From Deep Learning Indaba to Algorithmic Sovereignty with Benjamin Rosman | [Sidecar Sync Episode 121]

Summary:

If you think the future of AI is being decided only in Silicon Valley or Beijing, this conversation will stretch your perspective. Amith Nagarajan and Mallory Mejias sit down with Benjamin “Benji” Rosman—Professor at the University of the Witwatersrand, Founder of the Machine Intelligence and Neural Discovery (MIND) Institute, and recently named to the Time 100 Most Influential People in AI for 2025—to explore how Africa is building its AI ecosystem from the ground up. From the origin story of Deep Learning Indaba (now the largest machine learning summer school in the world) to the concept of algorithmic sovereignty, Benji explains why diversity of thinking fuels scientific breakthroughs, how constraints drive innovation, and why leaders must balance exploration and exploitation in an era of exponential change.

Timestamps:

👥Provide comprehensive AI education for your team

https://learn.sidecar.ai/teams

📅 Register for digitalNow 2026:

https://digitalnow.sidecar.ai/digitalnow

🤖 Join the AI Mastermind:

https://sidecar.ai/association-ai-mas...

🎀 Use code AIPOD50 for $50 off your Association AI Professional (AAiP) certification

https://learn.sidecar.ai/

📕 Download ‘Ascend 3rd Edition: Unlocking the Power of AI for Associations’ for FREE

https://sidecar.ai/ai

🛠 AI Tools and Resources Mentioned in This Episode:

ChatGPT ➔ https://chat.openai.com

TIME Article ➔ https://time.com/collections/time100-ai-2025/7305865/benjamin-rosman

Deep Learning Indaba ➔ https://deeplearningindaba.com/2025

Lelapa AI ➔ https://lelapa.ai

MIND Institute ➔ https://www.wits.ac.za/mind

👍 Please Like & Subscribe!

https://www.linkedin.com/company/sidecar-global

https://twitter.com/sidecarglobal

https://www.youtube.com/@SidecarSync

Follow Sidecar on LinkedIn

⚙️ Other Resources from Sidecar: 

• Sidecar Blog
• Sidecar Community
• digitalNow Conference
• Upcoming Webinars and Events
• Association AI Mastermind Group

More about Your Hosts:

Amith Nagarajan is the Chairman of Blue Cypress 🔗 https://BlueCypress.io, a family of purpose-driven companies and proud practitioners of Conscious Capitalism. The Blue Cypress companies focus on helping associations, non-profits, and other purpose-driven organizations achieve long-term success. Amith is also an active early-stage investor in B2B SaaS companies. He’s had the good fortune of nearly three decades of success as an entrepreneur and enjoys helping others in their journey.

📣 Follow Amith on LinkedIn:
https://linkedin.com/amithnagarajan

Mallory Mejias is passionate about creating opportunities for association professionals to learn, grow, and better serve their members using artificial intelligence. She enjoys blending creativity and innovation to produce fresh, meaningful content for the association space.

📣 Follow Mallory on Linkedin:
https://linkedin.com/mallorymejias

Read the Transcript

🤖 Please note this transcript was generated using (you guessed it) AI, so please excuse any errors 🤖

[00:00:01:10 - 00:00:10:13]
Mallory
 Welcome to the Sidecar Sync Podcast, your home for all things innovation, artificial intelligence and associations.

[00:00:10:13 - 00:02:03:10]
Mallory
Today we've got an interview episode of the Sidecar Sync lined up for you. If you think the future of AI is being decided only in Silicon Valley or Beijing, this episode might change your perspective. Today we're joined by Benjamin Rosman, also known as Benji, professor of computer science and applied mathematics at the University of Vidvatiran in Johannesburg, founding director of the Mind Institute and recently named to times 100 most influential people in AI for 2025. But this conversation isn't just about big models or global tech races. No, it is about something much deeper. We talk about building the African AI ecosystem from the ground up. Why diversity of thinking is essential to scientific progress. How constraints can actually drive innovation and what organizations, including associations, can learn about community building in an era of exponential technological change. We also dig into the origin story of deep learning in DABBA, now the largest machine learning summer school in the world. Why algorithmic sovereignty matters just as much as data sovereignty. How to think about AI strategy when everything feels like it's changing daily. And why exploration and exploitation, a concept from reinforcement learning, might be the best mental model for leaders navigating this moment. If you listening to this podcast care about AI, global innovation, or how to build resilient future-ready communities, this one is going to stretch your thinking in all the right ways. Everybody, please enjoy this conversation with Benjamin Rosman.

[00:02:03:10 - 00:02:28:07]
Mallory
 Benji, thank you so much for joining us on the Sidecar Sync podcast. We're thrilled to have you here and have this conversation. You were just named to Times 100 Most Influential People in AI for 2025, so I've got to say congratulations on that. And I'm hoping you can tell our audience a little bit about your background and what got you to this moment here in AI.

[00:02:28:07 - 00:02:32:07]
Speaker 7
 Cool. Well, it's great to meet you. Thanks so much for having me here.

[00:02:33:08 - 00:02:38:20]
Speaker 7
 I'm based in South Africa in Johannesburg.

[00:02:40:20 - 00:02:51:07]
Speaker 7
 My background is very much in computer science, applied maths. I spent about six years doing my master's and PhD in Edinburgh in the UK.

[00:02:52:11 - 00:02:57:05]
Speaker 7
 And I've been back in South Africa for well over a decade now.

[00:02:58:24 - 00:03:20:03]
Speaker 7
 And really, I care about a couple of things. One is building and advancing AI as a core fundamental science. And two, about building the African AI ecosystem and making sure that what happens in the field kind of translates to impact out there in the world.

