“This technology will be the center of future crises, future financial crises.”
Webs We Weave
Looks like SEC Chair Gary Gensler has been harboring some serious — as in, destroy-the-economy-level serious — AI apprehensions for some time now.
As Axios reports, back in 2020, Gensler — still a professor at MIT at the time — penned a paper arguing that embedding an array of too-similar deep learning programs into our economic structures could stand to undermine those systems to the tune of a crisis-level crash.
Coauthored alongside MIT engineer and computer scientist Lily Bailey, the paper contends that the “broad adoption” of AI could push economic systems to the point of deeply fragile uniformity and interconnectedness, which would leave our financial systems vulnerable to, say, a disastrous mass sell-off triggered by machine predictions. Our “existing financial sector regulatory regimes,” meanwhile, designed to manage now-outdated human-speed fintech and analytics, wouldn’t be able to keep up, leading to similarly consequential regulatory gaps.
As AI “moves to a mature stage of broad adoption,” the authors wrote, “it may lead to financial system fragility and economy-wide risks.” Big gulp.
Data Hegemony
The paper is expansive, and among other reasons to fear model hegemony, Gensler and Bailey warn of the risk of data “crowding” and “herding,” or the reality that “models built on the same datasets are likely to generate highly correlated predictions that proceed in lockstep.”
So, basically: if you train two models on the same data, those models can be expected to draw the same or similar conclusions. And if too many advanced models come to the same conclusions, at the same time — well, it probably wouldn’t be good.
The paper’s warnings, however, go beyond basic model predictions. It also touches on machine bias, and the reality that AI systems — even with the best existing guardrails — are embedded with human social prejudices. If our financial systems were to be rebuilt on bias-ridden predictive programs, the reasoning goes, it could be disastrous for marginalized groups.
It’s all concerning, but if there’s any silver lining, it’s good to see that the person who was considering these wide-ranging impacts of AI back in 2020 is the guy in charge of America’s biggest financial regulator now. And between calls for Wall Street crackdowns and extremely pointed speeches, he’s at least paying lip service to AI’s potential financial harms.
“This technology will be the center of future crises, future financial crises,” Gensler recently told The New York Times. “It has to do with this powerful set of economics around scale and networks.”
But of course, whether any meaningful regulatory action follows still remains to be seen.
More on the economics of AI: AI Is Starting to Look Like the Dot Com Bubble
Share This Article