In our fourth and final episode of the AI for Health Series, we take a peek under the hood of the maternal health chatbot supporting pregnant woman delivering in public health facilities across South Africa. To begin, Debbie Rogers, CEO of Reach Digital Health, shares her personal take on the origin story of MomConnect, South Africa’s maternal health chatbot. Sid Ravinutala of IDInsight then takes us on a tour of how AI will change chatbots, governments, science, music, and health. Along the way, Sid and Debbie share a number of specific tools, tips, and trends relevant for AI practitioners everywhere.
Listen now wherever you get your podcasts (Apple Podcasts, Spotify, Overcast, etc.).
This is the fourth and final episode of our series on AI for Health. Thank you to our sponsor, Reach Digital Health, for making this series possible.
Connect with Africa Health Ventures
📰 Sign up for our newsletter to stay informed about deals, news, events, and opportunities related to healthcare ventures in Africa
👍 Follow us on LinkedIn
🎙️ Subscribe to this podcast to get notified when new episodes air
🪙 Nominate an African startup for seed funding
🤝🏻 Invest with us
Show Notes
In this interview, we cover:
(2m56s) - An introduction to MomConnect, South Africa’s maternal health chatbot.
(9m30s) - What is the future of MomConnect? Debbie looks ahead to how AI will allow chatbots to address the specific and personal needs of mothers across South Africa. This hyper-personalization will unlock behaviour change and precision health at scale.
(13m53s) - IDInsight has supported MomConnect to better understand what mothers need when they interact with MomConnect and guide them to the right information. They are using AI to infer meaning from words i.e. intent classification.
(17m07s) - Large Language Models (LLMs) like ChatGPT will allow the average citizen to better understand and tap into government programs and benefits. Sid cites the example of Indus Action’s work in this space in India.
(20m12s) - AIs are now pretty good at making music. We share a clip from a song generated by AI.
(22m07s) - Sid anticipates how AI will usher in a new era of qualitative science.
(25m00s) - “Every AI Application in the Social Sector Should Have Guardrails". See below for links to specific guardrail tools recommended by Sid.
(27m40s) - Sid unpacks the role of regulation to manage the risks of AI. He discusses the balance between regulation and acceptable risk.
(32m18s) - Sid shares a set of technical tools and resources for getting started with AI. Also linked below.
(32m58s) - A few shoutouts to other organizations doing interesting work in this space.
(33m49s) - The next 10 years according to Sid and Debbie.
Tools and References
NeMo Guardrails and Guardrails: Sid recommends that every AI application in the social sector should be using guardrails. Here are two open-source toolkits that programmers can add to their applications to implement such guardrails.
To get started quickly with AI applications, Sid recommends a suite of tools:
LangChain is a programming framework to make it easy to build LLM tools.
Llama Hub is a centralized repository and platform for integrating various tools and plugins specifically designed to enhance the functionality and performance of the LLaMA (Large Language Model Meta AI) series of models.
LammaIndex is a toolset designed for indexing, querying, and managing data to improve the efficiency and accuracy of search and retrieval operations in large-scale language models.
Suno AI is the tool Sid used to generate his song on AI for good. It’s pretty good.
Reach Digital Health is a South African nonprofit that scales digital health programmes to improve delivery of healthcare services.
IDInsight helps leaders combat poverty worldwide by designing, deploying and promoting evidence-generating tools, including data science and artificial intelligence.
Indus Action is an organization working with IDInsight to make government services more accessible and understandable to the average citizen with the use of AI.
Sid’s first shoutout goes to Jugalbandi, an open platform that integrates Generative AI and Indian language translation models to create conversational AI solutions. It supports over 50 Indian languages and aims to democratize access to information in various sectors such as government services, law, and healthcare through chatbots on platforms like WhatsApp. Jugalbandi is a part of the Bashini mission of the government of India.
Jacaranda Health: Sid gives a shoutout to Jacaranda Health for building a language model for Swahili in Kenya and releasing it open source. They also provide a maternal messaging service in Kenya called PROMPTS.
Dimagi: Sid also gives a shoutout to Dimagi for developing an open-source platform called Open Chat Studio to rapidly deploy LLM-based solutions for global health.
Share this post