The Great AI Conversation

Beyond the Algorithm

A Collaborative Piece - Part 2 of 3

Some moments don’t announce themselves. They just arrive and you recognise them.

Part 1 of The Great AI Conversation went out last week Wednesday. What came back was recognition. Contributors who said it felt meaningful to be part of something like this. Readers from Brazil connecting it to Latin America and other parts of the Global South. People sitting with thoughts they had been thinking of about Africa and others with questions they hadn’t thought to ask themselves before. A conversation that had been happening in separate rooms between separate people suddenly in the same space.

It felt like a moment that had been waiting to happen. Long overdue.

Which is what makes Part 2 feel necessary rather than just sequential. Because Part 1 was about first impressions, what surfaces before you’ve been asked to think carefully. Part 2 is harder. It asks for a position.

The options in the question are not neutral. Each one carries a different power dynamic. Builder implies agency. User implies consumption. Testing ground implies someone else’s agenda. Regulator implies institutional authority. I offered them deliberately not to box anyone in, but to see which box people reached for first, and what that reaching revealed.

What came back was more contested than Part 1. Some contributors pushed back on the framing entirely. Some named things the question didn’t offer as options. And one went somewhere none of the others were willing to go.

Here is what they said and what I think it means.

---

Question 2: Do you imagine Africa primarily as a builder, a user, a testing ground, a regulator, or something else entirely in the global AI ecosystem?

Category 1: The User Frame

The contributors in this category lean toward user as their primary frame but with different degrees of openness about what that means and where it leads. One sees user as a starting point that will evolve. One stays closest to user with genuine honesty about not having seen enough to say otherwise. One sees builder and testing ground but holds that view lightly, aware of the limits of his own picture.

T.D Inoue

“Primarily as a user. I haven’t seen much about Africa as builders beyond a few inspirational TED Talks, so I don’t have much of a basis on which to comment. But as users, I can imagine a wide range of uses. On the other hand, I’m not sure what uses go beyond that of a glorified search engine. “How do I fix this?” “How can I make that?” So definitely primarily as a user.”

Brad Leclerc

“What little I’ve seen specifically leans me to the “builder” and “testing ground” ideas. Africa is jam packed with people coming up with creative solutions to issues (out of necessity or otherwise), and given the steep imbalance of resources for AI related compared to other countries, they’re all edge cases and very little large system-level projects (as far I’m aware at least), which makes some things a lot harder, but it actually can make innovation in unexpected directions more common, which is something the AI world is in DESPERATE need of. So, I wouldn’t be at all surprised to start seeing ideas that start in Africa to solve some niche issue or a use-case that feels very specific, to end up being the key to larger changes in unexpected ways at some point, just from the sheer number of “little” problems people are solving in creative ways.”

Kevin Guiney

“I would initially see Africa primarily as a user of AI, focused on identifying and deploying value-added applications. However, that phase is unlikely to remain static. As these tools are adopted and adapted, I expect locally driven innovation to emerge, leading to solutions that are not only relevant within Africa but marketable globally.

In many regions where computing resources and telecom infrastructure are still developing, constraints often drive creativity. This creates the conditions for meaningful innovation in sectors such as agriculture and mining, where Africa could develop world-leading applications. Over time, as economies strengthen and infrastructure matures, some nations may evolve into leaders not only in AI, but also in robotics and even quantum computing. While there are important differences across countries, the broader challenge in many nations today is managing debt, growing GDP, and continuing to build out infrastructure. At the same time, it is important not to overlook the foundation already in place, particularly the talent, adaptability, and problem-solving capacity that will ultimately drive that progression.”

Editorial commentary: The user frame is the most instinctive frame. When knowledge runs out, when the TED Talks are the primary source, when the infrastructure gaps are more visible than the innovation happening inside them, user is where the mind goes. That is not a criticism of these three contributors. It is a description of a visibility problem that shapes the entire global AI conversation, and that this group, more honestly than most, makes legible.

