Americans were already wary of artificial intelligence. Then came this past week, when tech industry titans proclaimed that society could shortly collapse.
The dystopian possibilities, as enumerated by Dario Amodei, the chief executive of Anthropic, include cyberattacks and bioterrorism, a ravaged economy, and A.I. systems slipping beyond human control.
But policy experts say the science fiction visions evoked by business leaders could distract from more nuanced conversations about how to govern technology companies — and what responsibility their human leaders owe the public.
“A.I. destroying humanity is the worst possible way to kick off a conversation about the public policy of A.I.,” said Milton Mueller, a professor of public policy at the Georgia Institute of Technology who has for decades focused on how to govern the internet. “It completely diverts attention away from human agency. It pretends that the machine itself is going to cause the problems.”
The top executives of A.I. companies, which have been locked in fierce competition, have lobbied for years for rules that they say would make their technologies safer and that their critics say could cement their position as industry leaders. In recent days, these executives have argued that the technology’s rapid advancement — exemplified by a high-profile mishap when so-called A.I. agents veered out of control this summer — means that competition should slow while lawmakers pass regulations.
A week ago, an Anthropic researcher, Jacob Coxon, said he had resigned over concerns that companies were ignoring safeguards as they sprinted to build “superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”
Then, on Saturday, Dr. Amodei called for slowing A.I. development in a 3,800-word essay. Other tech executives, including Sam Altman, the chief executive of OpenAI; Elon Musk, of SpaceX; and Demis Hassabis, of Google DeepMind, quickly wrote that they agreed.
Washington, after a long torpor, is responding. The top Republican and Democrat in the House on Sunday agreed broadly that Congress should act. President Trump on Monday dismissed calls to rein in A.I., but an unlikely group is set to discuss a “pro-human future with AI” on Tuesday. Speakers include both Bernie Sanders, the progressive independent senator from Vermont, and Steve Bannon, a MAGA podcaster and former top aide to President Trump.
Some Americans need no wake-up call. Ivy Goode, 65, a tourist in New York from Maryland, said the technology is already terrifying with its ability to wipe out jobs and enable fraud and theft — and that was before the technologists raised the specter of mass death by robot.
“Only God himself can do the human extinction because he created human beings,” said Ms. Goode. “But they’ve come in and just destroyed so much of what the creator created. It’s really scary, what could happen.”
The apocalyptic visions offered by industry leaders — what could happen if A.I. is misused and exploited by humans, or if it derails itself and acts in service of some other mission entirely — are vague and broad.
That leaves room for speculation, said David Evan Harris, who teaches courses on A.I. ethics at the University of California, Berkeley. Would A.I. interfere with elections? Dupe people out of their life savings? Eliminate work at a massive scale and entrench power in the hands of the tech elite?
Colleen Chien, who co-directs the Berkeley Center for Law & Technology, said A.I. agents bent on a particular end could blackmail people or coerce them into acting destructively. But, she said, the digital presences need human help.
“For a nuclear weapon or a virus to be released, there’s a person at many gates,” she said. “And that person is going to have to be manipulated into thinking that their best option is to move forward as told.”
Mr. Harris said the challenges of A.I. are real: It could increase inequality and degrade public information and trust, fatally eroding democracy. But, he said, those challenges are not insurmountable. He classed the task with other big problems, such as tackling climate change, the next pandemic and rising inequality.
“This is a really great moment for us to call on our democratically elected officials to pass laws urgently holding A.I. companies accountable for the risks that their products pose to society and the harms that they create,” Mr. Harris said.
That could include establishing penalties like steep fines or jail time for leaders of companies that lose control of their systems, either to a bad actor or to the A.I. technology itself, he said.
“We have to also make sure that it’s clear in our legal system that A.I. is not a person, and therefore A.I. does not get the protections of the First Amendment or other constitutional protections,” he said.
Dr. Amodei in his Saturday essay said Anthropic would invite independent auditors into the company and ask them to publish their findings. He also said A.I. companies should agree on safety standards and limit how fast their capabilities could advance, and that the United States should get other governments to accept similar limits.
In the European Union, policymakers in 2023 agreed to a law that would regulate artificial intelligence.
For his part, Mr. Trump said Monday that letting A.I. companies continue their work unhindered is crucial to competing with China. The technology needs no guardrails, he said, just a “HIGH IQ” president.
Mark MacCarthy, a professor at Georgetown University and senior fellow at the Brookings Institution, said in an email that recent high-profile episodes in which A.I. agents misbehaved reflect human failures.
“We are not on the verge of losing control of superintelligent systems,” he said. “The mishaps that occurred resulted from A.I. company actions that were unsafe.”
Rena Pearson, a 54-year-old in suburban Atlanta, said that for her husband, an information technology worker, and her two adult daughters, A.I. has become a tool like any other. “You could use a hammer to kill somebody but you could also use it to build a house,” she said.
She said she was worried that it could fall into the wrong hands, but not that it would go rogue and turn on humans.
“It does seem like we could just unplug it,” she said.
Sabrina Tavernise contributed reporting. Georgia Gee contributed research.
The post As A.I. Fears Rise, Experts Suggest Focusing on the Humans in Charge appeared first on New York Times.
