The list of tech failures in development is long.  Whether it be drone pilots getting in the way of emergency relief in Nepal, or computers gathering dust in Indian schools because teachers don’t know how to use them, the outcome is rarely positive when techies fall in love with their preferred solutions rather than taking time to understand real problems faced by real people.

The best example of a technology that has driven great development gains in recent years is of course the mobile phone.  But mobile’s extraordinary impact in Africa did not flow from some grand design.  Instead, it was the result of African citizens innovating their way around information and money transfer problems in the absence of banking systems and other infrastructure.  That’s why Kenya leads the world in mobile payment systems.  

Given all this, only a fool would make predictions from London about the opportunities, and risks that artificial intelligence will bring to developing countries.  But bear with me, because that is what I’m about to do, based on the most comprehensive survey of work to date in this emerging field, which we have just published in conjunction with Sir Tim Berners-Lee’s Web Foundation, based on interviews with local experts all over the world.

Around the world individual researchers and small teams are working on AI solutions to deliver services where currently they are absent or inadequate.  In Uganda, mCrops is developing AI image-processing tools to help farmers in remote areas  diagnose crop disease.  Geekie, an adaptive learning start-up, provides tailored virtual tutoring to pupils in Brazil.  And in Nigeria, Farouq Oyebiyi, a machine learning engineer, is piloting automated image recognition to improve diagnosis of malaria from blood films.

Elsewhere researchers are dreaming of using AI to strengthen democracy.  For about a decade Google Translate was almost entirely (and often comically) useless.  Then almost overnight, thanks to advances in machine learning, it became frighteningly good. Extraordinary advances in translation, voice recognition, and natural language processing create opportunities for greater political inclusion, especially for those who speak a minority language or who are illiterate.

So far so good. But aren’t we meant to be scared witless by the downsides of the AI revolution? We’ve all read the stories about machines taking over our jobs, and the World Bank says that jobs in developing countries are even more at risk of automation. As algorithms become cheaper than people, manufacturing and low-skilled service jobs that are carried out by workers in low and middle income countries may be ‘reshored’ and replaced by work done by machines in rich countries.  The very things that made certain jobs easy to offshore may make them easy to automate.   

This means the economic path pursued by the East Asian Tigers, and later by China and India, tapping into a comparative advantage of large pools of cheap labour, may be blocked to countries in the future.  Instead, low and middle income countries may need to pursue strategies based on building skills and expertise, leapfrogging older technologies, and filling new niches in the global economy.

AI could also undermine democracy. Authoritarian regimes will use AI-based surveillance tools to help identify and target political opponents. And a lack of algorithmic transparency will make it harder to hold decision-makers to account.  AI-generated filter bubbles and news silos will limit the variety of information upon which citizens make important decisions. And politicians in developing countries will surely exploit AI-powered online profiling to target voters with particular messages.

And while the potential for AI to improve delivery of public goods such as healthcare and education is tantalising, these benefits could be out of reach for some time to come. AI is only as good as the data on which it is trained and the lack of availability of good quality local training data, is a major restriction. In Ghana many records are still paper based, while we heard of projects in Nigeria resorting to approaching individual hospitals for data, with limited success.  

The answer to the question “Will AI be more of an opportunity or more of a risk for developing countries?” is of course, “it depends”.  And what it depends on is whether the right conditions are in place to maximise the opportunities and minimise the risks. Skills and infrastructure must be built. Data must be made available, but privacy and accountability guaranteed. Serious leadership is required from within politics and civil society.  But the greatest responsibility of all is surely with Google, Facebook, Amazon, Apple, and Microsoft to ensure that this story ends well.  They should start by further opening up their ‘Partnership’ to include many more voices from the developing world.  It is time to start laying the foundations for a truly global approach to AI development that maximises the opportunities and minimises the risks for everyone.