T.D. is the entry point. He hasn’t seen much about Africa as builders beyond a few inspirational TED Talks. He says so directly, without dressing it up. That honesty is worth more than a confident answer built on thin ground, because what he is naming is not a gap in Africa’s AI activity. It is a gap in what gets covered, what gets amplified, what reaches someone sitting outside the continent trying to form a view. And his uncertainty about what African AI use looks like beyond a glorified search engine “how do I fix this, how do I make that” has a concrete answer he may not have seen. In 2025, Kenya was the world’s number one user of ChatGPT, with 42.1% of internet users aged 16 and above using the platform monthly, ahead of the UAE, Israel, the United States, Japan, and China. That is not a glorified search engine statistic. That is the highest adoption rate on earth, driven by students, freelancers, entrepreneurs, and professionals using AI for education, coding, content creation, and productivity across semi-urban and rural regions alike.

But the Kenya data also complicates the user frame rather than simply vindicating it. Because leading the world in usage of a tool you didn’t build, don’t govern, and don’t profit from is a particular kind of leadership. Critics have pointed to what some describe as a data-as-currency dynamic, Kenyan users feeding data into global platforms while the resulting insights, algorithms, and commercial value accumulate elsewhere. In agriculture, for instance, farmers provide planting and financial data to apps whose outputs are controlled not by the farmers but by the banks, insurers, and tech firms behind the platform. High use, in other words, does not automatically mean high benefit. T.D.’s instinct that Africa is primarily a user may be more accurate than he realised, not because Africa lacks the capacity to build, but because the structures currently in place make use the most accessible entry point, and leave the terms of that use largely in someone else’s hands.

Brad reaches for something more generative. He sees the resource imbalance clearly, the steep gap between African AI practitioners and the compute and infrastructure available to them and he reads it not only as constraint but as the condition for innovation in unexpected directions. Edge cases, niche problems, creative solutions born of necessity. His instinct is that something built to solve a very specific local problem could turn out to be the key to larger changes in ways nobody predicted. That is a genuine and important argument. It is also, notably, an argument made from the outside looking in, an observation about what constraint produces rather than an account of what is actually being built. The distinction matters not because outside observations are wrong, but because they tend to see the condition more clearly than the content. The creativity is named. The specific work is harder to see from that distance.

Kevin’s framing is the most structured of the three, and the most revealing in what it assumes. He sees Africa primarily as a user first, then as an innovator over time, as economies strengthen and infrastructure matures. That is a developmental narrative, a progression from consumer to creator, from recipient to producer and it has a logic that feels reasonable on the surface. But it sits in direct tension with what category 2 of contributors described. One says Africa will get there. The other says Africa is already somewhere the conversation hasn’t caught up to yet.

What connects these three contributors is not a shared view of Africa. It is a shared experience of limited information, handled with varying degrees of awareness. T.D. names the limit explicitly. Brad works creatively within it. Kevin builds a confident structure on top of it. Together they reveal that the global AI conversation’s image of Africa is not neutral. It is constructed from what gets funded, covered, translated, and amplified. And when that image is thin, user is the role that fills the gap. Not because it is accurate but because it is available.

On visibility: If the primary image of African AI innovation available to a global audience is the inspirational TED Talk, what is not being shown and who is making that editorial decision?

On progression: What does it mean to frame Africa as a future builder rather than a present one and what does that framing protect us from having to know right now?

On constraint: Is there a difference between observing that constraint drives creativity and understanding what that creativity is actually producing and does that difference matter?

---

Category 2: The Builder Frame

The contributors in this category see Africa primarily as a builder, but each with a different logic through: 1)industriousness and making things work, 2)Ubuntu and community, 3)building from the ground up in African languages.

Daria Markava

“A builder, on its own terms. African AI isn’t simply taking a model developed in the global North and adapting it for African languages. Instead, it’s built from the ground up, with the characteristics and nuances of African languages in mind from the very start. This innovation wouldn’t be measured by whether the models’ performance hits a standardized benchmark, but by whether it can accurately capture and convey the intricacies of Africa’s languages - especially the low-resource or oral languages.

A trend I notice with African AI is that it’s rarely built for the sake of it. The majority of startups and businesses aim to provide local solutions to real problems, rather than taking the “let’s build it first and then see where it takes us” approach of the global North. Adopting Africa’s current solution-based approach worldwide could finally help us bridge the gap between the promises and the reality of AI.”

