Philanthropy needs to embrace AI and fast, experts say

Philanthropies are opening up to the idea of using artificial intelligence in their internal operations as well as in their grantmaking. But they will need to quicken the pace if they want to push the emerging technology toward inclusivity and equity for low- and middle-income countries, experts say. Philanthropy, like other sectors outside of technology, has been slow to adopt AI. However, it is now embracing AI language tools such as ChatGPT, a chatbot, as well as visual tools such as Beautiful.ai. The technology has the potential to change how philanthropy works, from how organizations review and write grant proposals to deciding which projects to fund. A few globally minded philanthropies are at the forefront of this movement. The Bill & Melinda Gates Foundation recently issued a call for proposals for projects that would promote equitable AI uses in health care and other areas that would benefit LMICs. Meanwhile, the Center for Effective Philanthropy, a U.S.-based philanthropy research and advisory organization, is publishing a series of blog posts to educate funders and nonprofits about AI. And grantmakers such as The Rockefeller Foundation are providing grants to researchers at the forefront of AI development. Philanthropy is making good progress on AI adoption but needs to step up its efforts, and soon, said Vilas Dhar, president and trustee of the Patrick J. McGovern Foundation. If the sector can fully engage with this technology as it emerges, it also has a better chance of helping shape it around priority issues such as promoting equity, he said. Like all new technologies, generative AI was developed by human programmers and therefore reflects human biases. If that isn’t addressed early, AI could widen existing equity gaps that the work of global development and philanthropy aims to close. “Philanthropy was so slow to adopt technology, period” --— Vilas Dhar, president and trustee, Patrick J. McGovern Foundation The window of opportunity to prevent that from happening is now, Dhar said. “I think philanthropy faces a critical choice about whether it will be slow adopters, or whether it will be early architects of an AI-enabled future,” he said. “This set of tools will transform economic relationships, social relationships, will transform power relationships.” Leveling the playing field The Boston-based Patrick J. McGovern Foundation is one of few organizations whose mission is entirely focused on advancing AI and other new technologies — which has not been philanthropy’s strong suit. It was started by the publisher of PC World Magazine and the “For Dummies” series of explanatory books. The McGovern Foundation works with organizations such as UNESCO and Physicians for Human Rights to help develop new mobile apps and COVID-19 data tracking that uses AI. “Philanthropy was so slow to adopt technology, period,” Dhar told Devex. But with AI, there are opportunities for funders and nonprofits to rectify that. For example, nonprofits can use generative AI tools to “create more captivating application materials,” he said. Within AI, large language models that understand and generate text can be used to draft fundraising letters, formulate grantmaking strategies, and perform research. Staff can use ChatGPT to write proposals or mid-term reports and Beautiful.ai to design “compelling presentations,” according to Dhar. Dall-E, an AI system that can create realistic images and art based on basic text descriptions, also can be used for multimedia storytelling, he said. Funders also can use chatbots to provide a real-time conversational interface between them and the grantees and to create opportunities for feedback, Dhar added. And on both the funder and nonprofit side, “AI can help level the playing field for under-staffed organizations, and decrease the risks of confusion, errors, or bureaucratic hurdles in the application process,” he said. Ethical concerns Like most technology developed by for-profits, AI has been optimized for scalability and profitability. And without clear regulations calling for developers to address biases or equity issues, it will be left up to consumers such as foundations and nonprofits to put pressure on innovators to prioritize equity in their work. McGovern is among those AI advocates pushing fellow funders and nonprofits to lead the charge on the development of equitable AI tools, especially those that can be accessed by the lowest-income countries and most marginalized communities. It is a member of the World Economic Forum’s Global AI Action Alliance, a coalition of philanthropic and technology organizations and global institutions seeking to accelerate AI adoption globally and across sectors and for causes such as the global transition to clean energy. Other organizations also are stepping in to provide funders and nonprofits with some basic knowledge to begin the process of integrating AI into their operations. The Center for Effective Philanthropy’s article series will show potential ways that AI can benefit philanthropy and how it is already being used