Why Hasty AI Regulation Could Hurt Africa

“It’s embarrassingly hard to get Google Gemini to acknowledge that white people exist,” remarked venture capitalist Debarghya Das on X. This, along with other “white lash” sentiments, led to the pausing of Google’s latest artificial intelligence (AI) tool, Gemini. The Gemini debacle roused diverse opinions. For people of color, particularly Black people, the mishap appeared to be poetic justice, given that Black people’s history and existence has often been the target of censure and erasure. On the other hand, AI-cautious factions regarded this as an opportunity to cement the regulation narrative. If the bots can make the mistake of wiping a whole race's existence, what else can they do?

In part because of public controversies like this one, AI governance has dominated the global conversation in recent months. And 2023 saw a significant rise in AI regulation. On October 30, 2023, US President Joe Biden issued his executive order on AI. China instituted a number of regulations governing different aspects of AI, such as the draft Measures for Generative Artificial Intelligence Services. The European Union passed what is regarded as the world’s first comprehensive regulation on AI: the AI Act. These regulatory moves address numerous concerns regarding AI, such as transparency, safety and security, discrimination and bias, Big Tech domination, and misinformation. The recent trend in regulation has prompted heated conversation everywhere that AI is reshaping the future, including Africa.

But too often, regulatory attempts do not consider ecosystem heterogeneities. Of course, AI regulation will eventually come to Africa, but the timing of its arrival is crucial. Given the nascency of AI technology in Africa, imposing regulation now could detract attention from the pressing challenges facing the African tech and AI ecosystem, stifle innovation and undermine the transformative power of AI on the continent.

A functioning ecosystem is a prerequisite for effective AI regulation. But AI remains an emerging technology in Africa. The ecosystem needed to impose and enforce regulatory measures effectively is almost nonexistent in most African countries. What’s needed now is an in-depth evaluation of the current state of the AI ecosystem in Africa focused on four essential elements as seen through the lens of my organization, Qubit’s proposed ‘four horsemen’ framework, comprised of Data and Data Systems, Digital Infrastructure, Talent, and AI markets.

Data is foundational to effective development of AI. Models require vast amounts of high-quality data for training, testing and validation. In this regard, Africa faces serious data-set constraints, including unavailability, enormity, accessibility, inconsistency and relevance. These limitations currently hinder the development of bespoke AI applications, which perform best when trained on contextual data. Although indigenous data creation initiatives have emerged across the continent, the enormity of the task requires significant public sector investment to create the needed volumes of high-quality localized data.

Aside from dataset limitations, accessing affordable and efficient computing is a big challenge for many African AI researchers. Training, refining, deploying and running advanced AI models demands significant computing power and financial resources. To put this into perspective, consider the cost of training a GPT model, estimated to be over $4 million. Presently, many African researchers rely on cloud providers for compute services, but the prices can be exorbitant.

As AI-powered technologies become embedded in everyday life, the demand for computing resources shows no sign of slowing. Given the current computer system infrastructure, this is a tall order for the continent. Of the top 500 most powerful commercially available computer systems known to us, none are in Sub-Saharan Africa, and only one is located in Africa—in Morocco. Without adequate investment into computing infrastructure, AI in Africa is off to a slow start, which hasty regulation would only exacerbate.

Beyond data and computing, African policymakers must address human capital constraints within the ecosystem. Increasingly, AI startups are sprouting across the continent. These companies require talent to advance the design and development of solutions specific to Africa. Currently, very few institutions on the continent offer AI and machine learning (ML) courses, a situation that is compounded by geography. As educational institutions in some African countries do not offer any AI and ML courses, aspiring students are required to relocate or study these crucial subjects remotely, which has downsides. Taking into account socio-economic constraints, relocation is a tough call for many. Establishing first-rate AI and ML curricula at the country level across Africa, enhancing the visibility of these courses and increasing the share of female talent is imperative to raising human capital levels.

Ultimately, markets are crucial to the vibrancy and sustainability of the AI ecosystem. In Africa there are fundamental questions on this front. To what extent is the general population cognizant of the ability of AI-related interventions to meet their needs, or the dangers of AI systems? Can AI solutions in Africa, with limited data pools and nascent infrastructure, generate profit? Amidst biting constraints – such as limited rural development, low literacy levels, and financial inclusion – how can AI practitioners design and build contextually relevant solutions? These questions are critical to the commercial viability of the ecosystem. Although ecosystem players continue to exert themselves in their individual capacities, they lack the infrastructure needed to address these questions at scale.

The Gemini mishap, a clear sign of how AI can go wrong, appears to support imposing global legislative guardrails. But given the relative newness of AI technologies in Africa and their tremendous potential to serve the continent’s pressing needs, such guardrails should not be so limiting as to stifle innovation. African countries deserve a fair chance to grow their tech ecosystems without being straitjacketed and without falling prey to AI pitfalls.

Enforcing ‘hard-core’ regulation on the African AI industry at this point in time would stifle both innovation and economic growth. For now, existing regulatory frameworks on privacy and data protection are sufficient guardrails. AI policy initiatives should be directed towards alleviating ecosystem bottlenecks in Africa. Otherwise, there will be no ecosystem to regulate.