[00:03:21:09 - 00:03:47:10]
Speaker 7
 And really doing that through different kind of vehicles of bringing people together. And so I've been involved in a number of initiatives and starting a number of initiatives focused on kind of really doing just that. I'm based at a university here in Johannesburg called the University of the Witte Waddesrand, which we call WITs. And kind of try and do everything from there, although wearing all these different hats.

[00:03:48:14 - 00:04:16:00]
Mallory
 You mentioned building the African AI ecosystem. And I wanted to point out something interesting I read in your time profile, which is that you were tired of being asked about the African angle on AI. You responded, and I think this is pretty clever, what's the unique American angle on AI? And that's something I myself, as an American, would have a hard time articulating to you. So can you talk about that frustration perhaps and why that matters to you? Sure.

[00:04:16:00 - 00:04:27:03]
Speaker 7
 Yeah, I think it's interesting to have seen this evolution because you kind of rewind 10 years or so and there really wasn't an African AI ecosystem.

[00:04:29:05 - 00:04:33:20]
Speaker 7
 And there were the pockets of this all over the place and we can get into exactly what's happened subsequently.

[00:04:35:03 - 00:04:54:22]
Speaker 7
 But over time, different things have happened. And really, a few years ago, if you'd asked kind of what the African angle was, then there'd be very specific things that I would answer. So I'd say, well, you know, there's a lot of work being done in African languages and natural language processing.

[00:04:56:04 - 00:05:28:13]
Speaker 7
 There's kind of 2,000 plus languages spoken across Africa and they really aren't standard ways to transcribe or translate kind of most of them. So that would be a very strong thing because now there's a community that's been built around that. There's areas in agriculture and healthcare and so on. And I think that for a while, that was really where the emphasis is. And I think to a large extent, that's still where a lot of the emphasis is in the African community.

[00:05:30:01 - 00:05:32:23]
Speaker 7
 But that's not to say that's where the emphasis should be.

[00:05:34:17 - 00:05:40:22]
Speaker 7
 I think there's a kind of boxing in that then happens that like, oh,

[00:05:42:04 - 00:06:19:00]
Speaker 7
 if those are the kind of core areas that people work in, then that's what should be funded, that we should encourage people to work in. But really, that's kind of historically, history being the last few years in this case, the reason for that is that's where some of the first successes happened in the field. Now, you don't want to double down and only be investing in where you've already had successes. We know with the amount that's happening in AI, the opportunities in like every single sector that AI could have impact in healthcare, but also finance, but also education, but also climate change. Whatever you want to look at.

[00:06:20:22 - 00:07:24:16]
Speaker 7
 And the same is true anywhere in the world. And so I think it's really important that we kind of frame the discussion that we want. We need to be capacitating different parts of the world with the same kind of technologies. I believe that an important component of this is the fundamental research for a number of reasons. But ranging from there's huge problems in the field that, again, the whole world faces, solutions could come from anywhere, from people with different kinds of backgrounds, different contexts, ways of thinking, that sort of thing. But by the same token, what's to say we don't find like someone in an incubation hub somewhere suddenly thinks of something interesting and new that could have an impact on people's lives, maybe here in Africa, that then could roll out across the world. So I think kind of boxing this in as saying there's like a unique angle to AI in a particular part of the world, I think is very limiting from like really what it could mean for people in that region, but also generally across the globe.

[00:07:24:16 - 00:07:54:01]
Mallory
 I think that's a really poignant point. I'm curious and also I want to add a little caveat here. I feel like when we're talking about AI in Africa, it almost sounds ridiculous, right? A continent of one and a half billion people in over 50 countries. But generally, what would you say has been the reception, maybe with your colleagues in the area that you live in, with AI adoption? Are people, do people seem excited about generative AI? Do they seem cautious, nervous, scared? What are you seeing?

[00:07:55:18 - 00:08:01:11]
Speaker 7
 I think it's probably quite a similar kind of distribution over hype that you're getting all over the place.

[00:08:03:19 - 00:08:43:19]
Speaker 7
 All the big corporates, the joke is a bit like we need an AI team. Why? I don't know, but our competitors have one. And I think that's kind of the global discussion. Everyone's going to be adopting things and using things and innovating and changing the way they do business and bringing in chief AI officers. And the same thing that we saw a few years earlier in data science and et cetera, et cetera. I think, again, there's similar kinds of distributions of people that are really excited, people that are kind of quite apprehensive about what this means. There's discussions always come up about what does this mean for jobs? What does this mean for various kinds of risks?

[00:08:44:20 - 00:09:05:22]
Speaker 7
 For me personally, for my organization, my sector, et cetera. So I think that's also, you know, in kind of engagements I've had all over the world, that seems to be similar kinds of things happening, which, you know, might be interesting for people that you think of Africa as being a very different kind of place, but these are fairly universal discussions that seem to be happening.

[00:09:07:09 - 00:09:17:01]
Mallory
 I'm hoping you can talk about an initiative that you co-founded in 2017, which is deep learning in DABBA. What is it? What inspired you to create it?

[00:09:17:01 - 00:09:28:09]
Speaker 7
 Cool. So the deep learning in DABBA, which we kind of call the end of DABBA, although many things are in DABBA, it's a Zulu or also a word for a kind of community gathering.

[00:09:29:17 - 00:09:57:15]
Speaker 7
 And back in kind of January 2017, there were eight of us who, actually at that point, I think we're all South Africans who'd either studied abroad or working abroad in kind of different AI research kind of institutions. And we wanted to give something back to our community. So we figured, okay, what we're going to do is organize like a summer school kind of thing, it will last for about a week.

