Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Who is left behind when AI moves fast? with Dr. Chinasa T. Okolo
4 Plays7 days ago
Dr. Chinasa T. Okolo is the Founder and Scientific Director of Technecultura and a consultant for the United Nations and the World Bank. She talks with Scott about the yawning gap between AI's pace of development and policymakers' ability to understand it. She also discusses what it means to do AI governance research with a focus on Global Majority communities that are too often left out of the conversation.
Recommended
Transcript
Can explainable AI provide user agency if technology is non-negotiable?
00:00:00
Speaker
That just lit up all the neurons in my brain with like a ah a fan out of like three different questions. Just because something is explainable, Can the explanation create agency if the user themselves, in this case Indian healthcare care workers, has no practical ability to refuse the technology? They're effectively being gaslit by it by a computer. Hey, friends, you probably knew that TextControl is a powerful library for document editing and PDF generation, but did you also know that they're a strong supporter of the developer community, and it's part of their mission to build and support a strong community by being present, by listening to users, and by sharing knowledge at conferences across Europe and the United States. If you're heading to a conference soon, maybe check if TextControl will be there. Stop by and say hi. You'll find their full conference calendar at textcontrol.com. That's T-E-X-T, control.com.
Introduction of Dr. Chinasa Okolo and AI governance
00:00:56
Speaker
Hi friends, it's Scott Hanselman. It's another episode of Hansel Minutes. Today I have the pleasure of talking with Dr. Chinasa Okolo. Chinasa T. Okolo is a doctor of philosophy in computer science from Cornell, and she is an internationally recognized researcher, a strategist, and a policy advisor on AI governance and safety for the global majority. How are you?
00:01:15
Speaker
I'm doing well, Scott. I'm very happy to be on your podcast. How about you? i'm I'm getting there. I will say that every day it is a new piece of confusing information about
Public concerns on AI's impact on jobs and development
00:01:24
Speaker
AI. People are generally freaking out. And I think that the freak out is not just happening amongst computer scientists like myself and like yourself, but it's starting to leak into government and it's starting to leak into like regular people because people believe that AI is coming for you and it's coming for your job.
00:01:42
Speaker
And it's a very pessimistic time, I feel. Do do you think that, are you pessimistic or you optimistic? Yeah, I would say for me, like I am optimistic, but I'm very cautious in terms of just generally, you know, a lot of the current sentiment around AI, but also just generally how it's pushed more broadly in the media.
Responsible AI: Cautious optimism and interdisciplinary approaches
00:02:01
Speaker
But I definitely would, you know, would say for people that, i may not be the most technically sound or technically advanced. you know Whenever they find out I'm a computer scientist or AI researcher, they always ask, like it is it ever going to get to a place where it actually replaces me? Do you think it's going to take over the world?
00:02:15
Speaker
um do you Or do you think it's going to, let's say, um significantly impact us all? um And so i would say I don't necessarily think so. I think you know humans, honestly, are very much so in control of how you know AI is developed, shaped, and deployed. And so we really just have to do it responsibly.
00:02:33
Speaker
Yeah. I feel like AI, of course, it's it's having a branding moment. And when we talk about AI, we're generally talking about large language models, but also machine learning and deep learning and all of those things. But when you say AI to a regular person, they think chat GPT. Exactly. When you do your work in in policy advising, when you're talking to like you know members of Congress or people in in government, and you say AI, are they also just thinking chat GPT or are they developing their own sophistication?
00:03:03
Speaker
Yeah, definitely. would honestly say it's still chatbots and other, let's say like language ah chat interfaces for the most part, because that's really how most, um I would say, a significant amount of policymakers got introduced um to AI technologies in the first place. However, there have been many governments i and other you know many governmental entities using predictive models um and you know other very kind of lower or basic AI or machine learning systems. for decades now. And so I think that's also an understanding of where, you know, this kind of automated sense, you know, of AI or whether it be machine learning has come into play.
00:03:39
Speaker
um For the most part, particularly when it comes to, let's say, like delivering government services, you know, trying to predict, um you know, what the economy and, you know, what issues can happen throughout there. And so it's still mostly chatbots, but, you know, there's a growing understanding about all the different areas where AI is and machine learning um are being impacted.
Policymakers' understanding of AI and its complexity
00:04:00
Speaker
Now you have, of course, an advanced degree and an expertise in computer science, but you also, your work spans, you know, ethnographic field work. You talk about healthcare, you're thinking about things in the, in the, you know, the emerging South and the global majority.
