How Artificial Intelligence could empower SMEs in Africa
Artificial Intelligence (AI) is a major driver of the fourth industrial revolution (4IR), concerned with creating machines that mimic human-like intelligence. Today, AI is creating tremendous possibilities for creation and invention, most recently being AlphaFold, which can accurately predict 3D models of protein structure and will accelerate our understanding of biology.
AI is an enabler of innovation and wealth creation. However, AI is not spread evenly. This article outlines how small, and medium enterprises (SMEs) in Africa, including Namibia, can benefit from the power of AI.
In a recent Ted Talk, Andrew Ng compared the rise of AI to the rise of literacy, whereby many years ago, people thought that not everyone needed to learn how to read and write. The people mostly depended on the high priests and monks to read the Holy Scriptures. However, literacy brought about a much better and richer society, allowing people to express their thoughts or ideas in written form. Similarly, the same can be said about AI. Today, AI is being shaped by a few highly skilled engineers. These engineers are the high priests, priestesses and monks that develop and maintain AI systems that tell us how to live and work.
The majority of these skilled engineers “The high priests, priestesses and monks” of AI work for big tech companies and many of us have only access to the AI they build such as Siri, Alexa and other applications we use to manage our health. Thus far, it is the big technology companies that are milking the benefits of AI, because AI systems require a lot of data and are expensive to build.
It is the large tech companies with the money and the millions of users, enabling big tech companies to create one-size-fitsall AI systems. For example, the algorithm that recommends what to watch on Netflix or Ads you will like on Facebook. For big tech companies, the use of one-size-fitsall AI can generate massive amounts of revenue. However, outside the big tech sector, this approach to AI will not work. For example, in the SME sector, there is hardly any project on AI because SMEs do not have 100 million customers. Thus, it does not make economic sense for SMEs to invest in AI.
Whenever I visit my grandmother in Endola, I always watch my grandmother with admiration as she bakes and sells her vetkoek. By selling vetkoek, she is generating data. If she had access to AI, she could leverage data to spot demand patterns to identify days of the week when the vetkoek sells well. AI can improve decision-making of SMEs, for example, if the vetkoek” sells well on a Friday afternoon, she may not need to wake up early on Friday. Rather she can use insights from data to determine the amount of vetkoek to make on a given day. Thus, maximising profits.
I know, AI requires large datasets, and large datasets can help improve the performance of AI. Contrary to the massive dataset, fine-tuning AI can work fine with modest data, such as data gathered by a single SME shop or vendor. The biggest problem is not the data, however, the small shop owned by my grandmother can never justify the cost of building an AI solution.
In Namibia, we have many SME vendors like my grandmother who sell vetkoek and collectively they serve thousands of customers. Yet every SME vendor is different, they serve different customers and record sales differently. So, no one-size-fits-all AI solution will work for all of them. Hence the need to build custom AI solutions for each business.
Typically, to build an AI solution, one needs to know how to code and understand complex mathematics (linear algebra and statistics). Even though I am a strong advocate of people learning how to code, I know that not everyone has the time to learn to code. New low-code AI-development platforms have been developed to enable people without coding skills to build AI algorithms that matter to them. Instead of writing 1000s lines of code, you provide data and finetune pre-trained models with custom data. Like how pen and paper were instrumental to widespread literacy. I think low-code AI-development platforms can help make AI accessible to all.
For example, an inspector of a start-up can use AI to detect defects in leather shoes by taking pictures of the shoes, and upload them on the low-code AI development platform. After they label the data to show the AI what tear, shear or discolouring on leather shoes look like. The labelled data is used to teach the AI what defects in leather shoes look like. After the AI learns this data, it might learn better to identify tears but not discolouring. The inspector can add more data showing discolouring to help the AI learn that better. Using low-code AI-development platforms, an inspector will be able to train their own custom AI in a few days.
Likewise, low-code AI-development platforms can empower a baker to use AI to check the quality of the cakes or a farmer to check the quality of vegetables or a carpenter to check the quality of the timber they are using. Platforms like this might need a few years to become easier for my grandmother to use.
Nevertheless, some of these platforms are useful to someone tech-savvy with a bit of training. Therefore, instead of relying on “The high priests, priestesses and monks” to write AI systems for everyone, we can empower store managers, accountants and some SME owners to build their own AI systems to facilitate decision-making, improve marketing, improve sales and streamline business operations among other benefits.
No doubt coding is the new literacy and understanding AI is the next frontier of literacy that we all need. A wider understanding of AI systems has the potential to empower more people and could be Africa’s last chance for sustainable and inclusive growth.
*Lameck Mbangula Amugongo is a post-doctoral researcher at the Institute for Ethics in Artificial Intelligence (Technical University of Munich). He holds a PhD in Cancer Science from the University of Manchester. The views expressed are his own.