AI and development economics: Early evidence and how to keep up

I have been (trying) to keep track of the latest AI-related thinking and research that is relevant to low- and middle-income countries. It is hard to know where to look and what to trust, here are the best resources and platforms I have come across.

Editor’s note: I updated this blog on April 1^st (and March 11^th and November 11^th before that) to include more useful links. We are a long way from satisfactory answers to the important questions, but if you are interested in this area, there is no shortage of interesting resources to dive into.

It is increasingly difficult to keep up with the overwhelming volume of AI-related (and AI-generated) content. We have tried our best to at VoxDev, and in the process I have built up a bank of sources that I trust and find useful when it comes to the potential impacts of AI.

In this blog, I have compiled these in one place, including: early evidence on the impacts of AI from development economics; what development organisations are working on; how development economists can use AI for their research; and the resources that I am using to stay up to date.

Almost everything in this blog was included in VoxDev's weekly newsletter, so sign up here to stay up to date on all things development. And if you think I’ve missed something, please send it over ([email protected]) so it can be added to this blog, which I will continue to update.

Ideas in Development: A new series of podcasts on AI

Putting together previous editions of this blog highlighted to me just how little thinking and research is out there on the economics of AI in low- and middle-income countries. I wrote about why this is such an important issue on my Substack - There is no randomising a technological revolution.

So, Deena Mousa and I have been working on a series of podcasts on the development economics of AI which we hope helps to fill this gap.

You can check out these episodes of Ideas in Development wherever you get your podcasts (Apple Podcasts, Spotify, YouTube, Everywhere Else), we also release write ups of each episode on our Substack, and the VoxDev website:

The development economics of AI: Lessons & questions Oliver Hanney & Deena Mousa
AI policy in developing countries Umar Saif
AI, India, and the future of service-led growth Raghuram Rajan
Technology diffusion: The role of venture capital, universities and China Josh Lerner
Three ways India is using AI for development Claire Cullen, Utkarsh Saxena & Niriksha Shetty
Thinking like an economist about AI, labour markets, and AGI Anton Korinek
AI in Africa: Barriers, opportunities and policy Rose Mutiso
AI and the industrial revolution: Similarities, differences and lessons Bruno Caprettini
Economists vs Technologists on AI Oliver Hanney and Deena Mousa

What is generative AI and how is it improving?

I have two recommendations for learning how generative AI works and then keeping track of improvements:

Watch this video. I think it is worth the time (it breaks down into three distinct sections if you can’t face a single three hour watch). If you think about how long you will spend using these models in your lifetime, I think this is a fair investment. In the video, Andrej Karpathy does a great job explaining how LLMs like ChatGPT work in an accessible way.
Follow Ethan Mollick on social media (LinkedIn, Twitter, Bluesky) and Substack to keep up to date with the rapid pace of technological advancements.
1. This guide of which AI to use in the agentic era is must read!

Early evidence on AI’s potential impacts in low- and middle-income countries

We have featured a wide range of research on VoxDev that sheds light on positive and negative impacts of AI across a range of topics, from jobs to learning to the surveillance state:

Generative AI as business trainer or consultant: In our recently updated Training Entrepreneurs VoxDevLit, David McKenzie and Chris Woodruff include new evidence on AI.
How will AI impact jobs in emerging and developing economies? Marina M. Tavares, Augustus J. Panton, Carlo Pizzinelli, Mauro Cazzaniga and Longji Li
AI hiring tools exhibit complex gender and racial biases: Jiafu An, Difang Huang, Chen Lin & Mingzhu Tai
Calling all parents: Boosting early childhood outcomes in Uruguay: Juanita Bloomfield, Ana Inés Balsa, Alejandra Cid & Philip Oreopoulos
Improving learning efficacy and equality with AI training: Zhengyang (leo) Bao, Difang Huang & Chen Lin
How the surveillance state is exported through trade in AI: Martin Beraja, Andrew Kao, David Yang & Noam Yuchtman.
Autocratic AI dystopias: Science fiction or social science fact? Martin Beraja, Andrew Kao, David Yang & Noam Yuchtman.
AI, entrepreneurship and development: Podcast with Rembrand M. Koning.