Marcella Distefano

“I definitely see Africa as a builder, but with a different logic. In the West, we’re used to a very individualistic AI, focused on personal productivity or consumption. In contrast, what’s brewing in Africa is much more aligned with the Ubuntu philosophy: ‘I am because we are.’ They are building an ecosystem where AI is designed for the community. It’s not just tech for tech’s sake; it’s a search for collective solutions. That vision of ‘community-driven tech’ is a breath of fresh air that the global AI ecosystem desperately needs to break out of the self-centered bubble it’s currently stuck in.”

Dinah Davis

“I see Africa as builders and testers. There’s an industriousness there — a culture of making things work, of finding a way. That energy is going to translate powerfully into this moment. I wouldn’t put them at the forefront of regulation, I think the EU will probably lead that, the same way they’ve led on user privacy.”

Editorial commentary: These three responses agree on one thing: Africa is a builder. But they don’t agree on what that means, and the disagreement is where the more interesting argument lives.

Daria’s case is technical. African AI, as she describes it, is not adaptation, it is origination. Models built from the ground up, with the characteristics of African languages in mind from the start. Not translated. Not localised. Conceived differently. And measured differently not against a standardised benchmark, but against whether the model can accurately capture and convey the intricacies of languages that are low-resource, or oral, or carry meaning in ways that resist the assumptions baked into most existing systems. This is a precise and demanding argument. It is about the architecture of the thing, about what gets embedded at the level of the model itself when you start from a different place.

Marcella is making a different kind of argument. Hers is not primarily about the technology. It is about the philosophy the technology carries. Ubuntu “I am because we are” is not a feature you can add to a model after the fact. It is a way of understanding what a person is, what a community is, what technology is for. An AI built inside that philosophy looks structurally different from one built around individual productivity or personal consumption. Where Daria’s measure is linguistic accuracy, Marcella’s measure is something closer to collective benefit. These are compatible ideas. But they are not the same idea. You could build a technically sophisticated language model, one that gets the tone shifts in Swahili exactly right and still have it serve individualistic ends. You could build something deeply communal in its logic and get the linguistics wrong. The two framings need each other, but neither one contains the other.

Dinah’s framing is the most grounded of the three and in some ways the hardest to argue with, because it doesn’t reach for philosophy or linguistics. It reaches for character. A culture of industriousness, of making things work, of finding a way. That is a different kind of claim, less about what African AI is building toward and more about the disposition from which it builds. It is also the framing most vulnerable to being romanticised from the outside, which is worth naming carefully. “Making things work” can be a genuine description of resourcefulness as a design principle. It can also, depending on who is saying it and from where, slide into a way of admiring constraint rather than questioning it.

But Dinah also says something that sits in quiet tension with both Daria and Marcella, without quite resolving it. She wouldn’t put Africa at the forefront of regulation. She’d give that to the EU. The comment reads as observational — she’s describing where she thinks governance will come from, not necessarily where she thinks it should. But the implication is significant. Because if Daria is right that African AI is built from different foundations, and Marcella is right that it carries different values, then the question of who governs what gets built is not a secondary question. It is the question. You can originate from a different place. You can embed different values. But if the regulatory framework that determines what is permissible, what is safe, what counts as compliant, was written somewhere else for someone else, then the autonomy these three contributors are describing has a ceiling. And none of them fully reckon with where that ceiling is.

That tension doesn’t undercut what they’re saying. It is the condition under which they are saying it. Three distinct arguments (technical, philosophical, cultural) that together point toward something the global AI conversation has not yet fully absorbed: that building on different terms is not the same as building toward a different destination. The destination is different too. The questions being asked are different. The problems being solved are not edge cases waiting to be absorbed into a larger system. They are central human challenges that the mainstream AI world has not yet had the urgency, or the context, or frankly the need, to confront.

The fact that these three contributors name what Africa is doing as building, not despite the constraints but through them is itself a kind of argument. Because naming pushes back against the assumption that real building only happens in certain places, with certain resources, toward certain predetermined ends.

On architecture: Most language models are built on text. What happens when the language you're building for was never primarily written?

