The Massive DDoS Edition

Editor's Note: 

If you were hoping that the faiV would be a pleasant distraction from the overwhelming flow of AI-related content, sorry. For the record, the faiV had “ai” built in before Anthropic existed. And a housekeeping note: we’ve finally jumped ship from Mailchimp to NYU’s mailing platform—I guess if you’re reading this it’s been a seamless transition and if you’re not, well, you’re not. Feedback on the faiV’s content and format and delivery methods is always welcome, just drop me an email. 

- Tim Ogden

1. AI AND THE ECONOMY

In June, the Annual Bank Conference on Development Economics had Susan Athey in one of the keynote slots, talking about the potential effects of AI on global labor markets and how to study it. (There’ll be more on other topics at ABCDE, including Small Firm Diaries, in the next faiV). There’s plenty of thinking and writing on the topic, of course. Specifically from a development perspective, here is a view of AI’s effects on global inequality through the lens of the Heckscher-Ohlin model, considering the possibility of wage gaps narrowing while income gaps widen. Here’s a different take that focuses on the potential lag between when growth effects kick in in high-income countries versus when they arrive in lower-income countries. But there’s a lot more that’s part of that project including interactive models, so check out the links here. 

But my big take-away from listening to Susan’s talk was not about macro-growth, inequality, or even labor markets. It’s my AI “doomer” scenario: AI crashes the global economy via what is in essence a massive distributed denial of service attack on basically every part of our system. Put another way, our existing systems (from markets, to regulation, to financial services, to everything) are built for the scale of inputs they had pre-AI; inputs are increasing far faster than the ability to change those systems, because that is not just a question of deploying the same AI tools, but of human systems and human adaptability.

There’s an O-Ring Theory flavor to this (stay till the last item to see an even more unexpected application of O-Ring Theory); see here and here. 

But to take it from the realm of theory to examples of what I’m terrified about, consider what happens when everyone has a superintelligent lawyer. Then think about how much of the economy ultimately rests on inefficient markets and inattentive consumers. The credit card economy in the United States fundamentally rests on shifting costs between market participants in ways that are hard to track for individuals. In fact, the entire financial system in most market economies builds in all sorts of reallocations and misallocations. I’m not making a value judgment on what is good or better, just noting that the profit margins of huge parts of the economy will likely shift very suddenly. 

Pile on top of that an incredibly rapid erosion of trust, which also remains enough of a pillar of the functioning economy that millions of people have been deluded into throwing away trillions of dollars attempting to make transactions “trustless.” Now that it’s difficult to know if the person on the other end of the phone call or Zoom meeting is actually your counterparty (or relative) or an AI-generated clone, or if the expense report is real, or if the person you are interviewing is real, or you have to spend a substantial chunk of your time making sure your elderly relatives don’t get scammed, or you have to take them into your home after they have lost all their savings… Those are huge frictions being introduced to the basic functioning of economies. Who is prepared for that?

2. CYBERSECUITY

AI-enabled fraud is a big part of the erosion of trust. I’m going to assume that you know about the Hugging Face incident, the accidental hacking of a fitness center in Australia, and the many other scary stories out there.

Taking a step back, it’s time for some Game Theory via Josh Gans: the reason I’m so worried about the denial of service scenario is that attackers have to be far less careful than defenders. Attackers tolerate 99.99% failure of attacks. Defenders will have to choose between a) massively outspending attackers, and b) rates of false positives dramatically higher than current practice. I think they’ll choose b) simply out of necessity, and then friction multiplies. 

We had a faiVLive on the specifics of the emerging cybersecurity threats to the global financial inclusion community. You can watch the recording here, or see the summary here. Since then I’ve only become more concerned, and with the recent incidents (it’s the summer of hacks!), I think we have to assume that many parts of the financial inclusion ecosystem have already been compromised in some way (e.g., that some of the “credit builder” loans made by US CDFIs are to malicious AI agents building an identity to enable other types of fraud and attacks).

3. MICROFINANCE

Perhaps it’s an illustration of the gap between the pace of AI-driven change and human capacity to adapt that we seem to be revisiting the impact of microfinance debates all over again. 