[00:09:58:23 - 00:10:53:12]
Speaker 7
 We probably don't need funding for this, like we want to do like really a technical deep dive, the kind that we hadn't seen. We'd seen people already starting to do, again, at that stage, more in the data science space, but data science for business and then kind of very consulting-y, speak and that sort of thing. We wanted to do something that was kind of grounded, like for the really like postgraduate students, academics, and then people in industry and startups that were really your deep tech people. So we figured, you know, that eight of us can do the talks and run the sessions and like, let's try be quite ambitious, get people from all over Africa, and maybe we'll find 30 people who would care to attend something like this. So we decided, all right, let's do it this year. This was January 2017. We decided on August and then we do it at my university, University of Nevada, I just rented Johannesburg.

[00:10:55:04 - 00:11:47:15]
Speaker 7
 So kind of fast forward a bit and we had an event with 330 attendees. We had more than double that in applications and we thought, okay, wow, this is a thing. And a few things happened there. One, we got much kind of wider interest and we got like way wider than we'd ever anticipated. But all sorts of interesting things started happening. So first off the back of that, we're like, there's probably demand for us to do this again. But also while we were going through the process, you know, being academics, we had quite a interesting review process where we review, two people had to review each applicant and so on. And we kept seeing these applicants from like places that we didn't even know had universities, in fact, might not have had universities. And they'd send in and we'd look at what they've done. They're like, wow, we don't even know how to do the things that these people are doing.

[00:11:49:08 - 00:12:43:20]
Speaker 7
 So we'd accept people and we kept getting these kinds of emails. And one that I always kind of quote that stands out for me, there's an email that we got that said, thank you so much for accepting me. This is the greatest thing that's ever happened to me. Unfortunately, a ticket to South Africa is five times my father's monthly salary. And then we thought, oh, shit, OK, so we need to go and do some fundraising. And then we we started approaching various corporates and big tech and so on, as we had to raise some money to bring these people. I think in the end, I think there was 60 students that we paid for like everything, their flights, accommodation, food, absolutely everything. And this was just such a hit, it ran for six days, that we kind of repeated it. And so now we've run it. We ran a second year in South Africa and then we thought, OK, this isn't a South African thing. This is Pan-African. And we went the next year to Kenya. And then I think then we had a COVID break.

[00:12:45:05 - 00:13:38:17]
Speaker 7
 But then we went to subsequently Tunisia. We went to Ghana. We went to Senegal this year. We were in Rwanda. We'll be going to Nigeria next year. So we're really trying to move around the continent. The community now kind of spans tens of thousands of people. I think we had 1,300 attendees this year making it the biggest machine learning summer school in the world. And we've kind of spun out a whole lot of other programs along the way. So, for example, after the first one, being a bunch of academics that are somewhat incompetent, we had many left over. And we thought, OK, can we use this in a way to grow this? Because it doesn't seem reasonable to like, you know, it's a big continent. How do we like reach everybody? We can't just make like an infinitely large meeting every year. That's probably not the right way to do this. So we started a program called the Indaba X, kind of inspired by TEDx.

[00:13:39:23 - 00:15:53:13]
Speaker 7
 And we thought, OK, maybe we could do satellite events in other countries. And we stopped ourselves and said, wait, we're a bunch of South Africans. What do we know about other countries? And so this program runs now that organizing teams from different countries can apply to host an Indaba X as a satellite event at some point during the year. The format is completely up to them. So they've run everything from like one day to I think five days, 30 people to 500 people. And it's, you know, they've got to kind of subscribe to our code of conduct and various things like that. And then we give them some money. We help them with some of their advertising. We help them with finding speakers and various things like that. But like at the moment, I think we're supporting events in 47 different African countries. And these all have very different characters ranging from ones that are depending on the state of that country that run from like, you know, I think in South Africa, we've got quite a strong research focus. In Kenya, there's a very strong startup focus when it ran. I think in the DRC, it's more focused on like high school students. So it's really talking to the character of like where that community is. And then what we do is we try and we give some tickets to the main event to each Indaba X. So one that they can use at least or at least one they can use for one of their organizers and at least one they can use for like a prize for one of the attendees. And so we keep getting that that flow coming in from all these different countries. And then many of the like Indaba X organizers have gone on to join our main volunteer organizing committees, which I think there's now about 80 people involved as volunteers organizing this thing every year. And so we've got this really nice dynamic flow of people across the continent that are really getting to know each other, working on projects. We've got sessions now like idea funds. We're trying to encourage people to start companies. We've got sessions on how to apply for grad school or write research papers. We've got sessions in different languages, sessions on policy and governance and ethics. We invite government ministers in the host countries to attend. So it's become this very big dynamic space of all sorts of interesting people.

[00:15:54:17 - 00:16:11:10]
Amith
 Benji, you know, that reminds me so much of the origin stories of many associations that I'm familiar with. Do you consider yourself a membership or association type organization or is it a little bit different with the structure that you guys have built? It sounds familiar, similar to me, but I'm not sure how you guys think of it over there.

[00:16:12:10 - 00:16:21:22]
Speaker 7
 So it's very much kind of running as like a kind of decentralized volunteer based kind of thing. So people aren't really members.

[00:16:23:02 - 00:17:02:05]
Speaker 7
 Like anyone can apply to attend. We've got various mechanisms because we have such a high demand. We've got to kind of have various mechanisms for filtering and for applications to keep it fair. But that's really just to attend the events every year. It's like this is kind of registered as a charity, but that's really just so that we've got an account that we can distribute money to like when the students we need to fund so we can receive money from kind of donor organizations, that sort of thing. So it's really as kind of grassroots and kind of chaotically organized as you can imagine.

[00:17:02:05 - 00:17:03:08]
Amith
 Sure.

