The
Foundation
Layer
A philanthropic strategy for the AGI transition
by TYLER JOHN
Other Endeavors for the Common Good: Political spending and impact investing
I’ve argued that philanthropy has an irreplaceable role to play in AI security and has already led to outsized impacts on safety that are competitive with the impacts of AI companies and governments despite amounting to a tiny fraction of what companies and governments spend. But charitable giving is not the only way to use money for the common good, or even the only way to give it away.
Political Giving
One highly complementary approach to giving is political giving. For Americans with the common good at heart, political giving — including both donations to 501c4 (lobbying) organizations and hard dollar donations to political candidates and PACs — has the same motivation and priorities of charitable giving, but without the tax deduction. Just like charitable giving, political giving is outside of the incentive systems of markets and democracy, allowing it to focus on what is most important. The goals of political giving are ultimately the same as those of philanthropic giving: ensure good AI governance, for example by helping governments understand the AGI issue, encouraging them to support beneficial AI innovation in safety and defensive capabilities, and progressing nonproliferation efforts to prevent the spread and use of dangerous AI capabilities.
For American donors deploying small sums, hard dollar donations are often the most cost-effective use of marginal dollars, due to regulations on campaign financing. Because donors are capped in how much they can give directly to politicians, the market is highly distorted and non-competitive, and donors are massively underspending compared to what they would do by default. Instead of having to spend in a crowded competition with corporations investing enormous sums in lobbying and research, you’re only spending against individuals who can give no more than you can. As a result, hard dollar donations are extremely leveraged, and a highly effective use of small amounts of capital that sit outside of your philanthropic foundation or DAF.
There are a number of C4 organizations that have strong AI policy agendas, but which are extremely constrained by capital. Some of these organizations also coordinate hard dollar donors to help them give more effectively. You can find further resources in Appendix A.
Impact Investing
Almost all philanthropists are excited about the idea of impact investing, that is making for-profit investments in projects that have a beneficial social impact along with a return on investment — sometimes subsidizing projects that otherwise have too weak a market case to receive funding from traditional venture capital.
Typically, my advice to philanthropists is that you can’t have your cake and eat it too. Given the efficiency of the market, if something is profitable it will tend towards market saturation. This makes it more difficult for your dollars to have a marginal impact, since if you didn’t invest in an offering, a market competitor would quickly take your place. In general, it’s most efficient to keep investing for profit and deploying capital for the common good separate: when you are investing for profit you should go after the highest rate of return, whatever that is, and when you are deploying capital for the common good you should deploy it wherever your dollars go the furthest, no matter the return on investments.
However, there are a number of specific cases where impact investing can be a highly effective use of capital:
Market-scale solutions: Some problems, like technical work in AI, benefit from market-scale approaches that leverage capital to grow the size of the work to the size of the problem. Philanthropic projects, because they do not solve a problem in the market and do not make returns, cannot scale as quickly as for-profit projects.
Market subsidies: Sometimes a philanthropist is willing to take a loss and subsidize the market, making an investment in an area where investors seeking returns would not invest. This makes impact investing much more like charitable giving, because there is little market incentive for others to take on the project if you do not.
Faster spend-down: Nearly all major philanthropists struggle to spend down their philanthropic endowments faster than their rate of return. If you are struggling to find enough granting opportunities to meet your desired rate of spending, making larger VC checks can be an effective way to close the gap.
Beating the market: If you’re able to beat the market in a particular domain, perhaps because you have greater fundamentals knowledge from your understanding of the AI safety and governance ecosystem, you can make substantial returns and have a substantial impact by simply being ahead of the curve.
Controlling stakes: Microsoft and Google clearly have an enormous influence on the future of AI. By having a major stake in a highly influential company (which ordinarily requires eye-watering sums), you can steer the company towards more impactful research, better corporate governance, and more thoughtful policy advocacy and communications.
A year ago, when philanthropists asked me for impact investing advice, there was literally nothing I could recommend. Perhaps you could bet on the AI company you think is the safest, but this is extremely poorly leveraged given how many assets these companies already have to work with. But today there is a rich ecosystem of for-profit AI technical safety and assurance companies in critical areas like interpretability, model evaluations, and cybersecurity. For details on how to get involved with these companies, see Appendix A.
← Previous
Overview
Next →
Why The Problem Remains Neglected
The Foundation Layer is a project by Tyler John, with generous support from
Effective Institutions Project and Juniper Ventures.