On values: Ubuntu is a philosophy, not a feature. What would it actually take to build it into a system rather than describe it in a pitch deck?

On governance: If the values are African, the problems are African, and the builders are African, but the rules are written in Brussels, where does that leave the argument that Africa is building on its own terms?

---

Category 3: The Extraction Frame

One contributor went somewhere else entirely. His answer was extraction point and he didn’t say it abstractly. He named names, named places, named costs.

Tumithak of the Corridors

“Something else. Right now, the clearest role I see is extraction point. Kenyan workers were contracted by OpenAI to moderate content. They spent months filtering violent and sexual material so the models could ship clean. One of them, Mophat Okinyi, described severe psychological harm afterward. His mental health deteriorated. His marriage fell apart. Low-wage workers absorbed the worst of the process, then capable models replaced them.

Meanwhile, AI hardware depends on cobalt, copper, and nickel. In the DRC, artisanal miners dig without protective gear. Children work alongside adults. Communities live with poisoned water and soil. By the time the chip reaches a data center, the mine has become an abstraction.

Could some African countries become builders? Sure. But building an AI sector takes reliable power, connectivity, institutional stability, and above all capital. Even outside the major tech hubs, it’s hard to compete with companies that can spend billions on compliance alone. Scale that problem to countries dealing with food insecurity and the math gets brutal.”

Editorial commentary: Every other contributor in this collection offered a frame. Tumithak named people, named places, named costs.

Tumithak does not say Africa cannot build. He says the math is brutal. And that detail about competing with companies that spend billions on compliance alone reframes the entire conversation from the previous category. The question was never about talent or creativity. It is about what kind of building is possible inside these structural conditions and who bears the cost of the industry that makes the building necessary in the first place.

On labour: Content moderation makes AI products usable. Who does that work, under what conditions, and what happens to them when the models no longer need them?

On supply chains: If the device you use to access AI depends on materials extracted under conditions you would not accept in your own country, what does that make the industry — and what does it make you?

On the math: When a single company’s compliance budget exceeds what many nations can invest in their entire technology sector, is competition even the right frame?

---

Category 4: The Multiple Roles Frame

The contributors in this category resist the single label entirely. For them Africa is too large, too diverse, and too dynamic to be reduced to one role. They see builder, user, tester, and regulator not as options to choose between but as simultaneous realities, each playing out differently across 54 countries and as many contexts.

Cristina Patrick

“The “testing ground” framing has a history, unfortunately, the pattern of using African populations as infrastructure for someone else’s innovation is real and documented. But Africa right now is a builder, with serious barriers at times, yes, but with real momentum. And Africa is a huge continent with so many different cultures, languages, and communities, so I think it is hard to label it. But I hope to see more media about African researchers, regulators, developers, and builders, I know they are out there.”

Arylee McSweaney

“It wouldn’t be wise to box Africa into a single role. Trying to do so would limit both the continent and the global companies eager to engage with it. The data clearly shows Africa is already positioned to play multiple roles simultaneously in the global AI ecosystem. On the builder side, the continent has a fast-growing technical workforce. Africa accounts for roughly 3% of the global AI talent pool, with countries like South Africa, Tunisia, Egypt, Kenya, and Mauritius ranking highest in AI talent readiness and digital skills.
talentindex.ai +1

At the same time, only about 5% of African AI practitioners have access to the compute power they need for advanced work. This combination of strong human capital and infrastructure constraints is fostering a distinctive culture of frugal innovation, robotics experimentation, and climate-focused AI solutions tailored to real-world needs.
undp.org

Africa also serves as a natural testbed for frontier technologies. Robotics and AI are being deployed in agriculture, logistics, and healthcare — sectors with urgent local demands where global companies can learn valuable lessons from operating under real-world constraints such as limited infrastructure and diverse environments. AI is projected to add up to $1.5 trillion to Africa’s economy by 2030, particularly through applications in agriculture, climate adaptation, and healthcare, where African innovators are already creating context-specific solutions.
news.sap.com

Even in the USA, we’re starting to see participation in the innovation economy. Increasingly, African Ivy League graduates and diaspora founders are securing U.S. venture funding, positioning the continent as a bridge between global capital and practical, locally grounded problem-solving. Finally, Africa is emerging as a regulatory voice. With at least 16 countries having launched national AI strategies and the African Union advancing continental frameworks (including the Africa Declaration on Artificial Intelligence adopted in Kigali in 2025) the continent is asserting itself as an active policymaker rather than a passive recipient of foreign technologies.
intelpoint.co +1

In short, Africa is capable of being all of them at once. This multidimensional role is precisely what makes the continent strategically vital in the global AI era: a place where innovation is born from necessity, tested in diverse realities, scaled with local insight, and governed with growing sovereignty.”