[00:17:04:10 - 00:17:28:04]
Amith
 One of the things that comes to mind, I have a bunch of questions about the event and the way you organized it, but it just seems like you've found something deeply community oriented that's also deeply technical in a way that's very relevant and allows people to not only advance in their lives and their careers, but to build their skills essentially. Is that a fair characterization? That's kind of that intersection that you guys found a unique market for?

[00:17:29:04 - 00:18:35:22]
Speaker 7
 I think so. I think seeing the kind of the passion is definitely a big thing. Right. So it's been commented so many times by people that have come from all over the world that they meet these attendees who might be a master student who's also doing community outreach in their local community and also has a startup and also does mentorship and also works for a corporate and this kind of thing. And it's given this kind of sense of community and something that's interesting is before this, I could probably count on one hand to the number of people I knew who knew someone from like five different African countries. And now there's this like community of people that have connections across the entire continent, like really different backgrounds and cultures and languages that have collaborated on projects that are on organizing committees that have been reviewing applications, like all these kinds of and that's become fascinating. So I think there was a huge hunger for this kind of like almost like fabric across society to exist.

[00:18:35:22 - 00:18:56:07]
Amith
 That is that's incredibly powerful. It's also touching and it's interesting because you're at this section of thing that is often talked about as potentially maybe even competitive at the species level globally with us. But at the same time, it's an opportunity to amplify our humanity, which is exactly what I see you doing and what you described. Very, very cool.

[00:18:57:11 - 00:19:06:05]
Amith
 Associations, they oftentimes struggle with this idea. You know, our association folks who listen to this podcast and are involved with our organization,

[00:19:07:05 - 00:20:03:10]
Amith
 they are in many cases running organizations that are decades or even centuries old, people that are in various branches of engineering and science and healthcare and law and accounting and always other fields. And there's enthusiast organizations and trade groups and things like that, too. Many of them are struggling with the question of how to remain relevant, which we prefer to position the question is how do you really jump ahead and amplify their impact? But it's the same fundamental question is, you know, what are the products and services? And what I see in a lot of these organizations is they're doing similar things to what you're describing. They're connecting people. They're creating opportunities to build that fabric. But at the same time, a lot of times I think they find themselves on their heels. I think what I'm hearing from you, though, is the formula you're applying, yes, topics and the expertise you bring for our super contemporary, very relevant. But the way you're organizing people is very much a classical association playbook in a lot of ways and the way you do the way you're doing events. Would you agree with that? Or do you have a different point of view on that statement?

[00:20:04:18 - 00:21:33:01]
Speaker 7
 No, I think that's fair. You know, something that's kind of occurred to me, there have been a number of these initiatives I've been involved in over the years. And I think like kind of some common themes are like, I often describe what you want to do is create the space where you can bring the right people together and like kind of, I don't know, maybe put them in a box, shake it and wait for the magic to happen. And it's, you know, you get a lot of cool things coming out of this kind of critical mass. And I think really kind of a hallmark of the modern era is that like stuff is complicated, right? Our technology is complicated. Our politics are complicated. The way our society works is complicated. And, you know, we've for very obvious reasons, we've kind of built these models of like specialization where, you know, like people spend their lives getting good at things, but everyone gets good at different things. And like, it's kind of unreasonable to expect that the same person is good at all the things. But if you can kind of reduce the friction in idea flow between individuals, magic can happen. Whether that magic is about like, you know, crazily innovating on a technology or like finding new structures to organize around whatever theme you care about. And so like, to me, that's like the secret sauce is really like, can you find these different avenues to connect across people that otherwise wouldn't be together? I guess that's kind of really the purpose of many associations.

[00:21:34:16 - 00:22:23:23]
Mallory
 Well, at the top of this call, Benji and I were chatting about associations in general, and Benji kind of said, can you explain what you mean when you say associations? But yes, and me to your point, as you were talking about deep learning in DABBA, I thought, ah, okay, you're bringing the community together. You have in-person events, educational events, it very much to me sounds like an association. But I do want to point out one thing, which I really enjoyed the idea of in DABBA X. So you as a South African recognizing, okay, we can try to put on these events in other countries, but perhaps we don't know the best way to do that in another culture, in another country. And I'm curious, maybe this is more of a question for you, Amith, about the idea of associations, I don't know, maybe taking some similar inspiration from that, maybe having member organized events or kind of like satellite association events in different states. Amith, what are your thoughts on that?

[00:22:23:23 - 00:22:49:24]
Amith
 Sure. I mean, I've seen that. I think the description that you just provided is super applicable and interesting ways to collaborate. I mean, there are associations who are much more and they're focused on protecting their turf and all that stuff. A lot of that though comes from having been around decades in some cases where people settle in, especially multiple generations into leadership, volunteer leadership, as well as if these organizations are big enough to have staff leadership.

[00:22:51:02 - 00:23:43:11]
Amith
 They tend to just kind of repeat the same playbook. And after a while, there's a calcification that occurs and it's very hard for them to innovate. But I do think there's a lot to be inspired by here. The TEDx model I think is great. And I do think there's some branding challenges there that you might have run into. I don't know if you have where some deviates a little bit from your values, from your approach and to kind of pull them back in and find the right way to do things. But I think there's an incredible opportunity there. Many associations, I would say, tend to organize around the principle of a topic and then a fairly narrow community. So it'll be like the American Association of X, right? And whatever that topic is. And I think really the idea of country-specific associations are both totally irrelevant today and more relevant than ever, which I would explain as being in effect local and culturally aligned is super important, but we have to collaborate globally.

