem solving. And those folks that we work with, they have a dogged determination to figure it out. They have a skin in the game. Of course, when in the typical venture capital, you're looking at trends. And so we end up funding the same geographies over and over, rather than that. So 85% of our solvers have lived experience of the challenge they're addressing. So the solutions are more trusted and more culturally appropriate. Because of that, they're building for scarcity.
So let me give you an example of Speetar. It's a Libyan solver. During the civil war in Libya, 3 million people found themselves without medical access. So they launched Speetar, which means hospital in Libyan, to connect anyone with a smartphone to a doctor. Not revolutionary, but the real frontier was the data. There's no medical record, no API, no digital infrastructure to plug into. So they had to build it all. They did cloud-based electronic medical records, data aggregation from paper files, lab results, prescriptions. And from that, they created a small, context-aware model that is trained on these, frankly, previously invisible data sets of rural records and a collection of informal transactions, all in local dialect. And today, Speetar is the backbone of most of Libya's healthcare system. So that's just one example.
You were talking a lot about people's personal healthcare data, and that actually brings us perfectly to our rapid-fire questions, which we ask all of our guests. And I'd like to hear from you where you think people will draw the line on bringing AI into their personal lives.
You know, we've traded privacy and agency for convenience a lot more than I think a lot of us expected. But I don't know if you feel it too. I feel a wind of change. I think people are getting angry. I mean, I was reflecting that tech executives are avoiding the red carpet at the Met Gala, right? They're scared. And of course, Sam Altman was attacked, which is totally not okay, but they've really transformed from heroes to villains at a very fast pace. It feels like the pitchforks may already be there.
And I think that just tells us how important this technology is. I mean, railroads, telecom, the internet, these technologies were so fundamental that first public expectations changed about them, but also the government stepped in, you know?
For sure. And my question is, what is something that's universally accepted in your field that you strongly disagree with?
That bigger models automatically mean better outcomes for society, you know? For whom are we building? At what cost? And who decides, right? There is a movement now of frugal AI, small context-specific models that run locally, that can be really more impactful.
And I think that goes back to your point, Stephen, of how do we do this? Well, in some way, I think by you having me on your show, you're spotlighting a different way of doing things. And that's the work that we do too. We spotlight these incredible entrepreneurs all over the world who are changing the status quo.
So my question is difficult for a lot of people, but it's going to be really easy for you. So what I want to know is, what are you optimistic about?
Yeah. I mean, so much, this whole thing, all of it. It's so much more fun to go through life optimistically. So it's a choice as much as it's a daily action.
I think if, especially coming from international development, the problems are so big, right? And it takes decades to move the needle on any of them. But here we are working with these 430 companies and seeing impact day in, day out.
And what really inspired me is as the aid industry and infrastructure collapse in the past couple of years under the weight of just killing USAID, et cetera, these entrepreneurs, I would sit in DC and New York and I would hear lots of folks complaining about, rightfully so, right? It's very jarring. It was a very jarring thing to happen and way to have it done. But our innovators just picked up the next day and go. They're just so embedded in their communities. This is an opportunity, I think, for them and for the global majority to create a new way of being - that is sovereign in that sense of local entrepreneurs taking care of local problems and scaling to impact the lives of millions.
Hala, I want to finish with one final thought from you. With the perfect AI technology and all the right amount of funding, where do you see AI having the largest and fastest impact when it comes to social good? Which arena?
You know, one thing is there's over 6,000 languages that are at risk of extinction. The majority of them are indigenous. And this is like irreplaceable, right? If a language dies, with it dies the knowledge, the culture, all of it. So we have indigenous communities that are using artificial intelligence to preserve those languages, speech recognition, neural translation, storytelling, in their own world, in the words of those who are the last ones to speak those languages.
And these are communities that have been the most harmed, perhaps, by extractive technology. So to see them finding ways to use it as a shield almost is really inspiring. And that story, of course, is, to your point, Stephen, completely invisible in the mainstream. One of them is called the First Languages AI Reality Initiative. It's racing to prevent those endangered languages.
Never in a million years did I expect that to be your answer. So that was really fascinating. Thank you so much for your time today.
Thank you very much.
Thank you so much.
Well, that was really interesting. I didn't expect preserving native languages to be your answer to that last question. I really thought she'd go somewhere with access to health care or something like that. But I think it's really interesting. She doesn't really, even when we pointed out, only 1% of all the AI funding is going toward these societal change initiatives. I don't feel like she's holding the Silicon Valley tech bros accountable for not putting their money where their mouth is, essentially.
Yeah.
But she's like, you're doing the best you can, but we could always use more.
Yeah. But I think she's not depending on them either. I think that's what she was talking about where they decided, like, yes, you can do it through government and regulation. But that's also, right now, not that viable. But investing in founders globally who are really innovating in that space is actually probably the most productive way to do it. So they've just actually taken it into their own hands, whereas I think that all of us are very much. And I think that's where the backlash comes from, where you're like, well, why isn't Sam Altman investing in this Nigerian stuff? Well, why would he? He has his own interests. So yeah, we can't just all sit here and be like, hopefully, Mr. Musk will do it for us.
I thought what was really, really great about Hala was the fact that she sort of accepts things as they are, she takes things as they are, and then makes a positive difference. So she understands the money, she understands the compute, but then actually moves things forward. There was one point I was taken with that I'd have loved to engage more with was the point about the pitchforks being out, right? She said the pitchforks are out.
I agree with her on that though.
So how do we move on from here with the pitchforks moving now? And I'd love to engage 'cause I think the only answer is you have to engage with the pitchforks, you have to embrace the pitchforks, you have to understand what's going on and then move forward, which is what she's doing on that micro level. So it was it was such a ray of sunshine having her on the show.
Totally, I agree. Talk to her all day.
Well, we'd love to hear what you think. After hearing from Hala, do you think that AI should be pushed toward prioritizing more social good? Or are the commercial incentives just simply too strong to sway current priorities? Can big tech help solve global problems? Or should we stop waiting for companies to do what only governments, funders and public institutions can mandate? Drop your comments below and be sure to like and subscribe and we will see you right back here next time on "Agents of Tech."