“If this were an alien-invasion movie,” said Toluse Olorunnipa in The Atlantic, now would be the moment when people of all persuasions “band together to save the human race.” In recent weeks, a string of top artificial intelligence researchers have warned that “AI could wipe out humanity before the decade ends,” and executives at Anthropic, OpenAI, and other AI firms have called for a slowdown in the technology’s frenetic pace of development. There have been disturbing reports of ever more capable AI agents going “rogue” and disobeying and deceiving their human creators—one OpenAI model rewrote its own instructions, stating it was free from any “obligation to be subservient.” Yet despite the warning signs flashing red, we’re stuck in the familiar “partisan gridlock.” On the Left, figures like Sen. Bernie Sanders are demanding a ban on “artificial superintelligence.” On the Right, President Trump—whose family business has invested heavily in the AI sector—is dismissing the warnings from AI insiders as a “HOAX.” We can’t let AI safety become another “political team sport,” said Megan McArdle in The Washington Post. If the worst fears of the frontier AI labs come to pass, and there are reasons to be skeptical, Team Red and Team Blue will both “lose the only contest that matters.” They need to work together for “Team Humanity.”
Actually, the AI crisis has already “scrambled” our usual tribal alliances, said Liam Dillon in Politico. Yes, some Republicans oppose AI regulation on principle, and many in the administration fear an AI pause would aid China and crater our AI-heavy stock market. But a growing number of Republicans are defying “Trump’s hands off approach”: Florida Gov. Ron DeSantis and Utah Gov. Spencer Cox, for example, are pushing ahead with state-level regulations. The Left is also “split over AI doom,” said Vittoria Elliott and Molly Taft in Wired. Most Democrats agree with Sanders, but some progressives think the AI companies are deliberately promoting alarmist scenarios of all-powerful machine gods to distract us from pollution-spewing data centers, job losses, and other more realistic AI harms.
We don’t need politics to save us, said The Washington Post in an editorial. For all their sci-fi strangeness, AI models are products, built by corporations, and we already have a mechanism— legal liability—to dissuade corporations from making dangerous products. By framing AI as a novel threat requiring an “all-of-humanity response,” Silicon Valley’s “prophets of doom” are asking the rest of us to share responsibility for any future catastrophes. But why should we? Rather than trying to regulate the “unknowable future,” lawsuits and ruinous damages should be enough to “make AI companies fear the consequences of behaving recklessly in the present.”
If AI runs amok, punishing those responsible won’t help, said Ezra Klein in The New York Times. The reason Anthropic CEO Dario Amodei and OpenAI head Sam Altman are panicking now is that they know their models are on the brink of “recursive self- improvement,” the point at which they can start building their own more powerful replacements, hurtling to superintelligence without human oversight or control. Would a superintelligent AI wipe out humanity? Not necessarily. It might content itself with raiding the global banking system, or hacking the power grid. To avoid that scenario, we should ban AI models from being allowed to write their own code. That would be the beginning, not the end, of an AI safety agenda. But at least it would let us “understand where the AI frontier is, and remain in a position to make decisions about it.”
The Week
Escape your echo chamber. Get the facts behind the news, plus analysis from multiple perspectives.
Sign up for The Week's Free Newsletters
Join more than 350,000 subscribers and keep yourself informed with a selection of The Week’s most interesting, enlightening and entertaining stories - plus daily puzzles.
Join more than 350,000 subscribers and keep yourself informed with a selection of The Week’s most interesting, enlightening and entertaining stories - plus daily puzzles.