Mike Pezzullo is worried about the apocalypse. In a speech at the Australian National University last night, the powerful Home Affairs boss outlined his vision of the various events which could end civilisation as we know it.

A year ago, few would have seen a global pandemic coming. Now Pezzullo warns we should start thinking about other security threats, like pandemics, which could come from nowhere and dramatically alter the course of modern life as we know it.

Sure there’s climate change and extreme weather events but Pezzullo wants us to worry about other stuff straight out of the realm of science fiction: super-volcanic eruptions blocking out the sun; a powerful geomagnetic storm rendering all electronic technology obsolete; a crippling cyber-attack sending us spiralling offline; the threat (seriously) of terminator-like AI; an asteroid hitting Earth; a nuclear apocalypse.

What are the odds of these apocalyptic visions ever being reality? Turns out, it’s pretty hard to say.

We started out by going to the bookies to see if anybody had run the numbers on the world ending. Turns out nobody would touch it — a market on the apocalypse was simply too crass even for betting companies.

So we asked University of Technology Sydney mathematician Stephen Woodcock about how one might go about predicting the scenarios. The answer really depends on what you’re trying to predict.

“There is a difference between events which are naturally occurring — pandemics, earthquakes, solar flares which have happened and will happen again — and the idea of the entirely man-made scenario,” Woodcock says.

With earthquakes and volcanoes, there were years of seismic data and early warning systems, he said. And pandemics tend to happen once in a century so there would be another in the future.

But when it came to human-made events like a nuclear war or a hostile AI takeover, guestimating was a fool’s game, he said.

“I don’t believe that anybody could give figures within orders of magnitude — you could be hundreds of thousands out,” Woodcock says. “You could obviously guess, but you’re making a whole suite of assumptions.”

The key reason it was so hard to figure out the odds was because, unlike natural phenomena, there’s no precedent for them, or any understanding of what the human response would be.

Put another way, to model the likelihood and impact of a future nuclear war, the best input we could get would be … an earlier nuclear war.

So, as terrifying as Pezzullo’s bolder warnings are it might be some relief that trying to work out when they’ll happen is pure guesswork.