by organizations. “I think we just feel AI has such far-reaching implications for all of us and that donors need to be engaging the questions of how their goals can be helped or hindered by AI,” CEP President Phil Buchanan said in an email to Devex. AI has the potential to impact any issue that a funder might be working on such as environment, economic development, or arts and culture, he said. “So, while it all feels a little overwhelming we need to be considering the potential implications, for good and ill, and how we can mitigate the latter and promote the former,” Buchanan said. Catching up A recent piece in CEP’s series written by researcher Sarah Di Troia dug into some of the ethical questions around using it, such as whether foundations should work with AI vendors whose tools aren’t tested for data biases. “Thinking about the intersection of equity and AI is a strength of the nonprofit and philanthropic sector — and it is time for our sector to get active,” wrote Di Troia, who is a senior strategic advisor of product innovation at Project Evident. “AI is quickly coming into the nonprofit sector, but we do not have the same 20–year runway the for-profit sector had to explore how AI could create value,” she said, noting that “only a few pioneering funders” such as Google.org and the Patrick J. McGovern Foundation were providing grants to AI-focused nonprofits. Project Evident is a Boston-based nonprofit that promotes an evidence-based approach to help nonprofits and other groups increase their impact. It receives funding from the Gates Foundation, the Chan Zuckerberg Initiative, MacKenzie Scott, and the Robert Wood Johnson Foundation, among others. Project Evident has been hosting AI 101 webinars for funders and nonprofits to encourage them to explore AI use. “AI can feel a bit amorphous. It can feel a bit scary for folks. But there are ways in which AI is actually powering the work that you’re already doing,” Di Troia said during a webinar last month. Among the tips Di Troia gives to funders and nonprofits is to encourage staff to start learning about AI. As a first step, they could experiment with using large language models to reduce time spent on repetitive tasks — which could bring efficiencies to things such as the grantmaking process. For example, smaller organizations that have online grant writing processes will soon be able to turn on AI grant writing functions similar to ChatGPT that understand and generate text in a human-like fashion instead of hiring a grant writer from a potentially small budget, she said. Foundation program officers tasked with reviewing grant proposals also will be able to use AI to help with their analyses. However, she also suggests funders and nonprofits make sure that the AI vendors they work with demonstrate a commitment to decreasing data biases to make sure that their AI tools don’t help create or reinforce discriminatory practices. “One of the reasons AI vendors don’t address equity is for-profit customers don’t ask about it,” Di Troia wrote in her piece for CEP. Equity and inclusion Major foundations such as the Gates Foundation and the Rockefeller Foundation also are pushing for philanthropy to use their dollars to push AI developers to prioritize equity and inclusion when creating tools. The Gates Foundation's recent request for proposals offered $100,000 grants for projects that would not only encourage equitable uses of AI but ensure that low- and middle-income communities could access and help develop the tools. And Rockefeller is among the foundations supporting groups leading global conversations about the ethical use of AI, such as the Distributed AI Research Institute. Last year, Rockefeller provided $200,000 in grant funding to DAIR, a nonprofit started by Timnit Gebru, an AI ethicist that was allegedly ousted from Google in 2020 because of her research on the “shortcomings” of language-based AI programs. It also gave $300,000 to Black in AI, another group co-founded by Gebru that aims to increase the number of Black people working in the AI field. The MacArthur Foundation and the Ford Foundation also have provided funding to DAIR. Rockefeller has had a longstanding interest in promoting AI use that only intensified when its president Raj Shah came on board in 2017, according to Kevin O’Neil, Rockefeller’s managing director of data and technology. The foundation wants to support people who are taking an equity-focused approach to building AI tools and the data sets that inform the application of those tools, O’Neil said. It’s important that philanthropy back those who are thinking about AI’s social impact and who might not otherwise receive the funding that they need, he said. “The world is building and is using AI tools, so if philanthropy, the public sector, the social sector, if we wait to answer all the questions, we’re going to be behind the curve,” he said. “The best thing we can do is both build the infrastructure, start using it, but also build in those safeguards, those norms. … That’s a great way to think about setting the course as opposed to responding to it,” he said.