Only Congress can (and must) tap the brakes on AI
You could almost hear palms smacking onto foreheads all over the techier corners of the internet last week, after a Google artificial intelligence program began generating pictures of Black founding fathers, a female Pope and other notions that would exist only in the most fringeprogressive alternate reality.
This was the unintended result of a legitimate but rushed attempt by Google to program out AI’s disturbing tendency toward racism and misogyny.
The episode illustrates just one of the many unforeseen challenges ahead regarding a technology that holds more promise — and more potential for societal havoc — than any since the creation of the internet itself.
Yet all the incentives for the companies developing the tech — Google, Meta, Microsoft, Amazon and a galaxy of less-familiar names — are toward rolling out their programs as quickly as possible in what amounts to an AI arms race.
The implications are too far-reaching to leave to the profit-driven feeding frenzy that is the tech industry. Congress must stop dithering on AI and set up a regulatory structure as soon as possible to govern its development and use.
The positive practical applications of AI are already astonishing. Medical diagnostics, drug discovery, manufacturing robotics, educational materials, cybersecurity, transportation advances, retail efficiencies — these are only a few of the many real-world endeavors where some form of AI is being regularly utilized to carry out complex processes faster and better than people alone can.
The darker side of the technology has been demonstrated as well, though, particularly in the area of misinformation and “deepfakes.” One of these cloned President Joe Biden’s voice in a robocall urging voters to sit out the New Hampshire primary. Another produced a vulgar video involving the president and his granddaughter. Yet another shows Hillary Clinton endorsing Republican Florida Gov. Ron DeSantis for president.
None of this actually happened, but all of it looked and sounded real.
Distrust in institutional norms is among the biggest threats to American democracy today, and malignant use of AI technology could worsen that situation in almost unfathomable ways. And that’s before even considering how identity thieves, con artists and even terrorists could abuse this technology. The Biden administration has nibbled at the edges of the issue with an executive order offering largely nonbinding guidelines for development and use of AI. Congress is talking — and talking and talking — on the issue. But it hasn’t offered a coherent legislative strategy.
If the federal government doesn’t slow this down for the sake of societal safety, profit-driven competing entrepreneurs have already shown they aren’t going to tap the brakes on their own. Last week’s Google debacle is instructive. Bloomberg said it best in a recent headline: “Google’s AI isn’t too woke. It’s too rushed.”
The same could be said of AI technology more broadly. And it’s a failing that could ultimately cause far bigger problems than just embarrassingly ahistorical imagery.