We started this research with five questions:
Where might AI add most value?
What are the different types of AI?
How do you go about developing an AI model?
What are the barriers to developing these solutions, and how do practitioners address them?
What are the unique ethical issues which arise with applying AI in international development, and what can we do about them?
Our approach throughout this module has been to surface preliminary answers to these questions through the story of different pilot projects, and their journey to impact. Our focus has been on the practical side of applying AI in international development, and uncovering basic technical introductions which help make the practical tangible. We hope that through these case studies, readers can untangle the real potential of AI as it currently stands from maximalist claims about what it might do in the future.
Our reason for taking this approach is that many of the answers to these questions are contextual. The barriers for building impactful AI solutions cannot be considered in abstract from the sector in which the solution is being developed, the people it is being developed with, and the unique conditions of the place in which it’s being developed. These case studies aim to surface some of the decisions that practitioners have to consider when moving from the generic capabilities of AI to actual impact.
Limitations
It should be noted that a comprehensive framework to explore these questions, and the specific opportunities, challenges, and risks of applying AI is beyond the scope of this work. There are many key issues which we have not explored in depth, in particular the need for compute to power systems and the availability of skills to develop them.
Additionally, many of the pilots which we have explored are at a relatively early stage, and as such, the question of scaling solutions at a community, national, and international level has not been our focus. We hope to see more comprehensive analyses of the AI ecosystem in international development and the enablers for developing an ecosystem which supports high-potential AI systems to develop and improve lives at scale.