In 2020, Oxford philosopher Toby Ord published the book “The Abyss” about the danger of humanity’s extinction. He estimated the odds of an “existential catastrophe” for our species in the next century at 1 in 6. This is a very specific number and it is alarming. This statement made headlines at the time and has been influential ever since; most recently in a speech by Australian politician Andrew Lee in Melbourne.

It’s hard to disagree with the idea that the coming decades hold some troubling prospects for us, from climate change, nuclear weapons, and bioengineered pathogens (which I think are all serious problems) to uncontrollable artificial intelligence and massive asteroids (which I think are). why) less anxiety).

What about this number? Where is this from? So what does this really mean?

Coin toss and weather forecast

To answer these questions, we must first answer another question: What is probability?

The most traditional view of probability is called frequency, named after the game of dice and cards. From this perspective, we know from observing the frequency of threes in a large number of rolls that the probability of a fair roll getting a three is (for example) 1 in 6.

Or consider the more complex case of weather forecasting. What does it mean when the meteorologist says there is a 1 in 6 (or 17%) chance of rain tomorrow?

It’s hard to believe that the meteorologist wants us to present a big total, some of which will get precipitation tomorrow. Instead, we need to look at a large number of such predictions and see what happens next. If the meteorologist is doing a good job, we should see that when he says “There’s a 1 in 6 chance of rain tomorrow” it actually rains the next day. So traditional probability depends on observations and procedure. To calculate this, we need a series of recurring events on which to base our prediction.

Can we learn from the Moon?

So what does this mean for the possibility of human extinction? Well, such an event will be a one-off: once it happens, there will be no room for repetition. Instead, we may find some parallel events from which we can learn. In fact, Ord’s book discusses a number of potential extinction events, some of which could potentially be examined in the light of history.

For example, we can estimate the probability of an extinct-sized asteroid hitting Earth by examining how many space rocks have hit the Moon throughout its history. A French scientist named Jean-Marc Salotti did this in 2022 and estimated that the chances of extinction in the next century were around 1 in 300 million.

There is uncertainty in this estimate, of course, but it is supported by something approximating the corresponding frequency calculation. Ord, by contrast, estimates the asteroid’s risk of extinction at 1 in 1 million, but notes the significant degree of uncertainty.

Result ranking system

There is another approach to probability theory called Bayesianism, in honor of the English statistician Thomas Bayes. It focuses less on the events themselves and more on what we know, expect, and believe about them. Simply put, we can say that Bayesians view probabilities as a kind of ranking system. According to this view, the specific number given to a probability should not be taken directly, but should be compared with other possibilities to understand which outcomes are more or less likely.

Ord’s book, for example, includes a chart of potential extinction events and his personal estimates of their likelihood. From a Bayesian perspective, we can consider these values ​​as relative rankings. Ord says the chance of extinction by an asteroid impact (1 in 1 million) is much less than the chance of extinction due to climate change (1 in 1,000) and the chance of extinction due to what he calls “disconnected AI” (1 in 10), both of which believes it is. )).

The challenge here is that initial estimates of Bayesian probabilities (often called priors) are highly subjective (for example, I estimate the probability of AI-based extinction to be much lower). Traditional Bayesian reasoning moves from “before” to “after” and again combines observational data about relevant outcomes to “update” probability values.

And again, the implications for the likelihood of human extinction are small.

Subjective evaluations

There are two ways to think about the accuracy and utility of probability calculations: calibration and discrimination.

Calibration is the accuracy of the true values ​​of the probabilities. We cannot determine this without relevant observational information. Discrimination, on the other hand, is only about relative ranking. We have no reason to believe that Horde’s values ​​are correctly calibrated. Of course, that wasn’t his intention. He shows that they are intended to give “order of magnitude” readings.

Even so, without any relevant observational verification, most of these predictions remain merely in the subjective realm of a priori probabilities.

Not very well calibrated but still useful

So what do we do with “1/6”? Experience shows that most people do not have a perfect understanding of probability (as evidenced by, among other things, the ongoing volume of lottery ticket sales). In such a case, the “likelihood” assessment need not be fine-tuned if you are discussing it publicly; it just needs to have the right psychological effect.

From that perspective, I’d say “1 in 6” ticks pretty much all the boxes. While “1 in 100” may seem trivial enough to ignore, “1 in 3” may cause panic or be viewed as a doomsday delusion.

As someone interested in the future, I hope that risks such as climate change and nuclear proliferation receive the attention they deserve. But as a data scientist, I hope that the careless use of probability will be cast aside and replaced by widespread knowledge of its true meaning and correct use. Source