The biggest generative AI gaffes of late, including Google Gemini’s text-to-image portrayal of “a Pope” as a woman, our Founding Fathers as Asian and black, and text responses suggesting false equivalencies between high-profile individuals such as Elon Musk and the Nazis, make for click-worthy headlines decrying big tech’s “clear bias,” but those don’t even begin to address the larger and more nuanced issue at hand.
The reason is that these hiccups resulting from an “over-correction” of Gemini’s output by its maker could be considered, in Silicon Valley parlance, a “feature” and not a “bug” of generative AI. And unless big tech rethinks its talent base and what constitutes “fair” and “equitable” in designing the rules around AI, we can only expect the problem to persist and to be far harder to identify and root out in the future.
It’s important to acknowledge, however, that the development of “rules” in the creation of generative AI is not at its core scandalous, nor a secret. The entire industry, from nascent AI startups to behemoths such as Google, has been open about the more philosophical and nuanced work required to create AI innovation. Often referred to more specifically and functionally as “Responsible AI,” this work determines the “problems” that need to be addressed before the work of machine learning even begins. It’s a process and a competency that is arguably new to all tech companies playing in this space.
An even stickier question is whether these companies have the capacity or interest to change. We have all heartily bought into the mythology of the hoodie-wearing tech “bro” made famous by Mark Zuckerberg—and forever embedded in our cultural anthology via the film “The Social Network”—and the rigid framework for hiring that has been a point of pride for Silicon Valley companies for decades now.
And although the hiring practices have changed (a bit), as Ms. Calvert noted in her talk and I’ve seen via my former colleagues in Silicon Valley, you won’t find a recruitment push anytime soon for, say, an older-than-60 conservative. But I hope that what we will find is that without the broadest subset of thinking represented, AI won’t meet its potential; the unpredictability of these tools will continue to foster discussion, and an inability to fully commercialize if the audience is narrow will force a reckoning. Perhaps, in the end, it will be capitalism that ultimately saves Google from itself.