Sources
talentindex.ai. Talent pool (~3%): African Union statements and related reports. au.int
Compute access (5%): UNDP analysis. undp.org
Economic impact ($1.5 trillion): SAP Africa report. news.sap.com
National strategies (at least 16): Intelpoint insights (as of mid-2025). intelpoint.co
Continental framework: Africa Declaration on AI (Kigali, 2025).c4ir.rw

Jax NiCarthaigh

“I see Africa as all of those, though unevenly. I mean 54 countries, it’s going to be uneven. It is far too large and diverse to be reduced to a single role. At the same time, I think many people in the global North still imagine Africa too narrowly, often in relation to extraction, labour, and exploitation at the early end of the supply chain. That can’t be the whole of it. I imagine Africa is also building, using, shaping, and questioning AI in ways the rest of the world should take more seriously.”

Dr. Sam Illingworth

“All of those simultaneously, which is what makes the question interesting. African researchers and developers are already building, often with fewer resources and more creativity than their counterparts in the Global North. From my own position in the UK, I am also aware that the global AI ecosystem tends to assign roles along familiar lines: certain countries build, certain countries regulate, certain countries get built upon. I think Africa’s place in AI will be shaped less by capability, which already exists, and more by whether the structures of the global ecosystem make room for it.”

Mike D

“A builder, user, and regulator, for sure. One idea that automatically strikes me is how AI will enable African developers to build hyper-local solutions for regional issues, like agricultural yields and financial tech. Africa is also in an excellent position to benefit big time from artificial intelligence as end-users. Why? Because traditional higher education has long been gated behind expensive institutions that are out of reach for most people worldwide. There are countless millions in Africa, and elsewhere, who cannot afford to attend the fancy Ivy League schools in the regions that surround my hometown in Greater Boston, Massachusetts. I’m not bragging, by the way, because I could never afford those fancy universities either, and I live right in their backyard. This topic all leads me to my single greatest hope for AI. Artificial intelligence can fix a lack of access to education. AI can potentially distribute education like nothing else I’ve ever seen before in my 10 years of online teaching. Imagine an elite AI agent that can speak to you in a simulated video call and literally teach you anything. At that point, we can finally democratize world-class education to all regions of the globe. Africa isn’t the lone beneficiary, either. I believe AI-backed education will also benefit countless millions of Americans and citizens worldwide who cannot otherwise afford to attain higher education. Respondents of the Anthropic study I referenced earlier also cited AI-backed education as an exciting opportunity to escape poverty, which I think is the biggest and most promising benefit of AI.”

Editorial commentary: Where the first 2 categories of contributors answered the question with a specific role, this group resists it. Not out of evasion but out of a considered argument that the question itself is too small. Africa is too large, too internally diverse, too dynamic across 54 countries and as many contexts to be reduced to a single role. That resistance is worth taking seriously, because it is not just a refusal to choose. It is a claim about what kind of thinking the global AI conversation needs more of.

Cristina is the sharpest entry point into this group, because she is the only one who historicises before she answers. She names the testing ground framing directly, the pattern of using African populations as infrastructure for someone else’s innovation and she names it as documented, not speculative. That is a different move from the others. It is not throat-clearing. It is a reminder that the roles being offered in this question are not neutral options. Some of them have histories and choosing not to accept one of them is also a choice with a history behind it.