00:04:18
Speaker
People seem to want to take, and this is me speaking as someone who does not have an advanced degree, ah degrees and specialties and put them in silos. Like you're a computer science expertise, or you know you know about AI. But AI being presented as universal is very much cross-cutting.
00:04:37
Speaker
Do you think we're teaching it wrong? We're explaining it wrong? Because it affects infrastructure and safety and and healthcare and all these other things. It's not just a computer science thing. Exactly, and that's a great point. you know it took me actually getting into policy just to see you know just the different side of AI and actually you know being at a place where um i started off my career at think tank and most people there are literally economists, political scientists, some sociologists, you know maybe some psychologists or other humanities or social scientists, but it's very rare for computer scientists to be at you know a place like Brookings, for example, or other many many other think tanks as well.
00:05:15
Speaker
I mean, so I feel like though those perspectives have really enhanced the quality of my work, but also really had given given me an opportunity to interact with different policymakers or different other stakeholders across policy, civil society, et cetera, um that I wouldn't have you know had if I just went straight into a postdoc at MIT, which was actually which I was planning to originally do before I kind of like got very and um entranced by the world of policy.
00:05:44
Speaker
Do you, and this is again speaking from a place of ignorance, but I worry that the people who are making the policy decisions have just such a limited understanding of the complexity of these systems.
00:05:58
Speaker
And then the complexity of these systems is only enhanced when you drop AI on top of it. It just, AI makes me feel like everything is so, so much more interconnected than we already knew it was. Yeah.
00:06:09
Speaker
Does that concern you? Are you just doing the best you can to educate our leaders? Yeah. So, I mean, you know, AI, along with many other technologies are integrated in and a lot of different
Rapid AI integration: Sector challenges and user agency debate
00:06:20
Speaker
fields. I think this is maybe for the first time it's, let's say in my lifetime, i just turned 30. So, um, uh, I did come up through, I guess, a digital revolution in a way with the internet and all that stuff.
00:06:31
Speaker
um And it's, you know, continues to have an impact, you know, even how many years, the decades later. and so I think when it comes to AI, just because one, it's moving so fast, but also, again, there are so many different sides, you know, outside chatbots, which again, most people...
00:06:47
Speaker
um have interacted with for the most part. um We're seeing lots of issues around agentic AI. you know Those capabilities honestly are still very much so unknown. And I don't think that you know policymakers will have the respective or sufficient expertise, particularly even in countries like the US s that are leading or at the frontier of AI development, for at least another decade or so, because I don't think there are really incentives ah you know for ah people with technical expertise to go into government, not just the financial and incentive, but just generally um you know career growth, career progression, actually feel feeling valued um you know for that expertise as well.
00:07:24
Speaker
Yeah. um i'm I'm fascinated with the concept of explainable AI, right? Explainable AI means how do we explain this system to the user? And i when I started teaching myself and teaching others about AI, I tried to think if I could just explain it to them, they would understand. But now I'm realizing that maybe we should ask a different question, which is, should this AI have been imposed on the user at all? Exactly.
00:07:51
Speaker
What do you think about that perspective? Yeah, I think it's a little bit of both. I mean, i'm explainability was something that was a big part of my PhD work when I was at Cornell. And it really came about because we just, um I did all my core dissertation work pre you know the release of CHAT-GBT, the public face inversion. um you know, in late 2022. And so um why do I was doing my studies in 2020 in India um and also had the chance to actually go in the field and back ah in 2022. And essentially the community healthcare workers did not know what AI was. And this is really concerning to us just because we knew that these tools were being rolled out to them. And also they were getting a lot of pressure from the Indian government to adopt these technologies um alongside disgenerally with having increased responsibilities due to due to the pandemic. And so because they didn't know what AI was, um there could be
00:08:37
Speaker
but if Let's say they are using an AI tool and it presents them, you know, like a false decision. And just because, you know, they kind of have, they so put a lot of value or trust into AI systems, they actually kind of like,
00:08:52
Speaker
let's say, don't go with their gut feeling or actually just like you know leverage their domain expertise and defer to the AI. And that could actually have very detrimental detrimental outcomes. And so really, you know explainability, if we help them if if we explain to them you know what AI is and they also get explanations for how this decision was made by a system, um they can then it can help in these very tricky instances and also really ah maintain their autonomy um as they continue to leverage um and use these AI systems in the future.