[00:23:44:21 - 00:24:46:05]
Amith
 And actually that's a question I really have had in the back of my mind the entire time, Benji, you were speaking, was Africa is placed in global innovation landscape. So a lot of times I think people in locations like the US or Western Europe, they think about Africa or India or places in what we'd call the developing world. Let's say generally, obviously South Africa is very different from that. But when you think about it from that very narrow point of view of the Western world, when we think about it that way, the mindset essentially, it's not said explicitly, but the mindset essentially is, "Okay, we need to bring everybody along. We need to make sure people don't fall behind." Which I think is noble on the one hand, but it also, I think misses the opportunity where there's an opportunity for leapfrog moments from places that think differently and are willing to explore these technologies perhaps from a different lens. I was just curious how you felt about that perspective. I think the whole American and Chinese race, so to speak, on frontier AI and models and all the craziness happening there leaves a lot of opportunity. So I'm just curious how you guys are thinking about that.

[00:24:46:05 - 00:24:58:13]
Speaker 7
 Cool. And I mean, again, being a huge place, there's quite a diversity in the way people think about things, what kind of resources they have, whether that's the size of teams,

[00:25:01:14 - 00:27:11:04]
Speaker 7
 university density, all these kinds of things for different places and different populations and so on. I think a bit of a common theme is kind of frugal innovation. We see this across a whole lot of different disciplines where it's like, "Okay, we don't have all the resources of somewhere else, but we've got certain problems we need to solve." And I think there's a lot of interesting things that come out of this. And we see this, there's all sorts of interesting things people have done ranging from this kind of this mobile money idea, which in Pezo, which came out of Kenya, that this was before anyone else was doing it, paying for things using your phone and was your phone account linked to your bank account. And even if you're making donations or tipping people, you could do this using and still do using these apps. And these kinds of things that kind of came up as financial solutions, not relying on stuff from other parts of the world. And there's many examples of this. We've got a big health insurance company here called Discovery that a lot of their philosophy is based on kind of shaping nudging behaviors to basically lower insurance costs. So if you can incentivize your people and then they've got these amazing incentive schemes to be exercising regularly or driving safely and things in that low is your premiums and gives you various benefits. And even to the extent of things like services that you can call to have potholes repaired because turns out that's cheaper than paying for car claims. So there's these different kinds of innovation that come up under different sort of situations. And I think that's interesting. But also it happens in kind of the social space because if you're trying to... We've got examples where maybe in the healthcare space,

[00:27:12:10 - 00:28:09:11]
Speaker 7
 you've got these rural clinics and you don't have specialists anywhere near there. And certain kinds of diagnostics, you would need a specialist to see that, but there's no appropriate specialist anywhere nearby. And there's these very long kind of logistic chains to get samples looked at. Well, turns out then people die in certain situations. And so you need those people go and talk to the appropriate tech people and kind of jerry rig some sort of solution. And then it works and it's like out there. And so I think in many cases in more developed situations, it would be just harder to make those conversations happen. Like the person with the demand would have to go up the whole food chain to get the person on the top to talk to their counterpart versus just going and finding the person who can solve the problem. And so I think all of this talks to kind of a necessity driven way of thinking about solutions to problems.

[00:28:10:13 - 00:28:58:14]
Amith
 So Benji, with regards to infrastructure and the whole idea of the way we think about it out here, I live in New Orleans and we're just having some internet problems. So over here in New Orleans, our infrastructure is not the best. And so you mentioned potholes and actually a friend of mine just started an AI company here in New Orleans that's aimed at helping city governments do a better job of providing government services, including specifically the use case of how do you find a pothole, take a picture of it and get it resolved quickly. He's got a large mountain to climb on that. But I suspect compared to where you live, New Orleans is far behind in terms of infrastructure and a lot of other things that we have. We have some advantages. There's very nice things in this city. There's lots of wonderful aspects, but we're a little bit behind in terms of a lot of these things, including our internet connections, which our regular listeners probably are aware of.

[00:29:00:13 - 00:30:27:20]
Amith
 So I wanted to ask you, there's this opportunity to think differently when you are in a position where constraints are perhaps higher, at least the classical constraints people think of. I want to remind our listeners that AlexNet was trained on two off-the-shelf GPUs that were purchased in an electronics store. I want to remind our listeners that Jensen Wong saved NVIDIA in their third iteration back in the 90s because he had no money left. He had six months to live basically. So constraints are the mother of the religion. I think the opportunities are there for people in all areas of the world with very limited budgets to think creatively and sometimes come up with solutions because they don't have 80,000 GPUs in a data center that they can do whatever they want with. And so for our associations, bringing it back to our listener group, many associations often have this mentality of we can't do what Amazon or Netflix or so-and-so is doing because we're a small association. I really don't think that's true anymore. I don't think it's ever really been true because constraints have always driven this type of innovation. But Benji, in the world of AI, when the speed of change is happening and the science is also, while it's been around for a while, much of what we're doing now is changing is so new every six to 12 months. Here's radical acceleration. Talk to us a little bit about that in terms of your perspective about being able to create fundamental innovation in Africa. More on the research side of what you do at the university or with your colleagues, what are you excited the most about in terms of opportunities for research on the fundamental side?

[00:30:28:23 - 00:30:41:16]
Speaker 7
 Cool. That's a great question. So if you're working at Google or OpenAI or something and you're not trying to innovate on the thing that requires the world's biggest data centers,

[00:30:43:02 - 00:30:45:10]
Speaker 7
 you're kind of throwing away your competitive advantage.

[00:30:46:10 - 00:30:54:19]
Speaker 7
 But that's your competitive advantage. It doesn't mean that's the playbook everyone else should be using. This is very much the philosophy that we take over here.

[00:30:56:10 - 00:31:04:18]
Speaker 7
 In the AI space, I think of it as this way. We know of two things that are capable of writing poetry.

[00:31:06:07 - 00:31:11:19]
Speaker 7
 The human brain, the blob of jelly between our ears and our large language models.

