Catalyzing Equitable Artificial Intelligence (AI) Use

Background

The power of science and innovation can improve global health and development outcomes and dramatically reduce global inequity. Harnessing the potential of Artificial Intelligence (AI) can improve the lives and wellbeing of women, children, and vulnerable communities everywhere.

As AI technology continues to swiftly evolve and advance, the global community must move with urgency to ensure low- and middle-income countries^[i] (LMICs) are included in the co-creation process. AI solutions targeting LMICs must be locally driven and owned and relevant to each person's needs. Local communities must be given the opportunity to provide their own perspective and cultural context so that they can decide on both 1) their own thresholds for safe usage and 2) the overall utility of AI within their own lives.

Additionally, LMICs must have equitable access to AI tools and approaches. Responsible global use of AI entails a safe, equitable, transparent, reliable, and beneficial process that is adhered to with a high level of accountability. As the world rapidly moves to seize AI's opportunity, it is imperative to monitor and mitigate the safety, ethical, equity, and reliability dimensions of AI deployment. This will allow the enormous resilience, creativity, and commitment of researchers, scientists, and policymakers to capture the full capability of AI for lasting good.

The Bill & Melinda Gates Foundation and Grand Challenges (GC) Partners (including GC South Africa and GC Brazil, with others to be confirmed) have jointly discussed the need for an equitable and responsible approach to the use of AI and specifically Large Language Models (LLMs) in LMICs. This initial call by the Bill & Melinda Gates Foundation represents an early step towards identifying, nurturing, and catalyzing the creativity, energy, and skills that researchers, implementers, governments, and technical partners have demonstrated in solving specific challenges in their countries and regions through LLMs. We are optimistic that this will lead to more investment from the Gates Foundation, GC partners, and other funders, to ensure that this potentially transformative technology improves the lives and conditions of the most vulnerable communities around the world.

The Challenge

This RFP seeks innovative and safe approaches to the use of Large Language Models (ChatGPT-4, or other credible sources with equivalent capability) to build an evidence base across the Bill & Melinda Gates Foundation Divisions of Global Health, Gender Equality, Global Growth and Opportunity (including Global Education), and Global Development contexts in LMICs.

We are looking for projects that can be executed over a 3-month period and will achieve at least one of the following objectives:

Demonstrate how the tool can be used to support decision making (in terms of speed, accuracy, inclusivity, quality) in the following domains.
- Accelerate the pathway from evidence to policy development.
- Accelerate the pathway from policy to implementation.
- Enhance the quality and timeliness of focused communications.
- Improve the quality of resource allocation exercises.
- Decision support for front line workers.
Achieve a better understanding of the opportunities and challenges related to ChatGPT-4, or other credible sources with equivalent capabilities, use in LMICs and how they can be effectively mitigated.
Conduct observational research that highlights end-user specific applications of ChatGPT-4, or other credible sources with equivalent capabilities, amongst vulnerable populations in LMICs.

Ideally the proposal will also:

Work on a specific problem that has been identified as a priority in a country or region.
Contribute to the body of evidence related to AI use in LMICs across health, gender, financial inclusion, agriculture, and education.
Foster innovation with AI in the interest of supporting decision makers and affected communities.
Enable an improved understanding of the potential and risks of AI for LMICs.
Place a strong emphasis on systematically observing, validating, and documenting the process and outcomes for each proposal.
Conform with global principles of AI use i) to do no harm ii) to leverage technology against the toughest/most relevant problem iii) ensures that projects are led from LMICs even though there may be a high-income country (HIC) partner iv) local voices and ideas are captured throughout the process, and v) there is a dissemination plan of the outcomes of the projects.

Priority will be given to proposals that have:

an explicit request for an AI supported project,
those projects that have already completed a pilot before this Grand Challenges call, and/or,
projects that have lessons/tools that can be transferred to other use cases / situations/ contexts with minimal change.

Funding Level:

The funding level is a grant of up to USD $100,000.00, for each project, provided to the organization, with a term of up to 3 months.^[ii] Application budgets should be commensurate with the scope of work proposed. The budget envelope for this Grand Challenge request for proposals is up to $3,000,000.

What We Are Looking For:

Proposals are led by investigators in LMICs.
Proposals that demonstrate clear applications of ChatGPT-4, or other credible sources with equivalent capabilities, to address a specific challenge in LMICs related to the SDGs.
Proposals that demonstrate the grantee has a firm grasp of the application, has done some of the stakeholder mapping, and has an engagement plan to ensure the proposal is successful.
Proposals that include engagement with affected/supported populations, decision makers.
Proposals that show the potential for AI to be transformative for application in the Bill & Melinda Gates Foundation Divisions of Global Health, Gender Equality, Global Growth and Opportunity (including Global Education), and Global Development.
Proposals that present a high leverage and scalability opportunity.
Proposals that can easily be adapted or modified either across i) geographies or ii) fields.
Proposals that outline a clear, feasible and reproducible methodology.
Proposals that focus on identifying risks to responsible use and develop ways to mitigate those risks.
Proposals that have timely access to data, decision-maker time and interest in using AI
Proposals that articulate how the project will lead to impact in the near-term and how those benefits will be sustained past the lifetime of the project.
Proposals that demonstrate engagement with local and regional decision makers.
Proposals that are driven by a shared commitment to open science, data sharing, and building collaboration and analysis infrastructure to enable discoveries that will benefit people everywhere.

Note: Global partners may be included. However, priority will be given to proposals that demonstrate at least 80% of the funding is going to LMIC institutions and where the investigator is a part of an LMIC institution.

We will not consider funding for proposals that:

Do not explicitly use/reference the use of ChatGPT-4, or other credible sources with equivalent capabilities, in their project execution.
Ignore the challenges and context constraints LMICs present.
Do not have timely access to necessary data, decision-maker time, commitment, and interest (a commendation/request letter will be advantageous).
Do not demonstrate that the majority of the work proposed will be undertaken by investigators and teams in LMICs.
Have a weak methodology.
Have no or weak validation plan.
Do not provide any thought towards scale up and sustainability.
Do not engage local and/or regional decision-makers.

^[i] See World Bank – Low Income Countries for definitions.

^[ii] Subject to the eligibility requirements in the Rules & Guidelines.