00:09:23
Speaker
Man, that just lit up all the neurons in my brain with like a ah a fan out of like three different questions. Just because something is explainable, can the explanation create agency if the user themselves, in this case Indian healthcare care workers, has no practical ability to refuse the technology? They're effectively being gaslit by it by a computer.
00:09:45
Speaker
Yeah, and I would say, fortunately, in those like frontline healthcare cases, you know usually the community healthcare worker can leverage her domain experience. Like, oh, okay, I've seen this case or these symptoms like grouped together a couple of times already. And I don't think that the AI system has actually, you know let's say, been trained or updated to understand these nuances, particularly in my context. So actually, you know, I will make the final decision rather than having AI
AI in healthcare: Autonomy and productivity concerns
00:10:11
Speaker
do that. And I think that in many cases, you know, um I would, I, mean AI is not actually being used as a final arbiter um a business of a decision.
00:10:20
Speaker
um in some cases it is, unfortunately, but I think and there's a little bit more flexibility in these frontline healthcare settings. But again, not every frontline healthcare worker in India or even you know across different countries in Africa, the Caribbean, have access to these technologies in the first place. And so it really is still the community healthcare worker making that final decision.
00:10:39
Speaker
Now, you mentioned that you just turned 30 and you finished your PhD in you about three or four years ago. And now you were thinking about doing postdoc. But you exist as a digital native in this interesting historical place. While I'm 20 years older than you, I was here as it was getting built. But I'm i'm not really. I'm a different kind of digital native.
00:11:01
Speaker
What do you think about the PhD students that you are that you teach and the ones that are like 10 years behind you? Are there, is there, and this might be a spicy take, but is their expertise going to be less deep because their, their thinking is more shallow or more AI augmented? And does that concern you as someone who has deep expertise in a specific area?
00:11:22
Speaker
Yeah, definitely. I mean, there are so many different ways. Like, you know, I'm a very active user or user of Reddit and, you know, I'm on different forums related to academia. how I frequent um the art professors subreddit and they're always mentioning how, you know, students cannot like, let's say, like operate a desktop computer or, you know, different yeah programs or applications. And that was something like I learned. I'm like very young and also just, you know, throughout my journey of experimenting with um or using computers on a daily basis or neared near daily basis as a child.
00:11:53
Speaker
I mean, so I think this is one sort of literacy that, you know, is definitely declining, but also when it comes to, you know, relying heavily chatbots large language models, you know, to augment or actually do a lot of your work, you definitely, you know, lose a lot of critical thinking skills because you're outsourcing some of the fundamental skills needed, you know, let's say like searching literature, actually reading the literature and understanding the different nuances in terms of being able to interpret what an author is saying and not just relying on what the chatbot is telling you. And then also synthesizing that, let's say like you're doing a literature review, synthesizing that and presenting your own interpretations of that work.
00:12:32
Speaker
And again, because we know that a lot of um LLMs provide this homogenized view um of the world more bodily, I think it makes research itself, um it will it kind of devalues the research process a little bit and also kind of weakens the empirical contributions to the computer the field of computer science, you know, and also the all the subfields more broadly. Yeah. My wife is currently on her second year of her PhD. She went back to school. She's the same age as I. And she has a very negative feeling towards AI, as do all of her PhD advisors, and they are being advised to just stay away from it, not even touch it.
00:13:12
Speaker
But at the same time, I know that someone out there right now, we don't know who they are, where they're located, is vibing their PhD. Right. Yeah, definitely. I mean, I've heard so many different stories. I mean, there's a lab at Stanford. I think they do work on like digital economy stuff. Like I just read an article and the professor, the PI who's leading that lab mentioned that they're literally producing like a paper every week um due to the to due to how they're leveraging, you know, authentic AI tools and in other AI tools as well. And I think like, you know, One, like we don't need that much research. And also two, it's just like, how can I trust actually what you're doing? um Just because knowing that, you know, professors themselves have so many responsibilities and, you know, you're trying to train up and advise students like, and they, you know, they may be advancing their skills, but you still have to like do a lot of this work ah manually and learn about the process and rather than having it outsourced.
00:14:07
Speaker
Yeah. A good friend of mine, ah Dr. Mark Rusinovich, wrote an application called RefChecker, which basically looks for fabricated references and citation errors, and he's run it on a huge corpus of material. And he says every day it's getting worse.