[00:31:13:20 - 00:31:41:00]
Speaker 7
 If you squint and half close your eyes, they're at some level of abstraction, similar kinds of ideas, but quite different technologies. They've got very different characteristics and pros and cons. One of them is about the energy use. This idea of your data center versus your brain using the energy of a light bulb, famously what everyone says.

[00:31:42:07 - 00:31:59:19]
Speaker 7
 What this means is we know of at least two ways to solve the same problem. Now, that doesn't mean by any means that there's only two ways to solve that problem. In fact, that'd be interesting if it was the case. But most of us have reason to believe there's other ways to build these kinds of technologies.

[00:32:01:06 - 00:32:16:14]
Speaker 7
 If you're in that space where you've got all those resources, you should be thinking about how do we take what we're building and make it better. If you're in a different space, you should be saying, "Are there other ways I can get to that space of, in this case, things that can write poetry?"

[00:32:18:00 - 00:32:55:03]
Speaker 7
 At my university, what we've done recently, in fact, just under a year ago, we launched this kind of AI institute, which I direct, called the Mind Institute, the Machine Intelligence and Neural Discovery Institute. This came out of a couple different things. One, with all the work we've been doing in Africa, there's really not much that's around innovating in AI itself. Everyone these days is doing AI. But for most people, that means taking off-the-shelf models and applying your own datasets.

[00:32:56:15 - 00:33:05:01]
Speaker 7
 Now, that's great, and that's what most people should be doing. But I think every region should have people that are looking at the innovation, and for multiple reasons,

[00:33:06:01 - 00:33:16:15]
Speaker 7
 from making sure that solutions are appropriate, making sure we can, again, apply our unique advantages and angles to this technology, which could have global benefits.

[00:33:18:00 - 00:33:33:09]
Speaker 7
 The way we thought about this, as we took what we'd been building in our School of Computer Science and Applied Maths, as our really core fundamental AI research groups, and wanted to expand this. The Mind Institute was really about, if we go back in the history of AI,

[00:33:35:08 - 00:34:32:17]
Speaker 7
 there were always these two angles to it. One was the engineering-slash-computer-science angle of, "I want to build the cool tech to solve the hard problems." But then there was more the biological angle of, "I want to understand how brains work." These always went hand in hand, and many people have forgotten that now. But if we're in the quest to build a community, get critical mass, that sort of thing, and famously, we're under resourced in things in Africa, can we find different groups of people that ask questions and work on things that have some sort of level of commonality? That's what this institute is about. We've pulled together a whole lot of academics that, about a third of them focus on fundamental AI, but then we've got these world-leading groups in areas ranging from evolutionary sciences to neuroanatomy to neuropsychology to policy and ethics and governance, and even in the creative arts.

[00:34:34:10 - 00:34:53:07]
Speaker 7
 We've pulled together all these people that have some interest in intelligence, whether it's in machines or animals or humans, and build this as a community that's meeting regularly. We're incentivizing crazy novel ideas. We're trying to give seed funding to people that can come up with something that's different.

[00:34:54:09 - 00:35:08:07]
Speaker 7
 And again, building this kind of community that's a different kind of community. There's a common theme, but it's towards trying to innovate, again, based on the various different strengths that we have to draw different inspirations that could affect the global AI space.

[00:35:08:07 - 00:37:15:09]
Amith
 That makes a lot of sense. I think the pendulum swings in one direction or the other often times with many fields. There's been this classical pendulum swing from all-in-one systems to point solutions and back and forth and back and forth. And the thesis was, put it all in one place and you'll have unification of data. And then, of course, that blows up because you can't really make that work. And then people say, "Oh, let's use specialist systems that are best of read." And of course, that becomes an integration nightmare. They go back and forth and back and forth. And from neural nets to symbolic systems and back and forth, they're just different kinds of core thesis. That's happened and people forget that. And they're like, "Oh, wait a second. We have symbolic reasoning. We don't know what that means anymore because everything has to be in the LLM." And then the inverse is also true. People say, "Oh, neural nets will never work." That was not very long ago when people were saying exactly that. And so now it's flipped. I think there's opportunities to have a fresh set of thoughts the way you're describing. I also really deeply believe that those constraints will lead to different thinking because of necessity. And the acceleration rate of capabilities is really quite stunning what's happening. People were really surprised in January when DeepSeek released their R1 reasoning model and it was like O1 caliber performance. And it was supposedly trained on single-digit millions of lower-end GPU time, not frontier, high-end GPUs. And there was this shocking thing and everybody was looking at it saying, "Well, of course that happened." And the fact that it came from China or anywhere else for that matter, the concepts that went into it, yes, there's certainly some great innovation that went into their approach. But fundamentally, it was almost guaranteed to happen at some point this year. And of course, that's happened again. There's now Moonshot AI out of China as well, released their Kimi K2 thinking model, which is blowing away a lot of the benchmarks and all this stuff is happening. So what's your general thought in terms of how people can level set in their own minds how to think about that when the world is changing so incredibly fast? If you were talking to a nonprofit leader who organized conferences and was thinking about the business that you're in with your community program, how would you help them think about how to plan ahead when things are changing at such an exponential pace?

[00:37:16:10 - 00:37:33:03]
Speaker 7
 Cool. I have this mental model I use for the global research landscape. And I think of research as one of those things where diversity and thinking is important. So you imagine people from the old prospect of days looking for gold.

[00:37:34:19 - 00:37:44:04]
Speaker 7
 You just imagine this broad field and everyone's digging in different places. And if someone finds something, then a bunch of people swarm towards that and dig in that area.

