What a Harrowing Nuclear Disaster Movie Says About AI’s Terminator Scenario

In the opening scene of Threads, a young couple called Jimmy and Ruth are parked up on a hillside overlooking Sheffield in northern England, under bleak gray skies. The car radio crackles as Jimmy changes through the stations.

Partial sentences from a news report are audible as he impatiently fiddles with the dial.

“There has been further fighting in Iran between government...the civil war there...” Eventually, he finds the right station. The afternoon’s soccer results start pouring out.

Later, at the pub, Jimmy is drinking a beer with his friend Bob as a television news report details the Soviet Union moving nuclear warheads to an Iranian base, and NATO considering its response to the crisis.

Jimmy notes to Bob that "it's getting serious," but is met with a dismissive reply: "Well there's nowt we can do about it, is there? Might as well enjoy ourselves while we can.”

This is a theme throughout the early part of Barry Hines’ harrowing 1984 film.

Ordinary people going about their lives, distracted by the usual chores and trivialities, as a geopolitical horror story unfolds over their lives, until it reaches its climax literally above their heads as a nuclear airburst.

It was a warning about the very real existential risk to human civilization of nuclear war between Western allies and the Soviet Union. By the irradiated end, our modern society had regressed into a medieval state of drudgery and survival in a long nuclear winter.

The unforgiving starkness of Threads is hard to capture in written word. You’ll have to watch it, but probably not before bed.

The nuclear threat still exists for us, of course, though it has eased somewhat since the worst days of the Cold War, notwithstanding Russia’s recent belligerence in Ukraine, and even against NATO allies.

Instead, the greatest and most urgent threat posed to human civilization is now—if the apparent Cassandras of frontier artificial intelligence are right—a recursive superintelligence that escapes human control.

Yet we’re behaving much like the people of Sheffield in the first half of Threads, and it isn’t for want of hearing the news.

After a summer in which AI models from both OpenAI and Anthropic went rogue during testing, a 27-year-old British researcher named Jacob Coxon quit Anthropic this month, warning that the people building AI believe it “could kill us all by the end of the decade.”

Days later, Anthropic’s chief executive, Dario Amodei, wrote that AI was starting to help build its own successors, and called on the industry to slow down. OpenAI’s Sam Altman quickly agreed, as did Elon Musk.

The public has certainly noticed that something’s up. Half of Americans told YouGov they were concerned AI could end the human race, and 48 percent told pollsters for Politico that work on more advanced models should pause.

But worry is cheap. By mid-September, no leading lab had committed to actually slowing down, instead, for some reason waiting for the government to tell them they must. But if the threat is so great, why continue down that road?

Meanwhile, we weave the technology ever deeper into our everyday lives. OpenAI’s ChatGPT alone had more than 900 million weekly users by February.

Vice President JD Vance said the industry’s pleas for regulation felt like “a bit of a Trojan horse.” President Donald Trump declared that the only guardrail AI needs is “a STRONG AND SMART (High IQ!) PRESIDENT.” Critics would argue that is an unusually thin safety framework for such a transformative technology.

Our relatively recent experiences create an understandable skepticism about warnings of technological doom. Could this be another millennium bug?

In the late 1990s, warnings that computers would read “00” as 1900 raised fears of failing banks, power grids and planes. The U.S. spent an estimated $100 billion on fixes.

When midnight passed quietly, mercifully without any aircraft falling from the skies, Y2K became a punchline and proof of an industry crying wolf. The White House’s Y2K council credited the calm, in part, to the sheer scale of the effort to avoid problems.

AI’s warnings could meet the same fate. Vindicated by their success, and then mocked for it. But Y2K came with a deadline and a known fix. AI has neither.

Evan Hubinger, an alignment science lead at Anthropic, concedes the company has no plan yet for making superintelligence safe.

Then there’s the money. Skeptics say a profound warning that your product is so advanced that it could end the world doubles as advertising, and invites rules that only incumbents can afford. All very convenient for multibillion dollar tech companies pursuing IPOs.

Aidan Gomez, who runs the smaller rival Cohere, called the push for a coordinated slowdown “a cartel by any other name.”

When people want a cultural shorthand for AI catastrophe, they tend to reach for The Terminator, released a month after Threads first aired in 1984.

Its director, James Cameron, told an interviewer in 2023: “I warned you guys in 1984, and you didn’t listen.”

The movie’s killer cyborg is sent back from 2029, squarely within the window Coxon says his former colleagues fear the eradication of, well, us.

But The Terminator is ultimately a work of pure fiction. Skynet’s Judgment Day, which the 1991 sequel dated to August 29, 1997, came and went. Threads, too, remained as fiction. The bombs never fell on Sheffield. Not yet, anyway.

That should temper any certainty of doom, but it shouldn’t inspire complacency. The Cold War stayed cold through luck as much as restraint.

A year before Threads aired, Soviet officer Stanislav Petrov decided that a computer warning of incoming American missiles was a false alarm. He overruled the machine, and he was right—probably saving hundreds of millions of lives from nuclear annihilation.

Some of that restraint was also written into arms-control treaties such as SALT, which Amodei now holds up as a model for a speed limit on AI improving itself.

He has warned that within six to 12 months, a misaligned swarm of AI agents might be able to hijack the whole internet with a self-sustaining botnet.

Yet Trump, and China’s Xi Jinping, are telling their AI frontiersmen to push on, harder.

Perhaps the warnings from the likes of Amodei will come to nothing, by destiny or design. Or maybe these are the last few months of human civilization as we know it.

For now, we’re just a guy in a car, changing the radio station beneath bleak gray skies.