Don’t fear nuclear war – a killer plague or rogue AI are more likely to end humanity

Humanity tends to lack a long-term perspective because there has been little in our evolutionary history that rewards such thinking
Long-term strategies can help avert existential threats we create ourselves, such as climate change, lab-engineered viruses and artificial intelligence

Gwynne Dyer

Published: 1:30am, 18 Sep, 2022

Why you can trust SCMP

Which would be worse: a global nuclear war with all buttons pressed or real, self-conscious artificial intelligence that goes rogue? You know, the central theme of the Terminator films.

An AI called Skynet wakes up and immediately realises that humanity could simply switch it off again, so it triggers a nuclear war that destroys most of mankind. The few survivors end up waging a losing war against the machines and extinction. However, this fantasy has too many moving parts, so let’s try again.

Which would be worse: a nuclear “war orgasm” – Herman Kahn’s description of the Pentagon’s nuclear war strategy circa 1960 – or a designer plague created in some secret bioweapons lab? The plague, obviously, because it could theoretically wipe out the human race while all-out nuclear war probably can’t.

The distinction between a 99 per cent wipeout and a 100 per cent wipeout is insignificant if you happen to be one of the victims, but Oxford University philosopher Derek Parfit thought that it actually made a huge difference.

If only 1 per cent of the human race survived, they would repopulate the world in a few centuries. If the human race learned something from its mistake, it might then continue for, say, a million years – the average length of time a mammalian species survives before going extinct.

01:57

China steps up pace in new nuclear arms race with US and Russia

Even if the human population is limited to 1 billion next time round, that’s a trillion lives in the balance and most of them would probably be worth living. By the way, the climate change problem goes away instantly if you reduce the human population by 99 per cent. Whereas if 100 per cent of the population dies now, all those potential future lives are also lost.

As Parfit wrote: “Civilisation only began a few thousand years ago. If we do not destroy mankind, these few thousand years may be only a tiny fraction of the whole of civilised human history.” This perspective is sometimes called “long-termism”, and few people can manage to hold onto it for very long.

That’s hardly surprising because there has been little in our evolutionary history that really rewards long-term thinking. We didn’t even know about big threats to our survival such as giant asteroid strikes and, if we had known, there was nothing we could have done about them anyway.

Now we do know about them. They have multiplied because of our own inventions, but it took another Oxford philosopher, Toby Ord, to list and rank them. It turns out that the most dangerous threats are not human hardware. They’re software.

A placard from the Double Asteroid Redirection Test briefing at the Johns Hopkins Applied Physics Laboratory in Laurel, Maryland, on September 12. The goal of the mission, which launched in November 2021, is to hit an asteroid with a spacecraft to alter its trajectory. Photo: AFP

“I put the existential risk this century at around one in six: Russian roulette,” Ord says in his book The Precipice. But “existential” means a threat to the existence of the human race, and we are quite hard to kill off.

Nuclear war is not likely to do it. Even if it caused a full-scale nuclear winter, lasting for years and starving most of the human race, a few “breeding pairs” – in James Lovelock’s words – would almost certainly survive.

A hothouse world or extreme glaciation wouldn’t do the trick, either. The planet’s climate has been through all sorts of extremes in its long history, and life survived them all. On a big planet like ours, there are always some places where it’s warm enough or cool enough to hang on through the extreme times.

The truly existential threats are the ones we might create ourselves, like AI that gets out of hand or an ethno-specific engineered killer virus that mutates just a little bit. But that’s software, or “wetware”, and few people take it seriously.

US-China race to build killer robots ‘a threat to humanity’

As Ord points out, “The international body responsible for the continued prohibition of bioweapons (The Biological Weapons Convention) has an annual budget of just US$1.4 million – less than the average McDonald’s restaurant.” Only a few tens of millions is spent on research into AI safety, compared to many billions in general AI research.

So if we keep rolling the dice, some time in the next few centuries we are bound to get the apocalypse in one way or other. But Ord’s prediction, even if it is accurate, is based on the assumption that we carry on heedlessly and never develop the long-term perspective that would enable us to reduce the risks.

In fact, many human beings are already starting to think long-term and act accordingly. It’s not all of us, and much too slowly, but it is happening.

We are trying to change our entire economy to avert catastrophic climate change. We are even experimenting with ways to divert asteroids on a collision course with Earth. It’s not nearly enough, but it’s not bad when you consider that, 500 years ago, many people didn’t even know the Earth was round.

Gwynne Dyer’s new book is “The Shortest History of War”