Wednesday, March 15, 2023

SVB, GPT, and the social sector

Photo by Bram Van Oost on Unsplash 

If you’ve been reading the news you know that Silicon Valley Bank (SVB), a bank that heavily caters to VC firms and start ups, collapsed and its depositors are being saved by the US Treasury. You know that hearings are being called for in Congress and the same old battle lines between the wealthy (people and institutions) and everyone else have been re-animated. And you can infer that there was (and is) a whole lot of backroom-ing going on.

You also know that SVB had lots of money in accounts held by nonprofit organizations, including affordable housing organizations. 

You also know that Open AI, the once not-for-profit-now-for-profit research organization has released GPT 4, a large language model (LLM) update to its previously (as in three months ago) released GPT 3. You’ve heard about generative AI, read stories about how “nasty,” “smutty” or “just weird” the outputs of the GPT models are, and you may have even “played” with or worked with these models. On mastodon I found a thread of nonprofit staff sharing stories of how they’re using ChatGPT to expedite the funder-driven time-suck of cutting the 1000 word description of your programs and their world changing effects required by Funder A into a 300 word description for Funder B. 

And you’ve probably seen, perhaps read, maybe skimmed the numerous articles and abundant research on how the LLMs are biased and the outputs are “hallucinations,’ (yes that’s what they’re called). As for SVB, you may have seen stories or tweets or blog posts about how the collapse of SVB will lead to an immediate funding disaster for all Bay Area nonprofits.

I want to posit two things. First, jumping to insights or conclusions right now about the effects of either the bank collapse or generative AI makes for good Twitter (if there is any such thing anymore), but isn’t reality. It’s punditry, lobbying, or sales. Second, think about the intersections between these things - emerging tech systems, corporate hype, cost of living, need for and role of nonprofits, risk management in banking, risk denial in corporations, risk and responsibility of governments, philanthropic product choices by wealthy individuals (DAFs, LLCs, private foundations, community foundations) and, finally, the overlap between these categories in terms of actual number of people involved. 

It’s too soon to know how these things will play out at a sector level. Those on the outside of SVB and/or OpenAI don’t know as much as we think we know. We don’t know all the ways they intersect. The best anyone can be doing right now is 1) finding out if they have exposure to SVB or Credit Suisse, either directly or through their funders (true for startups and nonprofits and mitigate appropriately at the organizational level; and check on your own bank, given potential for ripple effects of individual bank problems; 2) Put on your hype-goggles, convene your nonprofit’s data governance review committee (What? You don’t have one?) and start thinking now about who generative AI helps, what it does well and where it is dangerous, if and how it aligns with the mission of your organization (The mission - not the development or marketing departments' metrics, but the actual mission), where (within what software you use) are algorithms already at work, and what data (on whom) you’d be feeding to a third-party corporation (such as OpenAI) if you start using it and what that means for your constituents. 

These two things - a bank collapse and new technology - ARE likely to have BIG societal impacts. But understanding them will take time. And their impacts won't unfold along "straight lines" from A to B. There will be all kinds of additional "developments," intersections and interactions between impacts, and mitigations and responses. Don't fall for the quick analysis - it’s all operating on incomplete information.*

Just like the weather in California, judging from the winter we’ve had, forecasters (armed with actual meteorological and longitudinal data) are noting that we’re in for a long, strange Spring. That’s about all we can guess is coming from these two recent events. Strange times ahead. Keep your goggles on.

 

 *Speaking of incomplete information, Time Magazine is running a story describing how some of the biggest names in Effective Altruism knew about the financial shenanigans of their most famous, duplicitous member, Sam Bankman-Fried. Yet, they were still "shocked and dismayed" when his crypto-empire turned out to be built on fraud.

No comments: