Trump, Nukes, AI, and Civilizational Suicide

As human beings we have always been sitting ducks to extinction’s cosmic whims. We are a few astronomical units away from a life-ending meteor or comet strike, the kind that wiped out the dinosaurs after they were their own animal kingdom for a couple of hundred million years. Nothing we can do about that.

For the brief time human beings have roamed the earth–less than five seconds if earth’s history were a 24-hour clock–we’ve shown nearly equal aptitude and enthusiasm for reproducing, and for suicide. Aptitude for reproduction is nothing special. As Cole Porter put it, birds, bees and even fleas do it. But you have to be human to know suicide, by which I mean the act of engaging in knowing self-destruction, as most wars of recorded and (easy to guess) unrecorded history of the last 100,000 years have been.

Those mass suicides have ended tribes, empires and civilizations, but so far not humanity itself. It is only since 1945 that we’ve acquired that skill. A global nuclear war today would end life on the planet, which would become “a republic of insects and grass,” as Jonathan Schell, the Cole Porter of nukes, put it in The Fate of the Earth (1982).

There’s a cynical glibness to how the anxieties are being dismissed as just the latest chicken-little scenario.

Nuclear war was the top public concern of 1985, even ahead of unemployment, which averaged above 6 percent all year. We worried, we debated, and to Gorbachev’s and Reagan’s credit, in that order, we got a start on arms control deals that improved with time, at least until Obama, Trump and Putin decided to bring us back near the brink, where we are now: the Bulletin of the Atomic Scientists’ famous Doomsday Clock is at 85 seconds to midnight, “the closest the Clock has ever been to midnight in its history.” Almost half of Americans think a nuclear war is likely in the coming decade, though they don’t seem to be too concerned about it, as if it were something they might still get to watch on TV: This year’s list of top concerns doesn’t even include nukes.

Since the 19th century we’d been developing a new method of global suicide, only we didn’t know it until June 23, 1988, when NASA’s James Hansen testified to Congress that “the evidence is pretty strong that the greenhouse effect is here.” The implication was that global warming was a threat, but not yet a life-ending threat. Hansen and other panelists before the Senate Energy and Natural Resources Committee happened to speak on a day when temperatures in the Washington, D.C. area neared 100. He predicted that 1988 would prove to be the warmest year on record. It was, though the record was short-lived, as have near-annual new records since. The panel said the world had, with reasonable and prudent policies, time to respond.

It didn’t. For decades the world has been aiming to limit warming to 1.5 Celsius (2.5 Fahrenheit), above pre-industrial levels, the threshold climate accords had said we cannot safely cross. We will overshoot the target by the end of the decade. That’s not bad news just for a few Pacific islanders, but for places like Flagler Beach as seas continue to rise: that extra 10 feet the designers added to the new pier opening in a few months may not look so high in a few decades.

We’ve known since 1988 that warming is an existential threat to humanity. It is now our third means of civilizational suicide. As with wars and nukes, it’s still controllable. The solution has always been in our hands.

We now have yet another existential threat to humanity, and this one, like meteors, may soon be out of our control: artificial intelligence. Unlike meteors, the chances of extinction are not near zero. Anthropic’s Evan Hubinger warns there is a greater than 10% chance AI could kill all humans within the next ten years if unregulated. You don’t have to be a genius to understand how dangerous and already out of control your peppy AI bots are. If you use anything like Gemini or ChatGPT, you know that they hallucinate and fabricate, they mislead, they get an attitude, and they double down to hide their mistakes. Every day we’re learning of dangerous AI hacking their human creators never anticipated. Let’s hope they don’t know their way to the nuclear codes.

There’s a cynical glibness to how the anxieties are being dismissed as just the latest chicken-little scenario. I was surprised to hear even my usually more prudent colleague Brian McMillan claiming on the radio Friday that he “can’t tell whether that’s Y2K-scare level or not,” and calling fears an “extrapolation.” I don’t know how a reportedly informed journalist can make that analysis when Anthropic CEO Dario Amodei last week called for a pause in the technology’s development, as did OpenAI’s Sam Altman, SpaceX’s Elon Musk, Google DeepMind’s Demis Hassabis.

Every day we’re learning of dangerous AI hacking their human creators never anticipated. Let’s hope they don’t know their way to the nuclear codes.

These men hate each other. They all agree that differences must be set aside to build guardrails–on their business. I don’t know that you can point to an instance in American history when the chief executives of their own companies called on regulatory guardrails about their own product, when you could so easily “extrapolate,” to use Brian’s word, from their fears that they don’t trust their own product. Nor do they trust each other to singly address the crisis, knowing what they know of their product and of themselves.

As a New York Times report explained it, “A.I. leaders are girding themselves for a day when an A.I. system can rely on itself to grow more advanced rather than requiring the aid of human researchers. This dynamic, which computer scientists call ‘recursive self improvement,’ is what some top A.I. developers believe may lead to the software spinning out of their control.”

The technology is new. The existential crisis is not. Oppenheimer, Szilard, Bohr, Einstein and many other scientists involved or not in the Manhattan Project that developed the first atom bombs campaigned for guardrails about what they’d wrought, but the company they were working for (Einstein wasn’t) was the American government, and their audience were the American and Soviet public–similar to the AI industry’s audience now.

As a matter of self-defense, AI is the most important scientific, ethical and moral priority of the day. We can leave the ethics and morality of it to philosophers, theologians and academics. The science is our era’s Manhattan Project. As with the Manhattan Project, the only unifying and resource-ready force to build and enforce guardrails and draft other countries to do likewise is not Anthropic or Space X. It is government, guided by reason and the better angels of our history.

Of course I don’t know what the solution is. I don’t think the AI industry knows. Not yet, anyway. This is a collective effort that requires the pooling of unimaginable thought and resources. But we have guides, we have principles.

“Even if the universe should crush him, man would still be nobler than his destroyer because he knows that he dies and he knows the advantage which the universe has over him,” Pascal wrote four centuries ago. “The universe does not know these things.” Substitute “AI” for the “universe,” and Pascal was onto something: “All our dignity, then, consists in thought. By thought we must raise ourselves, he wrote. “Let us labor, then, to think well: that is the principle of morality.”

We don’t have a solution. But we know the essentials: we know how to get there. The rest is just thought, it’s just raising ourselves.

Instead, we are once again being debased. Donald Trump and his congressional harem now call the AI crisis a hoax. We knew they were immoral. We refused to know they would be suicidal. To his credit, Trump once fast-tracked the Covid vaccine. He has it in him, and could do with AI what he did with the vaccine, at warp speed. Instead, he is slow-walking us to a potential catastrophe as uncontrollable as a nuclear weapon’s chain reaction.

He is our dinosaur in chief, and we’re all sitting ducks.