And on our VoxDevTalks podcast, we have hosted bigger-picture discussions with leading economists working on AI:

What role can (should) economists play in shaping the future of AI? David Yanagizawa-Drott
Geopolitics, AI & the future of global development: Simon Johnson
David Autor on AI and the future of work.

What I’ve been reading, listening to, and watching outside of VoxDev

Some new resources from the past few months:

Everyone should be following Han Sheng Chia, who's the Director of the AI Initiative at the Center for Global Development. I'd highly recommend his recent piece on Reimagining Tech for Good. I also enjoyed his blog with Daniel Björkegren and Andrew Bredenkamp outlining a roadmap to AI that speaks the world's languages.
Johan Fourie's Substack - Our Long Walk - has a great piece on what South Africa should learn from the people already using AI.
W. Gyude Moore also discusses AI on his Substack, the Africa Project.
The Afroconomist delves into the dynamic world of business, economics, and culture shaping Africa's future, with lots relevant to the AI in Africa debate.
I wrote on the future of communicating science.
J-PAL’s AI Evidence Playbook – a detailed practical guide for policymakers, practitioners, and donors investing in or developing AI-enabled programmes.
I’ve really enjoyed Karthik Tadepalli’s blog on technological progress in developing countries – Beyond Imitation.
Rose Mutiso’s Substack Kibao – on AI, tech, energy and more!
Rose Mutiso has also set up the African Tech Futures Lab, which is well worth monitoring for all things energy, climate and technology in Africa.
Alex Imas has an amazing Substack – Ghosts of Electricity – on the economics of AI.
On The Argument, Matthew Yglesias writes we may miss the sweatshops.
The economics team at Anthropic are producing lots of useful resources and insights.
From the UNDP: Why AI may widen inequality between countries

Plus some reading from the last update in November:

Deena Mousa, a Works in Progress + Asimov Press Writing Fellow who is covering AI's impact on developing nations, investigates why radiologists have not been replaced by AI.
J-PAL is releasing a series of blogs on AI for social good. In the first of the series, Attaullah Abbasi, Sam Carter, Katie Chen & Iqbal Dhaliwal highlight six promising opportunities for AI to contribute to social good.
I had quite a few questions about AI’s energy use after the first edition of this reading list, thankfully Hannah Ritchie has summarised the limited current evidence on individual AI footprints, which mainly relies on self-reporting by AI companies.
- Hannah Ritchie also has good posts on how AI could improve weather forecasting, and AI’s total impact on energy demand.
On CGD, a podcast with Temina Madon and Han Sheng Chia on where AI meets development, and Daniel Björkegren discusses why SMS charges might be limited AI use for a billion people.
Maria Barron has outlined five lessons from the Global South for people-centered AI in education.
On IDB Nicolás Ajzenman, Ana Teresa del Toro Mijares, Gregory Elacqua, Catalina Hermosilla and Santiago Veleda discuss what they learned from scaling career guidance in Chile by moving from human counselors to AI agents.
Princeton Professor Arvind Narayanan has started a YouTube channel called Arvind on AI. The first video explores what happens if there's an AI crash?
Charlene Peters Chen outlines Ezrah’s thinking on AI for development in Africa.
IFPRI’s special blog and webinar series on AI for Food Systems Research.
- E.g. Using LLMs to code survey responses in native languages.