From there, Arylee makes the empirical case for multiplicity. Africa accounts for roughly 3% of global AI talent. At the same time, only about 5% of African AI practitioners have access to the compute power they need for advanced work. Those two figures sit right next to each other in her response, and the gap between them is the most important thing she says. It is the structural reality that the “multiple roles” framing has to reckon with. Human capital exists. Infrastructure does not always follow. What grows in that gap, frugal innovation, context-specific solutions, creativity under constraint is real and generative. But it is also a gap, and naming the generativity without naming the gap would be incomplete. Jax names what created it. Extraction, labour, exploitation at the early end of the supply chain. That is not ancient history. It is the economic context inside which African AI is being built right now.

Dr. Illingworth says something that none of the others quite say, and it is the pivot on which this entire category turns. Africa’s place in the global AI ecosystem will be shaped less by capability which already exists and more by whether the structures of the global ecosystem make room for it. That is a structural argument, not a cultural one. And it is quietly the most unsettling thing said in this group, because it shifts the burden. The question is no longer what Africa can do. It is whether the conditions exist for what Africa can do to matter at the scale it should. Multiple roles are possible. Whether they are “permitted”, resourced, and recognised is a different question entirely.

Mike goes somewhere else entirely and it is worth letting him. His frame is education. Not geopolitical, not regulatory, not linguistic but personal. He couldn’t afford the Ivy League either, and he lives in its backyard. His argument is that AI’s most transformative role in Africa, and elsewhere, may be as a distributor of access to knowledge that has always existed behind a paywall. That sits at an angle to the rest of this group. But it is not disconnected from it. If the multiple roles frame is ultimately about who gets to participate in the AI era on their own terms, then education is not a detour. It is one of the most direct routes there.

What this group is describing is not just multiplicity for its own sake. It is the argument that any single-role framing does a particular kind of violence, it flattens a continent of 54 countries into a function. Builder. Tester. User. Regulator. Each of those words, applied as a singular label, is also a limit. What these contributors are insisting on is the right to exceed the frame. Not to be assigned a role in someone else’s ecosystem, but to be present in it fully, unevenly, contradictorily, and on terms that are still being negotiated.

Whether those terms get negotiated, and by whom, and in which rooms, that is what Dr. Illingworth’s question is really asking.

On history: If the testing ground framing has a documented past, what does it take for a new framing to be more than a rebranding and who gets to decide when the shift is real?

On the gap: What does it mean to have 3% of global AI talent and 5% compute access and whose responsibility is it to close that distance?

On participation: If Africa’s role in AI will be shaped by whether the global ecosystem makes room, what would making room actually require from those who currently occupy the most space?

---

What the Role Question Reveals

Roles don’t assign themselves. Someone decides what counts as building and what counts as using. Someone decides which benchmarks matter and which problems are worth solving. Someone decides whose innovation gets funded, whose languages get trained on, whose workers get contracted and whose communities absorb the cost. The global AI ecosystem has a grammar, and like most grammars, it feels neutral until you ask who wrote it.

What this question surfaces is that contributors are grappling with that grammar whether they name it or not. Some push back against it directly. Some work creatively within it. Some are working from a partial picture which is itself a product of how the grammar operates, what it makes visible and what it quietly leaves out. And at least one contributor strips the grammar away entirely and replaces it with something the grammar prefers not to show: specific people, specific places, specific costs.

Together they reveal that the question of what role Africa plays in AI cannot be answered without first asking who gets to assign the roles. That is not a small reframe. It shifts the entire conversation from capability, what Africa can do to power, who decides what Africa gets to do, on what terms, and with what recognition. And once you see that shift, the four frames in this piece start to look less like different answers to the same question and more like different positions within the same unresolved argument about who the AI era is actually being built for.

That argument is not settled. It may not be settleable in the near term. But the fact that it is being had by African builders, by outside observers, by people working with partial pictures and people working with painful specifics matters. Because the conversation itself is a kind of pressure. And pressure, over time, is what changes grammars.

Question 3 presses directly on this. It asks what role African languages, knowledge systems, and local realities should play in shaping the future of AI globally. It is, on the surface, a question about inclusion. But underneath it is something more fundamental, a question about whose ways of knowing get to shape what intelligence means, and what it is for. The answers, as you’ll see, don’t stay on the surface.

Thank you for reading!

---

---

Support this work

Your support keeps it independent and community-rooted.

Thank you for standing with this work.