First there was a ridiculous WSJ story, and an even more ridiculous WSJ op-ed (really, how does the newspaper of capitalism come down on the side of “credit is an overrated fad”?). Coincidentally, as far as I know, CGAP published a new paper that same day on the topic of the impact of access to microcredit, “Opening the Black Box.” But of course one must respond to the WSJ, so there are posts from Sophie Sirtaine and Buhle Goslar, and from Shahid Khandker (if irony wasn’t already dead…). And VoxDev has a new short video with Simon Quinn and Muhammad Meki that’s worth watching. Oh, and I reviewed a couple of books on the topic for SSIR here.

4. PHILANTHROPY

Anyone with a passing familiarity with the financial inclusion space knows how central the Gates Foundation was to progress over the last 20 years, and that the financial inclusion program is winding down (NB: FAI was founded with a grant from Gates, and we remain a grantee today). I expect a non-trivial increase in the bounce rate for this edition of the faiV as many people in the sector are moving jobs as a knock-on effect of USAID cuts, other countries’ aid cuts and the wind down.

The massive question mark about future funding of financial inclusion and global health got bigger with the news that Warren Buffett further revised his plans for giving away his $150B+ fortune: no more gifts to the Gates Foundation, all will go to his family’s foundations. I’m quoted in a couple of Chronicle of Philanthropy stories about the news in ways that I think miss the point I was trying to make: there should be less attention on the “drama” and more attention on the question of how $150 billion is going to be given away by institutions that have no track record of handling anything in the same universe as that amount of money.

You may also be aware that there are expectations of much more philanthropic money coming into the global development and particularly global health spaces in the coming years, now being routinely referred to as the “Third Great Wave” based on this post, because of wealth driven by AI companies. GiveWell (NB: As of this writing, I’m chair of GiveWell’s board; I’ll be stepping down as chair next week, but remaining on the board), with funding from Coefficient Giving, is increasing its money moved this year by more than 2X to roughly $1 billion and looking to be ready to scale further. Coefficient has also launched Lightcone Commons and is participating in other efforts to influence that wave toward highly effective philanthropy. There is also Renaissance Philanthropy; and Stripe funding things like Works In Progress magazine. Hopefully you’ve seen In Development magazine by now as well. I will confess that although I’m at least adjacent to many of these things, I don’t understand how they all fit together and overlap (or don’t; could it be “participatory ambiguity”?). Despite that, they are helping me hold on to some hope.

5. CELL BIOLOGY = DEVELOPMENT ECONOMICS

Those of you who have been around for a while know that in addition to pretending to be an economist, I also occasionally have to pretend to be a cell biologist, because my oldest son has an ultra-rare genetic syndrome. I spent last week at a cell biology conference, being astounded by the conceptual overlaps between the two. Our bodies bear a strong resemblance to markets made up of independent actors—and primary cilia are the invisible hand!—all trying to optimize in a dynamic system leading to equilibrium homeostasis. 

As I was listening to a talk about the “bowtie model” of understanding cilia function and dysfunction, I thought, “Why are they talking about an application of O-Ring Theory?”  As papers using worm, fish and mouse models were presented, you could practically hear someone whispering, “All models are wrong, some are useful.” (OK, that was me). 

Why bring it up here? Mainly because I think the two disciplines can learn a lot by talking to each other about conceptually similar challenges (e.g., external validity, scale, identifying causal effects in complex systems, avoiding selection bias), so if you’re a development economist, amble over to the cell biology department and strike up a conversation. But more specifically, I’m hoping to have a summit of interested parties on both sides to talk about how the tools of causal inference in economics can help untangle some cell biology and genetics puzzles. So if that in particular interests you, get in touch! 

Image of the Week

Poetry Foundation

---

The faiV is written by Timothy Ogden, Jonathan Morduch, and Laura Freschi, and produced by the Financial Access Initiative at NYU's Wagner Graduate School of Public Service.
 Email: fai-wagner@nyu.edu
To translate the faiV into another language, view this email in your browser and click on translate at the top right.