[00:37:45:08 - 00:38:22:08]
Speaker 7
 But maybe that's not the best place to be looking. Maybe there's just a short-term amount you can get there. If the entire population of those miners went and only dug in that one place, they're probably going to end up competing and they're probably going to miss out on what might actually be the best place to dig for gold. So what you want is that some people will follow that trend and others will say, "No, I'm convinced that I'm in a better place." And so this is the picture that I use for the value of this. And altogether, there's this net win that we're, in that case, getting gold, but in general, getting knowledge and making progress.

[00:38:23:09 - 00:38:54:11]
Speaker 7
 So because of the amount that's happening here, AI is the fastest growing field in human history. We look at hundreds of papers being published a day, hundreds of billions of dollars, tens of thousands of startups. So to me, that means a few things. One, it means that whatever you're doubling down on is probably not going to be the right thing at the end of the day. Conversely, if you spend every day going for what's today's best thing, you're going to spend more time context switching than getting the most out of whatever tools you're using.

[00:38:55:15 - 00:39:26:02]
Speaker 7
 And so you kind of go to balance these things. And in my technical area of reinforcement learning, we talk about this as the exploration exploitation trade-off. You've got to commit to the thing you're committing to because that's how you exploit. You get the most out of it, but you need to do a little bit of exploration and find out, is there something out there that could be better suited to my needs? And handling that trade-off is key. I think what's important here is to focus on the commonalities.

[00:39:27:02 - 00:39:34:21]
Speaker 7
 So we often talk about the key principles here is being curious, being good at what you're doing.

[00:39:36:08 - 00:40:41:01]
Speaker 7
 It's this sort of thing. You can't be switching every day. You can't be spending your designing policies around AI systems where your policies are already out of date by the time they come up. You've got to have this level of flexibility and realism. You don't need to chase every single hype and trend as it comes up. You need to say, "What is the right thing for me?" So one of the other ways I think about this, I know there's a lot of controversy about the analogies you can use with AI, but I really think a lot of thinking in corporate type spaces, you can simplify it to being like, think of AI as some new employee. And the kind of employee I describe is straight out of university with a couple PhDs, but completely naive and green behind the ears and no idea really how the world works. So you've now got this employee. Now, maybe in the background, they're learning some new skills and stuff, but how do you best use this employee?

[00:40:42:07 - 00:41:16:21]
Speaker 7
 Maybe they've got some weird personality things you need to get better at just communicating with them in general. Maybe you need to think about how do they work in the pipeline of your organization and that sort of thing. Maybe if they were coming from a temp agency and it's a different employee tomorrow, you can't just come up with regulations around how do we work with Bob and then tomorrow how do we work with Jim. And also, you probably don't want to be swapping them out all the time, even if they tell you, "Oh, there's now one with three PhDs." I think if you use this kind of mental model, it's kind of useful to say,

[00:41:17:23 - 00:41:33:02]
Speaker 7
 rather than like, "Oh, we've got the next newest version of Excel," or whatever. If you think about it in that kind of way, I think that's quite useful for your resilience to the change because you're thinking about it at a slightly macro level.

[00:41:33:02 - 00:42:18:20]
Amith
 Yeah, it's super helpful. I think the point that I'd hit on for our audience is when you're thinking about AI road mapping, which really relates to what Benji's described, maybe think in 12-month intervals or six-month intervals, do really deep focused work without the churn, but pop your head up every once in a while and see what's around you so that if your core thesis has been shown to be not on path, then you can change it around and you should expect to change it around. Because I think what you said earlier, you said that if you double down on whatever you're doubling down on, it's probably wrong. And that's true for 99% of everything that happens. And even if it's not true, and if you're right, it'll be wrong eventually. So it's really important to keep taking that look. But it's

[00:42:18:20 - 00:42:23:13]
Speaker 7
 definitely wrong to be changing every day, right? Because you'll never get the most out of whatever you're using.

[00:42:23:13 - 00:43:19:16]
Amith
 Yeah, you'll spend all your time doing churn. I mean, in our software development shops, we're building AI agents and all these things. We're constantly test harnessing all the latest models and seeing how they behave against internal benchmarks, et cetera, et cetera, et cetera. And that's part of our job is it's not to put that into production, but to see how does KIMI K2 reasoning work compared to blah, blah, blah model. But really, a lot of it is in production, you need to stick with what works. And there's so much upside that I think another thing to remind our audience of is that even if AI stopped all progress at the fundamental scientific level today, which I certainly hope it doesn't, and the work you and your colleagues are doing are ensuring it continues along. But even if on the business community side, nothing ever got better, and the AI we had today was literally the end of the road, it would take decades for society to fully appreciate and diffuse the full power of what we have in our hands now. And so that also underscores the point you made, that there's so much value you can extract from what's available at this moment in time.

[00:43:19:16 - 00:43:21:00]
Speaker 7
 Absolutely agree.

[00:43:21:00 - 00:43:39:13]
Mallory
 In my research on you, Benji, I found that you mentioned this idea of algorithmic sovereignty. So the idea that data sovereignty isn't enough, you also need to control the refineries that are turning the data into intelligence. I was hoping you could speak a little bit about that and how you feel it pertains to AI in Africa as a whole.

[00:43:39:13 - 00:44:44:02]
Speaker 7
 Cool. Yeah. So again, kind of going back to a bit of what I was saying earlier, I think this comes down to saying that you don't need everybody in the world to have all the skills. And there's certainly value in using tools developed all over the place, but you need to have, or at least be close enough to people that are developing these things. So I think there's a number of potential risks if, it would take like a situation in Africa, but I think this is kind of a global problem. So let's say you're in some country that you're building your tech stack and it's dependent entirely on stuff from somewhere else in the world. Now, if you're building the tech stack, you probably got some idea of what's going on there and maybe you've got an idea of how to swap out pieces and so on. But let's say there's some sort of economic problem and your currency suddenly becomes worth a fraction of what it was before. Now, if you're paying some subscription model, it might not be affordable to keep your products going.

