In 1988, IBM invested $10 million developing a project – Deep Thought – to create a rudimentary artificial intelligence program that could eventually beat a grand master at chess. The original program could calculate 720,000 moves per second. In 1993, the project’s name was changed to Deep Blue and on February 17, 1996 world chess champion Garry Kasparov played against Deep Blue and Kasparov won four out of six matches and took home $400,000 in prize money from IBM. One year later, Deep Blue successfully beat Kasparov and became the first computer system to beat a human world champion in a standard chess match. In a press conference afterwards Kasparov insisted that he was defending the honor of mankind against the world’s best chess computer. Today, it is not just the honor of mankind that seems at risk but, for some, a greater loss as artificial intelligence becomes far more sophisticated and pervasive.

On the other hand, are there opportunities for philanthropy in the growth of artificial intelligence? What does AI do as well as or better than humans? Does its contribution need to be perfect or simply good enough to make AI valuable? Is it only good for massive amounts of number crunching?

Schwab Charitable saw a 20% uptick in dollars granted to charity in 2021 compared with a year earlier. Grants totaled a record $4.4 billion last year, with donors supporting 114,000 charities through 945,000 grants. That is just one fund. Imagine how many grant applications are sorted  by thousands of foundations every year. AI is fully capable of taking 10,000 applications down to a short list of 100 by using available data from numerous sources while also eliminating some of the inherent bias of human screeners. A foundation principal told me that his foundation spends more time saying no to thousands of unsuitable requests instead of saying yes to a few that fit.

With the appropriate criteria AI could be proactive and used to help the donor search for and find organizations that would be candidates for grants. How does a typical program officer look for those organizations now? They call their peers or send out a notice that the foundation is looking for organizations doing a particular kind of work. They might send out formal requests for proposals and wait for organizations to respond. Unfortunately, the request is seen by a very limited number of organizations and many candidates (sometimes the best qualified) never see the request. As well, it takes time for proposals to be written and submitted. The donors find themselves having to, again, sift through more responses than expected or they are prepared to sort efficiently.

For instance, when MacKenzie Scott gave away $6 billion to 500 organizations in 2020 her team of advisors sought suggestions and perspective from hundreds of field experts, other funders, and non-profit leaders and volunteers with decades of experience. “We leveraged this collective knowledge base in a collaboration that included hundreds of emails and phone interviews, and thousands of pages of data analysis on community needs, program outcomes, and each non-profit’s capacity to absorb and make effective use of funding. We looked at 6,490 organizations and undertook deeper research into 822.” That is an army of human resources required to make those 500 grants. This is after she had already defined her particular areas of interest.

We all know from experience that fund-raising professionals have access to sophisticated tools for identifying potential foundations and donors. The knowledge available for seeking grants grows more advanced every year. However, the collective knowledge of foundations is still far behind the development professionals. They know what we have but we don’t know what we know. We don’t know what foundations have done the best work in a particular area or those who have specialized in a field. We don’t know those who are doing work in which we have an interest and more often than we like our funding only duplicates the funding of others without adding any other value. AI has the ability when used appropriately to create shared knowledge and experience that enables us to learn from each other more broadly than our limited networks of relationships. Machine learning can not only map non-profit organizations active in our areas of interest but also identify other funders working there. Funders could avoid the “winner takes all” syndrome where funders tend to back the same projects and organizations because they have inadequate knowledge about other options.

Rhodri Davies, the expert in residence at the Pears Foundation in London, has coined the term “philgorithm” as a way of identifying those processes that would lend themselves most readily to AI. He admits that it has obvious limitations and the field is still in its infancy so the most likely use of machine learning will be mostly in augmenting human decision making by data-driven algorithmic tools. Still, while the best philanthropy is not a game we should not overlook an opportunity to improve the craft.