What AI Image Generation Taught Me About Progress, Identity, and Access

My name is Jeffrey Mdala, and I am an AI Engineer & Founder based in Lusaka, Zambia. I run Zambian Online Education Company (ZOEC), where I built eskulu, an AI-powered learning platform for the Zambian ECZ curriculum, and Zedpastpapers, which now serves more than 200,000 users every month. Because I work closely with AI systems every day, I pay attention not just to the big breakthroughs, but also to the small moments that reveal how fast this technology is changing.

One of those moments came when I generated images of myself and noticed something striking: AI image generation had become much better at showing my vitiligo accurately. Not long ago, these tools would miss it completely or smooth it out. Now, the pattern is visible in much greater detail. That may sound like a small technical improvement, but to me, it says something much bigger about where AI is heading — especially for people in Zambia and across Africa.

When AI Could Not See Me Properly

For a long time, AI image tools struggled with accurate representation. If you had any distinctive personal features, there was a good chance the output would ignore them, distort them, or replace them with something more generic. In my case, my vitiligo often did not appear at all.

That matters because AI is not just about generating impressive visuals. It is also about recognition. When a system fails to reflect how real people actually look, it reveals the limits of the data, the design, and the assumptions behind the model. In many cases, African faces, African skin tones, and unique physical characteristics have historically been underrepresented in global technology systems. So when I saw AI begin to capture my vitiligo pattern more clearly, I saw more than better graphics — I saw evidence of progress in representation.

Of course, the model still was not perfect. It got one detail wrong: my teeth. I have bunny teeth, and the image did not capture that properly. But even that says something useful. AI has improved a lot, yet it still has room to grow. Progress in this space is real, but it is not final.

How Fast AI Is Improving

What really stood out to me was the speed of change. The difference I noticed happened over roughly nine months. In technology, that is an incredibly short time. We are used to thinking of innovation as something that takes years, but with AI, major improvements can happen in less than a year.

As someone who has spent years self-teaching, building software, and working across engineering and AI, I find this pace both exciting and demanding. I started coding in Grade 12, and since then I have stayed deeply curious about how systems evolve. That curiosity helped me build platforms like Zedpastpapers and eskulu, and it is the same mindset that makes me pay attention to moments like this one. AI is moving so quickly that if you stop observing, testing, and building, you can fall behind almost immediately.

This is also why I continue investing in my skills through hands-on work and certifications such as AWS Lambda Foundations and Amazon Bedrock. In AI, learning cannot be static. The tools are changing, the capabilities are changing, and the opportunities for African builders are changing too.

From Premium Tools to Free Access

Another major shift is accessibility. Not long ago, generating high-quality AI images like this would usually require payment. Today, many people can create surprisingly strong results for free.

That change is bigger than convenience. It means more students, creators, entrepreneurs, and developers in Zambia can experiment with advanced tools without first needing a large budget. In an African context, that matters a lot. Cost has always been one of the biggest barriers to technology adoption. When powerful tools become free or significantly cheaper, innovation expands beyond elite circles and reaches ordinary people with ideas.

I have seen this pattern before in education technology. When I built eskulu during COVID-19, my goal was to make quality learning resources more accessible to Zambian learners. Today, eskulu has helped reach more than 500,000 students across Zambia with notes, past papers, marking schemes, quizzes, and AI-powered support for Grades 6 to 12. The lesson is the same: when access improves, participation grows.

AI image generation is following a similar path. What used to feel exclusive is becoming more available. That opens the door for African storytellers, marketers, educators, and startups to create at a much higher level than before.

Why Representation in AI Matters in Africa

In Zambia and across Africa, we should care deeply about how AI systems represent people. If these systems are going to shape media, education, business, and communication, then they must be able to reflect the diversity of the people using them.

That includes:

• Skin tones that are rendered accurately
• Facial features that are not generalized into one global template
• Conditions like vitiligo that should not be erased by default
• Cultural and local context that makes outputs feel relevant rather than imported

As an African AI builder, I believe this is not a side issue. It is central to building useful and trustworthy systems. If AI is going to serve our communities well, it must learn to see us properly.

This belief also shapes how I think about the future of ZOEC and eskulu. I do not just want to build tools that work in theory. I want to build systems that work for real Zambian students, real African users, and real local challenges. That is part of the larger mission behind my work: using AI to expand opportunity, not just imitate trends from elsewhere.

Small Moments Can Reveal Big Shifts

Sometimes people expect AI progress to show up only in research papers, billion-dollar announcements, or global product launches. But often, the clearest evidence appears in ordinary use. You try a tool again after some months, and suddenly it does something it could not do before. That is when the change becomes real.

For me, seeing my vitiligo represented more clearly was one of those moments. It reminded me that AI is becoming more capable, more accessible, and in some ways more inclusive. Not perfect — but noticeably better.

As someone whose journey has included building products from Zambia, reaching national recognition, placing in competitions like the X Pitchathon as Business With a Purpose, and continuing to explore AI across education and business, I see this as part of a larger pattern. The future will belong to those who not only consume AI, but understand it, test it, and shape it for their own context.

My Takeaway as a Builder

My biggest takeaway is simple: AI is improving fast, and access is widening. That combination is powerful. It means more people can participate. It means more African problems can be addressed with modern tools. And it means builders like us have a responsibility to create systems that are locally grounded, inclusive, and genuinely useful.

Whether I am working on educational platforms, AI integrations, or consulting through my other venture, MAY and Company, I keep coming back to the same principle: technology should reflect people more truthfully and serve them more affordably.

If AI can now better capture details like vitiligo that it once ignored, then that is a sign of broader progress. The next step is making sure that progress benefits more people in Zambia and across Africa.

Conclusion

AI image generation has come a long way in a short time. In my own experience, it has moved from missing important details about how I look to representing them with much more accuracy — even if it still got my bunny teeth wrong. That may seem like a small example, but it reflects a much bigger story about how quickly AI is evolving and how much more accessible it has become.

For me, this is not just about images. It is about visibility, inclusion, and opportunity. It is about making sure the next generation of AI tools works for people here in Zambia and across Africa, not just for users in markets that have always been prioritized.

If you are interested in AI solutions, edtech development, web platforms, or generative AI integration, feel free to reach out to me at jeffmdala@gmail.com. And if you are a student, parent, or school looking for quality ECZ learning support, explore eskulu — the platform I built to help bring AI-powered education to more learners across Zambia.