00:14:23
Speaker
the The references are being fabricated and made up, which is hugely problematic. Yeah, I just actually found a fake reference of mine. um It was cited in some food undergraduate dissertation at a university in the Czech Republic. And actually reached out um to the chair or respective committee of that because like, you know, this is not my research, you know, and I know what they were referencing. interesting. So the research was fabricated, but your name was applied to research that you didn't do?
00:14:48
Speaker
Exactly. Yes. Well, it so the name of the work was right, but the journal it was attributed to was fake. It was something, it was like something they published at Brookings and, right right right you know, they just made this, the whatever source they were referencing stuff they were using was not right.
00:15:04
Speaker
Yeah. um I want to go back to the, the the healthcare care workers that you studied because, so these are community healthcare care workers. They're already overburdened. They're already underpaid. How do we, and and we, where we is, don't know, a company, society, humans, how do we tell whether AI is actually helping them or just giving them another thing, another device, another form, another system to maintain? Mm-hmm.
00:15:28
Speaker
Yeah, so this is also something that interesting that came up in my work because the community healthcare workers, ASHAs, that's what they're called in India, um they understood that they would actually be responsible for learning how to use AI systems and also troubleshoot them. And so, because usually, you know, they're you alone in the field themselves, you know, they're using this mobile device. It may be like a basic or kind of mid-level ah smartphone, not super advanced. and um And so this is also an added burden, you know, to their work. I think that there's also just a cognitive, um there's a difference in terms of how ai systems are used or when it comes to getting predictions rather than using things like scales, um you know, or just like a
00:16:11
Speaker
measures you know to weigh and you know ah measure a baby because these are basically definitive um kind of outcomes rather than then trying to understand like all these different aspects that produce a a decision that may be, let's say, 85% confident. it And so it's also kind of a mental model they have to adapt to as well when using AI.
00:16:31
Speaker
Yeah. So this word productivity gets used a lot and it's being applied to all industries, healthcare as as well. We're trying to like hyper optimize with, you know, we we were trying to create productivity gains.
00:16:45
Speaker
Is that what these folks need? Do they need productivity gains? And if so, who gets it? Is it the worker, the patient, the government, or is it the company that sells the technology?
00:16:58
Speaker
Yeah, so when it comes to these contexts, I would say the healthcare workers themselves are pretty productive already. think it's just that they're underpaid. I remember that Asha's mentioning to us, okay, if we have this AI tool, maybe the government can actually see the full range of services that we're doing and increase our pay. Because usually know a lot of healthcare systems, particularly across Africa, are reliant on foreign aid funding. India has little bit more of a kind of self or government-friendly system that doesn't heavily rely on external funding as well.
00:17:31
Speaker
um But those are issues that have increased you know due to the different pandemics that are happening or epidemics that are that have been happening um across the world, but also just generally the reduction in global aid funding as well. And so i think really it's just like providing these workers with sufficient training and funding will also just generally increase their livelihoods and just more generally increasing the um a number um you know of community healthcare workers because healthcare care work um is definitely devalued um across the world, but also just like there's not enough workers um for the respective need that we have. And so I think that's also somewhere that investments have to be concentrated in as well.
00:18:10
Speaker
Maya, this is just a coincidence. I don't always bring my wife up in every interview, but my wife is is doing her PhD research on the use of Ubuntu in the global South and if it can be applied into the you know the Western culture and Northern Hemisphere.
Global majority vs. developing world: Terminology and influence
00:18:25
Speaker
Now, you deliberately, in the research that I've done on you, you deliberately use the phrase global majority. Yes. What does that phrase do or reveal or change that terms like developing world or the even the global South hide?
00:18:40
Speaker
Yeah, I would say it just generally, you know, it's in the name, but it makes you understand that, and you know, these people that you kind of, that people have put mostly at the margins actually constitute a significant part of the world's population. you know not necessarily, let's say, um a significant part of the world's GDP or let's say economic productivity or, you know, other kind of, let's say like economic indicators that people kind of use more so to imply and worthiness for the most part, but it's still, You know, the numbers matter. And I think that this is something we have to really be considerate of just because, you know, a lot of things really revolve around, you know, U.S. pop culture.
00:19:20
Speaker
I mean, we're just only a that's a we decent majority, you know, of the world's population, but it's still relatively small once we can group all these Western countries together. And so I think this is what um my youth my use of this word or phrase, um you know, intends to convey.
