In order to usher in the Fourth Industrial Revolution, Nigeria has to focus on training the necessary people on the foundations and expertise on some of the technologies involved. They also have to apply these technologies in our everyday lives. This especially applies to the field of Artificial Intelligence (AI).
We see an increase in AI training opportunities and events – some for free – and the ever-widening spotlight on AI-powered businesses in Africa and Nigeria.
According to Aniedi Udo-Obong – program manager at Google – there seems to be data everywhere. In weather, agriculture, and health, these data are abundant because they are constantly collected and sometimes measured using different methods.
“So when people think of all the problems that exist and the opportunities to solve these problems, if they want to use AI as a tool they have to go and operate in a space where there is abundant data [present].”
Unfortunately, many domains in Nigeria do not have a lot of accessible data online especially about the various industries in the economy.
A comparative look at two industries –healthcare technology and fintech – highlight the strides domains have made thanks to the accessible data that could be used for analysis and machine learning systems.
To do this, one would need to look at three factors – the methods of data collection, the abundance and accessibility of said data and the advancements the industry has made in terms business application and everyday use.
In the healthcare industry, the measurement of one’s blood pressure or pulse during a routine medical consultation makes up part of the data collected and fed to the AI model. But according to pharmacist, data scientist, and health-tech entrepreneur, Adeola Adesina, health-related data is not that easy to obtain in Nigeria.
“Healthcare data is one of the most complex data available,” says Adesina.
During a phone chat over the nature of this monster, he enumerated several reasons why the healthcare industry, though flowing with abundant data, has not ensured the application of AI on a broader scale.
One factor adding to the complexity of health-related data compared to banking or insurance is the presence of many components that come into play when dealing with health and illness. According to Adesina,
“Healthcare data is not [the same] as when we run machine learning on insurance or banking data. It is so complex that there are so many other things you need to look into. One problem, for example, is during my work [with] data on diabetes, [data] for diabetes on its own has about 50 rows.”
Another major problem that complicates data in healthcare, Adesina says, is the inherently poor bureaucratic practices found in Nigeria.
The current attitude of healthcare workers towards data collection is not inspiring. Call it ignorance or just a lack of exposure, to him probably everyone in that industry — doctors, pharmacists, nurses and others — do not know what to do with the data once collected. After after that, nothing gets done about it.
According to Adesina, most hospitals collect patient data on paper files. Even those that have an Electronic Document Management System (EDMS) only focus on using it to centralise patient data within the hospital alone.
Pharmacies are also a huge part of the equation.
“People just walk into pharmacies and purchase anything, no drug history, nothing.”
To be fair, collecting health-related data in Nigeria is not without its frustrations. Ubenwa, a Nigerian-Canadian machine learning system that analyses a baby’s cry to provide an instant diagnosis of birth asphyxia, reportedly had to obtain data of 1,400 infants crying from researchers in Mexico to power its AI engine. Although the company’s founder, Charles Onu, stated in a research paper published at Cornell University that they had obtained data acquisition approval from the University of Port Harcourt Teaching Hospital (UPTH), field trials have yet to begin.
At the other end of the data spectrum is the financial technology industry. The plethora of fintech businesses in Nigeria belie the fact that the finance industry seems to have its data organised enough to make something of it.
Domains like finance and telecommunications have abundant data because people make transactions and phone calls every day.
“So the opportunity to use AI or machine learning techniques is dependent on the availability of streaming data,” says Udo-Obong.
According to a source from a recognised Nigerian fintech startup, data strictly depends on the company and what it is trying to achieve. Consequently, there is no one way to access this data. But for most fintech companies, tracking transactional data is usually the easiest.
Transactional data – data that has a time dimension, numeric value, and refers to one or more objects – are gathered by financial technology companies that build their applications and gather data at every touch point where the user accesses the company’s services.
For instance, the use of banking apps banks can enable banks easily track the frequency of users’ transfers, withdrawals, and payments. A decision to analyse users’ activities using the information from their transactions easily gives insights into what they have been up to.
But there are different ways data can be gathered, and some methods, like a good old-fashioned survey, are enough for financial institutions to track users’ data, according to our source.
These surveys are created and sent to customers for the companies to understand how customers spend their money, which is still a form of transactional data.
These insights gained directly – or indirectly – from customers help fintechs know which services to provide as well as which market segment(s) to focus on.
Current and potential data privacy issues aside, it is very clear that data has the potential to make or break an industry in the Fourth Industrial Revolution. Without this very important resource that the Economist deemed more valuable than oil in the twenty-first century, all hope for implementing AI en masse in Nigeria may be lost.
As Udo-Obong stated during a brief chat over lunch, “You are not going to use AI to solve a problem where the domain space has very limited [data].”