ICTworks

The Gates Foundation recently announced the winners of their Grand Challenge for “Catalyzing Equitable Artificial Intelligence (AI) Use.” According to the foundation’s website, the goal of the funding is to “test a wide range of approaches to using AI to maximize and accelerate global health and development impact.”

At the African Digital Library Support Network (ADLSN), we are exploring how AI can help us in our mission to provide equitable access to African-created content. When I reviewed Gates Foundation’s release of their 48 funded projects I thought it would be useful to analyze them to see key areas of interest Global South.

Artificial Intelligence (AI) was the key trend. I hope my analysis will help to generate ideas for uses of AI across all domains. Here’s what I learned.

Domain

Given the Gates foundation’s focus on digital health, it is not surprising that 36 of the 48 projects are health related; the remainder are in the following domains: education (4), agriculture (3), financial (2), legal (1), environment (1) , and cross-discipline research support (1)

Geography

The majority of the Gates Foundation grants were awarded to projects in Africa (29). By country, the most grants were awarded in Brazil and South Africa, with six in each country; followed by five each in India and Kenya; four in Nigeria; three each in Ghana, Tanzania, and Uganda; two each in Malawi, Pakistan, Vietnam, and one each in Lebanon, Mali, Nepal, Rwanda, and Senegal.

Grant Projects

The Gates Foundation website includes brief project descriptions for each grant (learn more). From those grant application descriptions, I categorized the 48 grants into five broad groups. I have listed the groups below from the most to the least number of grant types awarded. Please note there is overlap in some of the groups and some grants can be in more than one group.

1. Improve Work Effectiveness

The largest grouping of projects is related to creating tools to support people in their work. The grant funds will be used to improve the quality and reach of patient care, to inform diagnosing in the medical and agricultural fields, and to equip frontline healthcare workers with diagnostics tools in areas with limited access to trained medical staff, primarily in rural areas.

The tool (an app or website) will allow end users to query via text or audio, while the platform leverages existing large language models (LLMs) and incorporates data, reports, and/or guidelines relevant to their local context.

Examples

• Access Afya in Kenya will integrate ChatGPT into their existing systems to “increase the scope, speed, and quality of responses to patients’ queries.”
• Munai Health in Brazil will create a conversational AI tool to enhance their existing platform to provide health workers with a more user-friendly method to access complex protocols to ensure adherence to antimicrobial therapy guidelines.
• Common across many grant projects is the use of AI to improve healthcare where there are few trained professionals. For example, the EHA Clinics Ltd. in Nigeria will integrate GPT-based tools into their existing THINKMD system to explore the possibility that minimally trained paraprofessionals, in low-resource settings, can take on some of the responsibilities of medically trained personnel.
• In an agricultural context, at the Sokoine University of Agriculture in Tanzania, they will develop a ChatGPT-powered Swahili chatbot to support farmers to help them detect crop diseases in time to take preventative action.

2. Provide Unmediated Information

Another group of grants will develop apps and/or websites to increase the availability of information to the general public, mainly for medical information, but also for legal and financial information.

In the medical field, the tools will provide quality, unmediated information to audiences who might not feel comfortable seeking such information in person (such as sexual, reproductive, or HIV information). The legal and financial projects centered mainly around simplifying complex information to make it more broadly accessible and comprehensible.

Examples

• The only grant that includes radio, Boresha Live in Tanzania hopes to broaden the reach of their message by integrating ChatGPT into community radio. ChatGPT will be trained on both general and local malaria information to produce “accurate malaria-related information that respects cultural norms, language preferences, and local challenges.”
• Myna Mahila Foundation in India is creating a chatbot (“Myna Bolo”) that will allow girls and women to ask questions about sexual and reproductive health in their own language and get responses that are tailored to their local context while respecting their privacy.
• Kwanele in South Africa will create a chatbot and mobile app to provide legal information on gender-based violence that helps to guide vulnerable groups through the complex judicial system.

3. Improve Work Efficiency

Another set of grants coalescences around improving work processes, such as improving operations or data management. The tools might conduct data analysis, facilitate data gathering, or streamline current manual processes, such as the creation of hospital exit summaries and the management of medical records.

Examples

• The THINK Tuberculosis and HIV Investigative Network (RF) NPC, South Africa and THINK International Denmark will create a toolkit that can be used broadly to perform data analysis and help harmonize fragmented datasets, initially focusing on tuberculosis (TB).
• A project in Pakistan at the Lahore University of Management Sciences seeks to build a voice-enabled conversational AI tool, Awaaz-e-Sehat, that allows for multiple language inputs and converts them into Urdu to populate electronic records.

4. Improve Education and Training

Four grants focus solely on improving educational results through AI tools. Three of the four grants focus exclusively on offering personalized education as a way to engage students and increase literacy. The fourth grant proposes to use AI to create new literature in local languages to inspire children to learn to read.

There were also a few education-related projects focused on creating tools for training medical professionals (and in one case volunteers). The proposed solutions are hoping to more effectively convey complex and up-to-date information in an accessible and personalized way.

Examples

• SOMANASI project in Kenya proposes to develop an AI application to provide personalized education in the form of a virtual tutor that “delivers tailored content, adaptive learning experiences and interactive guidance.”
• The Association Robots Mali seeks to use AI to create children’s literature in Bambara, which is widely spoken in Mali. There is a lack of children’s literature in Bambara (as the focus in schools remains on French materials).
• ARMANN in India will integrate LLM-powered tools into their existing learning applications to improve training for frontline healthcare workers.

5. Make Predictions

A smaller number of grants will work on projects related to making predictions (for example, for medical predictions of the impact of vaccines or predicting future pandemics). One grant hopes to impact and predict future policy development.

Examples

• Comzine Tech and Investments Ltd in Uganda will build a ChatGPT platform, DROMEDIC-AI, to predict future pandemics by monitoring zoonotic diseases. The DROMEDIC-AI platform, trained on large volumes of text and data, will be used by farmers to upload photos of sick animals and receive advice and also to generate risk assessments and maps of hotspots that will help health officials to monitor potential outbreaks.
• Child Health Research Foundation in Bangladesh will incorporate local data to predict the impact of introducing specific vaccines in Bangladesh.

By Rosalie Lack, Strategy and Engagement, African Digital Library Support Network (ADLSN)