When the Covid-19 epidemic reached a point in March 2020 at which trustworthy information about the virus, its impact and spread was becoming difficult to come by, the World Health Organisation (WHO) picked a fourth industrial revolution (4IR) South African public benefit organisation (PBO) to supply reliable, automated information in the form of easy to utilise WhatsApp messages. With this WHO Health Alert useful communications service, users simply send the word “Hi” to a dedicated number specific to their country to receive real-time, reliable news and information regarding Covid-19.
Started as an English language service, WHO Health Alert is now available in a large variety of languages such as Arabic, French, Hindi, Italian, Spanish and Portuguese. In South Africa, where the service was launched as Covid-19 Connect, the interactive WhatsApp helpline is already available in Sotho, Zulu, Afrikaans and Xhosa, in addition to English.
In April 2020 an additional WhatsApp Covid-19 self-assessment service as part of Covid-19 Connect was launched in South Africa for the National Department of Health. Called HealthCheck this tool allows for early detection, mapping and management of Covid-19 cases. By now more than one million South Africans have already used the HealthCheck Covid-19 digital self-assessment tool.
One of the remarkable aspects of WHO Health Alert and its associated services is the fact that this advanced machine learning product has its roots in two technology orientated public benefit organisations with strong South African ties. Originally incubated within the mobile technology orientated Praekelt Foundation, WHO Health Alert and related products are powered by Turn, a stand-alone software as a service (SaaS) product with significant machine learning capability.
According to Lieze Langford, the London-based growth and business manager at Turn, the origins of WHO Health Alert can be traced back to a Praekelt Foundation product called MomConnect. “Initially MomConnect was an SMS and unstructured supplementary service data (USSD) service for the South African National Department of Health to reach pregnant mothers with reliable, useful information. Turn was born when we started using WhatsApp as a communication channel to drive engagement at scale for MomConnect. Today Turn is a standalone product that helps social impact and non-profit organisations to connect to end-users through the WhatsApp application programme interface (API), via a simple user experience.”
Praekelt Foundation, created by mobile visionary Gustav Praekelt 12 years ago, aims to create products, services and experiences that address the real needs of people, especially through mobile technology. When mobile penetration was still as low as 10% of the South African population way back when, Praekelt saw the long-term potential of instant connected conversations from the palm of your hand.
“I’m fortunate enough to have witnessed three technology revolutions,” he remembers. “The internet came first, mobile second and now, with universal access, the third revolution is machine learning and artificial intelligence that allow conversational interfaces for almost everyone in the world.”
Types of artificial intelligence explained
According to Praekelt, the popular perception of artificial intelligence (AI) is machines that can act autonomously like humans do. But this overarching ability of computers to be able to mimic humans remains a long way off. “There’s an old joke about AI that goes something like: ‘The definition of artificial intelligence is what computers can’t do’,” he says. This kind of AI can be called artificial general intelligence and includes self-driving cars and robots that are human-like in their abilities and behaviour. Science fiction movies often feature such popularised artificial general intelligence.
Although completely autonomous machines that totally act like humans do not exist, over the last few years great strides have been made towards this ultimate reality. Praekelt calls what is already practically possible with computers, artificial narrow intelligence. “Specifically, deep learning is a set of techniques that models a very large set of available data and uses unsupervised reinforcement learning within a set of rules, coupled with algorithms that learn how to act on the data as if it were human intelligence.”
The reason why AI is talked about so much these days therefore relates specifically to artificial narrow intelligence, a field in which great strides have already been made.
Practically, this manifests in useful, but perhaps boring and repetitive tasks that can be performed better by computers than humans. “This only works in narrow domains,” Praekelt explains. “Chess programmes now regularly beat humans and a computer can now beat a human in the complex, abstract strategy board game Go.”
This kind of machine learning is already very useful in defined areas, such as health. “We can train an algorithm in a specific domain to act as well as humans do, or even better than humans in precise instances. Algorithms have been used successfully in detecting cancer tumours, for instance. In this example the programme can be trained by specialist human experts such as radiologists and oncologists, using a large range of magnetic resonance imaging (MRI) scans, to eventually be able to detect possible cancer tumours quicker, easier and even more accurately than humans. But only in this very narrowly defined, particular domain.”
Regarding Turn, Praekelt says its machine learning capability therefore rests within narrowly defined, specific domains, such as accurate, useful and easily accessible information about Covid-19 via WHO Health Alert. “Turn’s ability to leverage artificial narrow intelligence in the health space is what drives its product development,” Praekelt says.
Is AI taking our jobs away?
“No,” says Praekelt. “AI is not a panacea. The positive answer is actually, yes, but it will take away boring jobs. For very repetitive things, we now have algorithms that can do that. Humans will be relieved of doing boring, repetitive tasks.
“But it also means we have to train human beings to be able to use these incredible processing machines we already have [the brain], to deal with novelty. It means our educational system and the way that we learn must change. Things like memorising a whole bunch of facts becomes less important, because this is the kind of thing an algorithm can do really well. Pattern matching, answering questions that have been answered before and very simple customer service are chores computers can already do better than humans.”
Novelty is not going to be automated, Praekelt continues. “Complex questions, with subtle nuances cannot be handed over to an algorithm to answer. For tasks that require the ability to deal with novelty, you need humans.”
Inspiration from Turn
In explaining the purpose of Turn, Langford says they exist to solve problems. “We’re a small, agile, problem solving team. We ask ourselves how we can solve a problem first and then how we can scale the solution to benefit large numbers of people. This philosophy helped us to launch the WhatsApp health alert services so quickly.”
Apart from WHO, Turn has customers around the world in countries such as India, other parts of Africa and Latin America. “We aim to get Turn into the hands of impact teams around the world who need to reach their customers on WhatsApp, in health and other verticals such as agriculture and financial literacy,” Langford concludes.
The success of Turn comes from improving the lives of all people and it is a truly inspirational story of how a local South African managed public benefit organisation can change people’s lives for better around the world. In the fourth industrial revolution, Turn is the human face of artificial intelligence.