Matthew Smith, senior program specialist at the International Development Research Centre, talks with Charting Change about the opportunities and challenges of AI in transforming societies
Artificial intelligence is transforming societies worldwide. In developing countries, AI’s potential to benefit local economies, healthcare, agriculture, education and other sectors is sparking optimism and investment. Yet the challenges are great. Charting Change freelance contributor Brian Banks asked Matthew Smith, senior program specialist at the International Development Research Centre and lead author of the 2018 IDRC whitepaper Artificial intelligence and human development to walk us through some of the issues.
Charting Change: Your whitepaper cites AI’s enormous potential, but it also warns that “if we continue blindly forward, we should expect to see increased inequality alongside economic disruption, social unrest … with the technologically disadvantaged and underrepresented faring the worst.” Why frame it this way?
Matthew Smith: It’s really a call for developing a deeper understanding of these systems and how they interact with human rights. And also, fundamentally, a call for building the capacity of people working in developing country contexts, the global South, to be able to design and build and develop these technologies themselves and also to participate in the global discourse around them.
CC: You state those countries need to have effective AI policy and regulatory structures in place to ensure people benefit and their rights and privacy are protected. Where do things stand today?
MS: I don’t think we’re that far. There’s been a lot of work in Western countries. And a lot of the UN agencies or human rights agencies have done a lot of interesting thinking about the intersections between AI, say, and how it might interplay with human rights; how do you design an AI system in an ethical way; or what are the principles for responsible and ethical design of AI? But in terms of how it is in the development space, from a research perspective, I don’t see a lot of it.
CC: Last fall, IDRC helped to fund an AI research network in sub-Saharan Africa. Is that the sort of step that’s needed?
MS: Yes, that’s still being developed. There was a meeting on it in April in Kenya. The idea of bringing everyone together was to try to figure out what an African research and innovation agenda would look like. There was also a call for proposals for innovations to advance sustainable development, with winners and funding announced in August.
CC: What makes AI different than other technologies?
MS: It changes the way we can think about working, it changes the way we can organize. And it can do it massively and very cheaply.
CC: Where do you see AI having the most potential benefit in developing countries?
MS: It depends on what level you look at. The things that you probably hear about most are in agriculture. But if you look at prediction models for weather pattern changes or where areas will get flooded, they are not local innovations. From a local perspective, an area that I am personally interested in is education. I was at a workshop thinking about training teachers at scale. There are all sorts of interesting ways that you can introduce AI to help. One idea was to create a little AI coach on Facebook Messenger. It would check in periodically with teachers who are progressing to a different training module, saying, ‘Ok, how are things going, do you need any help with anything?’
CC: Is that easily scalable?
MS: It’s a package deal. The Facebooks of the world can assemble these things, put them out there and they make the barriers to development of AI and to developing something new quite low. In this case, teachers just need a cellphone and Facebook Messenger. And when that starts happening, then you get all sorts of really interesting experiments by people who are just trying to solve a local problem.
CC: Disease detection is another area that is often highlighted?
MS: In situations where you need some level of expertise, but you don’t have that expertise — say, rural areas or areas where they just don’t have enough doctors — AI systems are pretty sophisticated in doing diagnostics. It’s a kind of prediction problem: ‘What are your symptoms and what do you think it’s going to be?’ If you can get a fairly robust system to front-line health workers with mobile phones, they’ll be able to do an incredibly high level of diagnostics.
Photograph courtesy of Charting Change.