00:19:38
Speaker
So it certainly carries weight. I mean, majority, like words matter. People can complain about labels all they want, but for the purposes of the conversation, calling it the global majority is a reminder of the weight behind it, which is all of these different groups. Do you think, though, that grouping India, and Nigeria, Brazil, these are dozens of different societies together, risks creating another abstraction that is designed by academics? Yeah. Yeah, obviously, you know, and just because, you know, these these countries themselves are so diverse, not just across, you know, countries, but within them. I say like India, Nigeria, very similar, you know, hundreds um of tribes, hundreds of languages, um you know, ev varying you know amounts of socioeconomic progress or development throughout the countries themselves.
00:20:21
Speaker
and You know, even though Nigeria and India, that they still very much so struggle with poverty, um you know, more broadly. And so, but again, i think that um there definitely can be much other terms other terms created to just ensure that we're not homogenizing this very diverse set of people. i definitely encourage people to develop them and and use them.
00:20:42
Speaker
Yeah. So you're another one of your papers describes a global AI divide, describing inequities and and inequalities. that are kind of legion, infrastructure, education, ai development. And then you've also said in an essay in 2025 that AI is not Africa's savior.
00:21:01
Speaker
But if you go on Twitter, the dumpster fire that is Twitter, there's AI grifters of all flavors from all over the world. Every country's got their own flavor of ai grifter. They're saying that this is it. But I remember a couple of years ago, it was crypto that was going to be Africa's savior.
00:21:18
Speaker
So as we continue to try to parade technologies that are going to save us and save the global majority, it doesn't feel like technology is the is the problem or access to it.
00:21:30
Speaker
What is the real problem that we're not
AI in Africa: Infrastructure needs vs. techno solutionism
00:21:32
Speaker
talking about? Yeah. I mean, I would honestly literally just say like where the money is going. um I had an interview with a Nigerian newspaper a couple months ago and literally I just, you know, I was looking through, I follow lot Nigerian organizations on Twitter and and they were breaking down some of the budget allocations that was proposed for the upcoming fiscal year. And just generally, you know um a lot of the money ah you know were going to agencies and not necessarily being trickled down. And just generally knowing that Nigeria has a very big problem of corruption and seeing this firsthand, um through different accounts of you know politicians hoarding pallets of money um you know in their basements, what the note's actually rotting because they're not able to use that money. um and And just knowing more broadly that you know there's money being allocated towards healthcare, care but they're spending it on cars for hospitals rather than actually beds and medicine. and and patient care. And so I think um in most cases, it's really, you know, if you're if you focus on the fundamentals um and actually, you know, sufficiently allocate that money to ensure that it's, you know, being, it's going where it's supposed to go, um then you'll see a lot more progress rather than trying to slap AI, you know, onto a clinic that it's understaffed and actually does not even have, let's say, internet connectivity or actually stable electricity. And so
00:22:51
Speaker
I always encourage African governments, stakeholders, focus on the basics, and then you can use AI um to augment those capabilities. Yeah. So focus on the basics.
00:23:03
Speaker
What there are problems in Africa, there are problems in the global majority that are often described as problems that can be solved with AI. But if you break them down, they are connectivity problems. They are electric electricity problems. Like electricity is like the problem right now on the continent. And then not to mention wage problems, institutional problems, political problems.
00:23:23
Speaker
When someone shows up with an AI thing, whether they be an AI grifter on Twitter, they usually will say, look, We're with these poor people and they're in a well. And we added some AI and an Arduino. and there's And there's a more offensive phrase, but I'll just say inspiration fluff. Yes.
00:23:42
Speaker
Inspiration fluff. yeah I even heard heard heard someone call it, I love this term, techno solutionist theater. Right. Ooh, look at me. And then you pat yourself on the back and then you leave the Arduino running on that well. And I've just saved a village. What is a test? What is your test, Dr. O'Kolo, for distinguishing a valuable AI intervention from techno solutionist theater?
00:24:04
Speaker
Yeah, honestly, it's really hard to say, because I feel like when it comes to AI deployments more broadly, particularly for, let's say, like these rural low resource areas, The evidence base is not there yet. I mean, it's growing, um but it's not strong enough just to understand like the impact of long-term deployments. i mean A lot of my dissertation work is rooted in the field of ICTD, information and communication to technologies these for development.
