Investment in artificial intelligence is essential for our future health

Yet investment in AI is dominated by the private sector

We can use machine learning to analyse radio content and distil patterns of malaria outbreaks (AP)

Artificial intelligence may still be in its infancy, but it’s moving fast. Nowhere is this more apparent than in the data-rich health sector. AI has the potential to provide more precise, personalised care, as well as help us to shift our focus from treatment to prevention and tackle some of the world’s biggest global health issues.

The WHO estimates that achieving the health-related targets under the Sustainable Development Goals – from ending tuberculosis to ensuring universal access to sexual and reproductive healthcare services by 2030 – will cost between $134bn-$371bn (£97bn-£270bn) a year over current health spending.

AI startups raised $15.2bn last year alone, adding to investments made by tech giants like Google, Facebook, and Alibaba and a host of research institutions.

It’s not only smart for public and social sectors to start investing in AI: it’s essential.

Imagine. Where non-communicable diseases are now the leading cause of mortality worldwide, computers can scan for early signs of Alzheimer’s and melanoma. Machine learning predicts the spread of infectious diseases to help plan for and prevent epidemics. Natural language processing in chatbots delivers real-time responses to sexual and reproductive health queries in HID/AIDS-affected communities. And in low-income countries where mortality rates from car accidents are the highest, the deep neural networks in self-driving cars improve road safety.

There are several hurdles to realising our vision.

For a starting point, good AI needs good data. Machine learning, a subset of AI, uses extensive data to learn and improve without explicitly being programmed. Health data is vast, ranging from patient records and medical scans to population data and global health indicators. It is complex, as there are varied measurement standards, issues and around collection and storage, not to mention regulatory requirements.

There’s also a data gap. Millions of births and more than 38 million deaths go unreported each year. For registered deaths, 75 per cent fail to include a cause of death. A lack of information on the most vulnerable communities – minorities, the poor, those living in rural regions and in emergencies – biases systems. It is essential to create the infrastructure to collect, centralise and construct equitable datasets ensuring that every country, community and person counts.

Ethics is another issue.

“The world’s most valuable resource is no longer oil, but data,” The Economist declared last year when outlining the importance of regulating the internet giants. If we’ve learned anything from Facebook’s dealings with Cambridge Analytica, conversations on AI governance and fairness are essential to place people’s interests first.

There are several questions that must be answered here. How can we ensure that the benefits of AI don’t just apply to a small number of people by reinforcing bias and discrimination? What is under the black box of how algorithms get from point A to Z, reaching medical decisions that may be the difference between life and death? When handling confidential patient data, how can we guarantee privacy, security, and transparency to gain users’ trust?

Despite these hurdles, there are several recent examples of initiatives helping public and social sector understanding and adoption of AI innovation.

UN Global Pulse, an initiative that is working to incorporate the newest technology within the development and humanitarian sectors, analysed radio content in Uganda using machine learning to distil patterns of malaria outbreaks, the leading cause of death in the country. The Pulse Labs collaborate with local universities, co-hosting seminars to present and share research, ideas and tools.

It is also critical that we bridge the gap between the technological breakthroughs of the private sector and the social missions of civil societies, governments and NGOs. Working together will bring thoughtful perspectives and diverse talent from all sides. An example is Partnership on AI, which includes prominent tech companies and research institutions alongside a handful of non-profits like Unicef, Human Rights Watch, and the ACLU working to “ensure that applications of AI are beneficial to people and society”.

We need a sustainable, multi-stakeholder ecosystem, working globally and locally, to build the foundational safeguards to maximise the value generated by AI for the whole of society.

At the moment we don’t have the billions of dollars we need. What we do have is a wealth of data that can power promising, innovative solutions to bolster and accelerate the existing global health agenda. Investing in AI today means high returns for people’s health and wellbeing tomorrow.

Betty Chang is a researcher in gender and innovation at Unicef