[00:44:45:11 - 00:44:53:14]
Speaker 7
 Similarly, a new kind of tool comes out. Do you have the right way of saying, is this relevant to your community?

[00:44:54:16 - 00:45:50:05]
Speaker 7
 Are there challenges that it might impose? Are there ways that it might cause some sort of issues that potentially could have been foreseen if you had the right level of tech expertise? Again, that's not to say that every company should have their tech expertise. In most cases, that doesn't make sense. But I think at least different regions of the world and similarly, different kind of sectors in different parts of the world need to be tapped in some way. Whether that's some group of universities that have programs like doing AI in mining or whatever it might be. There's got to be some sort of group that you can talk to that understand these things at a more fundamental level. Now, I think, again, when we go to the African angle, there's a number of other reasons you want this as well.

[00:45:51:10 - 00:46:05:04]
Speaker 7
 Again, if the best tool to solve a problem is American, you should use that American tool. Same way if it's German or Japanese or whatever. But I strongly feel like you need to have the aspirations to be playing at that level.

[00:46:06:13 - 00:46:32:16]
Speaker 7
 If you don't have those aspirations, no one will treat you as having those aspirations. If you want to have a seat at the table and at the big table, you need to be at least trying to play that game. If you're not, you're missing out on potential economic opportunities. You're missing out on different applications that you might have been able to tap into that you wouldn't have if you weren't thinking about the right part of the innovation pipeline. All of this goes back to,

[00:46:33:20 - 00:46:39:21]
Speaker 7
 particularly as a sector, as a region, as a country, you can't just focus on one piece of this innovation pipeline.

[00:46:41:01 - 00:47:00:11]
Speaker 7
 As I often say, anybody would fund the fancy new health and fitness app that's running on your watch or something. But that only exists because some other people went and put satellites into space, GPS satellites. That only works because someone else went and invented relativity.

[00:47:01:11 - 00:47:35:21]
Speaker 7
 The successes we have as a society come from these being pipelines that work from the crazy person working in isolation on a blackboard all the way through to the cutting-edge tech startup that's doing all the raises and building their stacks and everything like that. If you're only investing in... You want to invest in different amounts in different parts of this pipeline, but if you're only investing in the later parts, you're missing opportunities and you're becoming dependent on things you probably don't necessarily want to be dependent on.

[00:47:37:04 - 00:47:51:18]
Amith
 I think that makes a lot of sense. It also leads people to question fewer assumptions in terms of how things are done. With that knowledge and that education at every level of that vertically integrated stack, you have the opportunity to think more creatively, I think.

[00:47:51:18 - 00:47:55:06]
Speaker 7
 Plus competition makes everything better for everyone, right?

[00:47:55:06 - 00:48:32:04]
Mallory
 Absolutely. I also wanted to share, as someone based in the US, I know a lot of our audience probably relates to this as well. We feel like ChatGPT can do anything, right? It's got multiple PhDs. ChatGPT can take on nearly any challenge we throw at it. But in my research on you, Benji, I realized that when it comes to African languages, ChatGPT does not do so well on those. I think the idea of innovation spread globally and making sure we tap into that nuance and understanding what these models do well and what they don't and how we can, to your point, put bright people in a box, shake it up, and then see if we can come out with something better.

[00:48:32:04 - 00:48:58:15]
Speaker 7
 Right. The language of things is important because that's like people's gateway into anything from being part of an innovation ecosystem to being part of a market that could be accessible by someone somewhere else in the world. If you could kind of ignore this language component, you've suddenly got a much bigger market, a much bigger set of potential collaborators, and so on. It's really critical that everybody has access to this kind of infrastructure.

[00:49:00:01 - 00:49:16:04]
Mallory
 Absolutely. Well, Benji, that takes us to almost the end of the hour. I want to say thank you so much for joining us on the podcast. In terms of the general huge field of artificial intelligence, what is something on the horizon that you are most looking forward to?

[00:49:16:04 - 00:49:57:03]
Speaker 7
 So many things, but what I'm really excited about in my own work is autonomous decision-making, which is a thing that we see a little bit with the kind of agentic AI that's very popular now, but it's really not doing the way many of us in the field think about it, which is doing sequential reasoning and almost to an extent simulating things, learning from counterfactuals, and so on. There's communities of people, areas like reinforcement learning, that are thinking about this a lot. We're starting to see ways that people are integrating this with the amazing power and things like LLMs and VLMs and so on. I'm interested to see what happens when we start bringing this all together.

[00:49:58:04 - 00:50:01:23]
Mallory
 Where can people keep up with you if they want to see the work that you're doing?

[00:50:01:23 - 00:50:29:07]
Speaker 7
 I mean, following on the socials, I think at the moment we're trying to get good at our mind institute, machine intelligence and neural discovery institute. I think probably best is LinkedIn at this point. The deep learning at DARBA is active on many socials. My company doing African natural language processing, which is called Lilapa AI is active on many socials. Somewhere among all of these, you'll find the work that we're doing.

[00:50:29:07 - 00:50:31:23]
Mallory
 We will certainly link to those

[00:50:31:23 - 00:50:37:09]
 (Music Playing)

[00:50:48:02 - 00:51:05:01]
Mallory
 Thanks for tuning into the Sidecar Sync podcast. If you want to dive deeper into anything mentioned in this episode, please check out the links in our show notes. And if you're looking for more in-depth AI education for you, your entire team, or your members, head to sidecar.ai.

[00:51:05:01 - 00:51:08:07]
 (Music Playing)