00:24:27
Speaker
and And a common problem, you know, even before um these modern day AI tools became really prevalent was that you know researchers you know would develop, let's say, kind of or this Raspberry Pi solution or some technology solution or mobile app, and they would just you know like deploy it throughout the length of the study, whether it be a couple of weeks to a couple months or maybe up to a year.
00:24:48
Speaker
um And then after that, you know they it's ah essentially abandoned by the community because there isn't necessarily sufficient technology transfer or even just generally resources to support um the longevity of these respective solutions. um And just generally the infrastructure, um you know, it it doesn't, it's not there to support long-term use and adoption. And so um i would say for um ai solutions, we have to really consider some of these underlying factors. You know, you're going to have, you have to realize like, okay, if you want the solution to be used long-term, um you need to set up some kind of, you
00:25:22
Speaker
Cloud support so it can actually run. You need to pay someone um or a group of people to actually you know help with debugging and other issues if you're not actually going to be involved in it. um and this Generally also, you know, be able to willing to come back in the field from time to time to see how you can ah adapt, um upgrade and extend the life of the solution just so we can actually run or maybe even encourage the government to invest um and in helping in supporting the project as a whole. So these are some I would love to see as AI solutions start to get deployed in low resource areas. Yeah, it is a huge problem because they got their case study and they did their polished fancy video and then they left.
00:26:03
Speaker
And then if you would return three years later, you'll find that those devices are just sitting there on the side of the road. So then if ah if a donor, and maybe that's not the right word, but a technology company arrives with an AI solution, then Who originally defined the problem? Are we flooding the market with solutions that have either the the wrong solution for a poorly defined problem? Or it was more like push technology as opposed to pulled? Like who asked for this is a question I find myself wondering.
00:26:32
Speaker
Oh yeah, I mean, and that's a great question too, because I think, you know, a lot of times it looks good. it's It's good PR, you know, to appear that you're solving a problem for a community. And maybe be you may be solving just like one aspect of a larger problem more broadly. And, you know, fortunately, a lot of these big tech companies do have resources and let's say like socio-technical expertise in-house um from philosophers, anthropologists, et cetera, sociologists you know to help um leverage you know participatory design mechanisms to ensure that they're actually working with these communities you know from the onset of this work. A lot of times it doesn't happen, but it it definitely should.
00:27:08
Speaker
Yeah. um A lot of people are talking about sovereign AI and sovereign
AI sovereignty in Africa and global supply chain challenges
00:27:13
Speaker
clouds. You know Germany's got a cloud and China wants their own cloud. um And then you've got folks like Lelapa AI and Pelanomi Moloa down in South Africa who are doing their own local models.
00:27:25
Speaker
What is the minimally viable form for sovereignty? Is it just local compute? It's running on a computer. Is it local models like our universities in this country? Is it local evaluation? is it that it's in a particular, there there's more languages on the African continent than there is pretty much anywhere. Like, is it about local language models or is it something, or is it just the power to say, no, no, thank you. I don't want any of it.
00:27:49
Speaker
Yeah, I would say honestly, it's a lot of basically what you just mentioned, but it really comes down to um the last thing you just said. And that underlying um kind of philosophy is just generally autonomy. You know, being able to, um you know, have the power to independently um say you want to adopt these systems, also being able to independently develop that capacity, you know, um to, you know, push out AI models, ah you know, curate your own data, um also compensate people equitably, you know, for that labor.
00:28:19
Speaker
um and curating the data sets in also, let's say, just generally having your own AI clusters, hyperscalers, data centers, whatever you want to call it. I think you know sometimes when it kind if you think about sovereignty more broadly, everything is still much so pretty connected. um like you know The SEPSI cables, they connect from country to country. that You can't just have a SEPSI cable that's just like only relegated to traffic in your country. i mean um African policymakers or ah dictators tend to, you know, cut off internet at times, but that's a whole other thing as well. um But also the chips that we rely on come generally from, you know, mostly China or or NVIDIA or, you know,
00:29:03
Speaker
ah ASML as well. And so there isn't enough diversification the value chain um for those kind of the infrastructure of AI to be really sovereign itself. And so I think we have to kind of co-hire up i mean and think about it more broadly. Okay. That's a great point because we, like you, one can be pro or anti-globalization, but the facts are everything's connected.
00:29:23
Speaker
And if a widget or a cell phone shows up in an African country, it is not really possible that it be entirely built from scratch. You know, the, the, the lithium didn't come from there. Like what all the different parts all the way up.