And some other resources from the original version of this blog:

Also from Han Sheng Chia, a range of insightful writing on generative AI, from powering government portals, to smart bets, to nonprofit AI use.
- Han Sheng Chia, Temina Madon, Sid Ravinutala and Donald Lobo outline four investments funders can make to raise the bar on AI evaluation.
AI and the future of development: This is a really useful write up on David Nash’s Global Development & Economic Advancement blog. And his monthly reading list has a section dedicated to AI which is well worth checking out.
If you are reading this, I imagine you have already seen this research on the World Bank Blogs: From chalkboards to chatbots: Transforming learning in Nigeria, one prompt at a time by Martin E. De Simone, Federico Tiberti, Wuraola Mosuro, Federico Manolio, Maria Barron & Eliot Dikoru.
From The Times of India “AI Can Derail GCC Story. Hereʼs What India Must Do” by Rohit Lamba & Raghuram Rajan.
On Brookings: Artificial intelligence is transforming middle-class jobs. Can it also help the poor? Luis Felipe López-Calva, Carlos Rodríguez-Castelán & Hernan Winkler
This syllabus for an upcoming summer course at the Digital Economy Lab in Stanford lists some of the key papers on AI’s impact on economic growth and labour.
Mapping the Potentials and Limitations of Using Generative AI Technologies to Address Socio-Economic Challenges in LMICs.
The Humanitarian AI today podcast covers AI applications in the humanitarian sector – Listen on Spotify.

Plus, some videos I would highly recommend:

The CEPR Webinar Series on the Economics of Artificial Intelligence. Register on the webpage, and catch up on the previous presentations and discussions on YouTube.
At the Harvard GEM Conference, leaders in AI convened to discuss AI innovations in emerging economies, and what emerging economies want from AI – recordings available on YouTube.
Two videos from the UBS Center’s Economics for Society channel with David Yanagizawa-Drott (AI Opportunities for the Global South), and Sendhil Mullainathan & David Yanagizawa-Drott (How AI can truly change the world).
AI and Economic Growth – an overview lecture of several of Chad Jones papers on artificial intelligence, economic growth, and existential risk.

How can development economists use AI?

AI has a range of use-cases for researchers. At this VoxDev event, Aniket Panjwani discussed how development economists can get set up to use agents for research, focusing on the cheapest options currently available, and general lessons for working with agents. Here are the slides from the presentation.

Aniket Panjwani has a useful YouTube channel of his own, be sure to subscribe!

Chris Blattman has created a one stop shop for using AI tools at all stages of knowledge work – Claude Blattman.

In 2024, Anton Korinek joined us to discuss how researchers and students in low- and middle-income countries can use generative AI for economic research. Things have progressed rapidly since then, so I would recommend checking out Anton’s companion website, Generative AI for Economic Research, where you can find the latest tips on enhancing your research with the power of generative AI. Sign up to Anton Korinek’s newsletter to keep up to date as these models evolve. In the latest edition, AI Agents for Economic Research, Anton demystifies agents and explains how you can make your own.

More recently, Han Sheng Chia, Samantha Carter and Farhan Abrol presented their four level approach to evaluating potential applications of AI in the development sector.

In terms of a concrete example, this recent article outlined how Ananth Balashankar, Philipp Zimmer, Lakshminarayanan Subramanian & Samuel Fraiberger created an AI-driven early warning system uses news data to predict food crises up to 12 months in advance.

I’d also re-flag the Ideas in Development episode with three organisations in India, who talk through their experiences implementing AI on the ground.

And I summarised how we are using AI at VoxDev for editing, writing, images and ideas.

What are development organisations already working on?

Impact AI, a DIME (World Bank) project, will launch soon. (I've seen a preview, it looks amazing!)
The Agency Fund, in collaboration with CGD and Open AI have launched the AI for Global Development Accelerator – check out the range of projects they are funding for an insight into how and where organisations are already implementing AI.
And here is the next cohort of Google’s Generative AI accelerator.
GiveDirectly use AI/ML algorithms to quickly target people during crises, they have useful reflections on how to explain their use of AI across the digital divide here.
J-PAL have launched their partnership for AI evidence.

Other AI-related research and thinking

When thinking about AI’s impacts, we can potentially learn from how other transformative technologies have changed societies and diffused across the world. Here are some previous VoxDev articles looking at technology and innovation in developing countries: