Kenya ready for Artificial Intelligence but where is the data?

Lack of access to free, up-to-date and machine-readable government data stands in the way of Kenya tapping into opportunities brought by artificial intelligence, a Nation Newsplex analysis reveals.

The country tops Africa and ranks 52 globally out of 194 countries on the government’s readiness to adopt artificial intelligence (AI) in public services but only manages positions seven in the continent and 78 out of 94 globally in availability of government data, according to two key reports.

The Government Artificial Intelligence Readiness Index 2019 report, by the International Development Research Centre (IDRC) and Oxford Insights, factors in 11 inputs grouped into four main categories: governance, infrastructure and data; skills and education; and government and public services.

No African country features in the top 50 AI readiness index, whereas only 12 are in the top 100.

“Not only will they not reap the potential benefits of AI, but there is also the danger that unequal implementation widens global inequalities,” the report observes.

The Global Open Data Index 2016/17 (GODI) by the Open Knowledge Network Foundation, on the other hand, gauges the availability of government data, and hints at how much data might be available to model and train AI algorithms to carry out various tasks.

Kenya scores a measly 15 per cent, meaning that there is little useful government data easily available for AI experts or learners in Kenya to use to develop credible AI solutions. This despite access to information, which includes data, being a citizen’s constitutional right. In comparison, the leading country, Taiwan, scores 90 per cent.

According to the Access to Information Act of 2016, every citizen has the right to access to information held by the State, except in special circumstances such as when it undermines national security or impedes due process of law.

While the government has made efforts to ease access, this information is largely not in as many formats as possible for widespread consumption, in this case, machine-readable formats such as CSV, XLS, XML and JSON.

Artificial Intelligence, which involves training computer systems to do tasks normally done by human beings, has been embraced by developed nations as the next frontier of development. The IDRC report cites concerns over low and middle-income countries located in the southern hemisphere lagging behind as their northern and more developed counterparts move to enjoy the gains availed by AI, such as improved quality of service and citizens’ experience of government.

In Kenya, it would come in handy in areas such as boosting education outcomes through personalised learning, disease control, citizen inquiries, weather forecasts and predicting food security issues.

Over the last decade, Kenya’s total value of investment in AI is estimated at Sh13 billion, which is tiny compared with South Africa’s Sh165.8 billion and Nigeria Sh60.3 billion, according to a Microsoft’s Artificial Intelligence in Middle East and Africa 2019 Outlook report. This investment is in the form of support for start-ups and entrepreneurs, with the government investing little.

Many ICT experts believe that data, more than any other factor, is key to the development of AI in the country and there can’t be take-off if that is not sorted out. “Data is the engine of AI, since AI creates algorithms for predictions mainly based on historical data. If that’s missing, the power and impact of AI will be sub-optimal, or worse still, have a negative impact,” explains John Walubengo, an ICT lecturer at Multimedia University and a technology blogger, who was also part of the Blockchain and Artificial Intelligence Taskforce, whose mandate was to provide the government with recommendations on how to harness these emerging technologies over the next five years.

One of the recommendations made by the task force that the government formed last year is to increase data availability.

“Local public data tends to be in short supply. The availability and amount of such public data would also need to be improved to ensure the development of AI solutions that address local needs,” states the task force report published in July this year.

In 2011, Kenya launched its first open data portal, providing the public government data in machine-readable format. However, several Newsplex spot-checks on the portal showed that some of the datasets are raw and not up-to-date.

South Africa leads the continent in access to government data with 40 per cent, followed by Tunisia (22 per cent), Tanzania (20 per cent), Zambia (19 per cent) and Lesotho and Ghana with 16 per cent each.

The openness of government data is measured under six factors in 15 categories. The factors include if it is openly licensed, downloadable at once, up-to-date, publicly available, free and in a machine-readable format. The Kenyan government scored nil in 10 out of the 15 categories analysed. However, it scored 45 per cent each in government budgets, national statistics, procurement, national laws and draft legislation. In the five areas, the information was available free of charge and publicly accessible, but none of it was in machine-readable formats that can be easily fed into algorithms.

Australia and the United Kingdom have the second highest open data score (79 per cent each), followed by France (70 per cent), Finland, Canada and Norway (69 per cent each). Myanmar scores the least (one per cent).

The World Economic Forum lists resistance to change as one of the five major barriers to AI adoption in government and public entities, owing, in part, to their long-established processes.

“In parts of the private sector, a strong culture for experimentation encourages employees to innovate, and positive performance is rewarded. In government there can be less encouragement for employees to take risks,” it says.

This willingness to adopt AI in the private sector is visible from responses given by executives of various companies in the Middle East and Africa. Three in four companies studied say the wish to use AI for predictive analysis such as forecasting customer behaviour and maintenance of their machinery, according to the Microsoft report.

Artificial intelligence and open data aside, Kenya and African countries are generally ranked lowly in embracing information technologies in the public service. The country ranks 122 out of 193 countries globally and 11th in Africa in the 2018 e-Government Development Index by the United Nations, which measures how a country is using information technologies to promote access and inclusion of its citizens.

Similarly, in the 2019 Change Readiness Index by KPMG, which measures a country’s ability to respond to changes like new technologies and climate change, the Kenyan government, at position 68 out of 140 nations ranks lower than the civil society (64) and enterprises (63).

To address the government’s poor showing, Mr Walubengo proposes the creation of progressive policies that would provide safe spaces for public sector experimentation while at the same time shielding the decision-maker from any form of punishment.

Second to the data challenge is the high demand and low supply of AI experts, as most people with such knowledge and skills are found in the private sector, at a cost too expensive for governments to hire.

“There are few AI groups centred in sharing AI knowledge practice, but they are focused on enterprise-type of AI interventions, largely in the banking, insurance, mobile lending and transport sectors,” says Mr Walubengo.

He adds that counties could also benefit a great deal from AI, especially in areas such as agriculture and health if policymakers in the devolved units would engage with AI experts.