00:29:37
Speaker
So then can you do, can, is it even possible to have sovereignty if your cloud, your chips, your models, and your technical expertise still come from American and Chinese companies?
00:29:48
Speaker
Yeah, I mean, i would say in a way, I mean, i've I've seen a lot of efforts, you know, from European countries, particularly, or european European governments, because they've seen a lot of the issues with big tech and having just like this this dominance in their respective government systems. France actually released a super interesting like open source, a kind of office set of a suite of software, which they call La Suite, I believe. And so I thought it's super interesting.
00:30:13
Speaker
um i would love to see, you know, more countries, particularly in Africa, i think about this, but again, that expertise or just generally capacity is not there yet to divest, you know, from, let's say like, it's usually, i don't think they're really adopting the Google Workspace, it's more so Microsoft, a lot of times like Soho and and other kinds of software is where as well. um But I think like, you can get to a sense of sovereignty, but understanding like if you if you don't have, let's say again, like the minerals to create the trips yourself and also the companies that can create you know the GPU clusters as well, um then you are still limited. But I think you can develop these alliances where you're aligned, um you know more generally, you know and ah values and and other areas um to ensure that you know sovereign sovereignty is done on your respective terms.
00:31:03
Speaker
So the podcast is my own, but I do work for a big tech company in my day job. So I'll put them on blast. When a big company says they want to support responsible AI in the global majority, what would you as a policy expert demand beyond funding programs and opening offices and putting representatives on
Independent governance in Africa and future infrastructure vision
00:31:22
Speaker
panels?
00:31:22
Speaker
Yeah, definitely. um I would say really just independence, particularly when it comes to like the actual how governance decisions or governance frameworks are developed themselves. um For example, I did some work with Nigeria on their national strategy.
00:31:36
Speaker
And there were stakeholders, you know, from Meta, Google, Microsoft, um Also, like UN agencies um involved um in that process. And um I'm sure a lot of these people that were in kind of like in the working groups were working in their individual capacity, but, you know, their affiliations and also these companies did fund um the strategy development itself. And I i definitely think that and Even though you can say it's unrestricted or we don't have you know these specific or there aren't specific outputs we want you to have, um there still is an influence because there's a certain kind of, um i don't know, mindset around like satisfying the funders or making them happy. And so I would just really you know advocate that there'd be really true independence and also like external kind of like more neutral stakeholders that can advise on this process to ensure that you know, there isn't too much influence, you know, from this company or that company um in defining like what the exact strategy or framework or law proposes. And so i think this is something that will be of really high concern i'm in Africa vve because we've seen already how big tech has, let's say, kind of impeded ah governance processes when it came to Brazil's AI Act.
00:32:48
Speaker
So I want to end on a positive note, but I also want to put you on a little bit of a spot. Try to describe a a global majority success story, whether it be African or Brazilian or Indian, from 2035.
00:33:03
Speaker
Who built it? Who owns it? Who benefits from it? And how is it different than just importing some Western system? Yeah, definitely. Yeah, I'm really big on infrastructure right now. And I think that is one thing I would love to see expand a little bit more. um and And there's a company, um I would say, Omni, they're based in Kenya, and they've been doing a lot of work um with Barbados and actually creating kind of like these micro data centers. And um I think they kind of like fit in um like a shipping container and they're modular as well. And so I would love to see something...
00:33:36
Speaker
I look across Africa and also even across the Caribbean and and Pacific Islands just due to kind of the nuances or challenges of those respective environments. like greater penetration um of this kind of infrastructure across these regions. But also just generally, I think within Africa, this this would be accompanied by greater ah like electrification um and more, let's say, like ah renewable energy leverngaging leveraging like those who like to technologies for um a solid functional grid. um And then also, again, like um connected a connected cloud across Africa, leveraging this micro data center concept as well.
00:34:14
Speaker
Very cool.
Conclusion and resources on Dr. Okolo's work
00:34:15
Speaker
You mentioned Amini. I had Kate Callow, the CEO of the on the, on the show. She was episode nine, nine, two, if folks want to follow up on that. So I love that you brought that up as well.
00:34:25
Speaker
Thank you so much, Dr. Okolo for chatting with me today. Thank you, Scott. Very happy to have chatted with you. We have been chatting with Dr. Chinasa T. Okolo. You can check her out at chinasatokolo.com. I'll put a link in the show notes. You can learn all about the research that she's doing and check out her media kit. This has been another episode of Hansel Minutes, and we'll see you again next week.