Serious researchers have tried to put a number on the odds of human extinction, even though nobody can calculate them precisely. Oxford philosopher Toby Ord puts the chance of an existential catastrophe this century at about 1 in 6. Astronomer Royal Martin Rees gave our civilisation no better than 50-50 odds of getting through the century without a serious setback. Professional forecasters give much lower numbers, around 1% for extinction by 2100.
A spread that wide tells you these are informed judgements rather than measurements. Each number comes from somewhere specific, and most of the people behind them think the risk can still be cut.
What are the odds of human extinction this century?
It depends on whom you ask, and what exactly you ask them.
| Source | Estimate | What it measures |
|---|---|---|
| Toby Ord, The Precipice (2020) | About 1 in 6 (17%) | Existential catastrophe in the next 100 years |
| Martin Rees, Our Final Century (2003) | About 50% | Civilisation suffering a serious setback by 2100 |
| Global Catastrophic Risk Conference survey, Oxford (2008) | 19% | Human extinction by 2100 (informal survey of attendees) |
| Forecasting Research Institute: domain experts (2022) | 6% | Human extinction by 2100 |
| Forecasting Research Institute: superforecasters (2022) | 1% | Human extinction by 2100 |
These numbers are not directly comparable, because they ask slightly different questions. But even the lowest, 1%, is far higher than the risk most people would accept from any single technology.
What is an existential catastrophe?
Ord uses a broader term than extinction. An existential catastrophe is any event that permanently destroys humanity’s long-term potential. That includes:
- Extinction: no humans left.
- Unrecoverable collapse: civilisation falls and never rebuilds.
- Unrecoverable dystopia: humanity survives but is locked into a terrible state, for example under a permanent global tyranny.
This matters because a catastrophe does not need to kill everyone to end humanity’s future.
Toby Ord’s estimates, risk by risk
Ord separates natural risks, which humanity has survived for hundreds of thousands of years, from risks we create ourselves. His estimates for the next 100 years:
| Risk | Ord’s estimate (next 100 years) |
|---|---|
| Unaligned artificial intelligence | 1 in 10 |
| Engineered pandemics | 1 in 30 |
| Unforeseen human-made risks | 1 in 30 |
| Other human-made risks | 1 in 50 |
| Nuclear war | 1 in 1,000 |
| Climate change | 1 in 1,000 |
| Other environmental damage | 1 in 1,000 |
| Natural pandemics | 1 in 10,000 |
| Supervolcanic eruption | 1 in 10,000 |
| Asteroid or comet impact | 1 in 1,000,000 |
| All natural risks combined | About 1 in 10,000 |
| Total | About 1 in 6 |
Ord compares the total to playing Russian roulette with the future of humanity. He also stresses that these are rough judgements, not precise measurements.
Why are natural risks so low?
Ord’s argument is simple. Modern humans have been around for about 2,000 centuries. If natural disasters regularly carried, say, a 1 in 100 chance of wiping us out each century, we would almost certainly not have lasted this long. Our long survival is evidence that the natural risk per century is very small.
Why is AI at the top?
Ord’s reasoning is that a technology that could become more capable than its creators is unlike any risk humanity has faced, and that we do not yet know how to make sure such systems reliably pursue goals that are good for us. Many researchers think his number is too high, and some think it is too low. We cover the full debate in Could AI really destroy humanity?
Why engineered pandemics?
Natural pandemics, even severe ones like COVID-19, are unlikely to kill everyone. A pathogen deliberately engineered to be more contagious and more deadly is a different matter. As biotechnology becomes cheaper and easier to use, the number of people who could cause such a disaster grows. Ord points out that the Biological Weapons Convention, the treaty meant to stop this, has operated with a tiny staff and a budget smaller than that of an average McDonald’s restaurant.
Why Ord calls our time “the precipice”
He dates the start of this period to 16 July 1945, the first atomic bomb test. For the first time, humanity had the power to destroy itself. His argument is that our power has grown faster than our wisdom, and that the next century or two will decide whether we survive long enough to fulfil our potential. If we do, he argues, humanity could have billions of years ahead of it.
Martin Rees: “Our Final Century”
In Our Final Century (2003, published in the US as Our Final Hour), Rees, the UK’s Astronomer Royal, argued that the odds are no better than 50-50 that our present civilisation survives to the end of this century without a serious setback. He worries most about bioengineering, cyberattacks and how small groups or even individuals can now cause damage on a global scale. As he says in his TED talk, the worst dangers now come from us.
What do professional forecasters say?
In 2022, the Forecasting Research Institute ran the Existential Risk Persuasion Tournament. It brought together subject-matter experts and “superforecasters”, people with a strong track record in forecasting real-world events, and asked them to estimate catastrophic risks by 2100.
- Domain experts put the chance of human extinction by 2100 at about 6%.
- Superforecasters put it at about 1%.
- For AI specifically, experts said about 3% and superforecasters about 0.4%.
The two groups discussed their reasoning with each other for months, and hardly anyone changed their mind. That disagreement is itself one of the most useful findings: careful, well-informed people can look at the same evidence and reach very different conclusions.
Should we believe these numbers?
They are estimates rather than calculations, and there are good reasons for caution:
- There is no track record. Humanity has never gone extinct, so there is no historical data to check the numbers against.
- Experts may be biased. People who study a risk may be more likely to think it is serious.
- Small numbers are hard to judge. Humans are bad at telling the difference between 1 in 100 and 1 in 1,000.
But sceptics usually agree on the core point: the risk is not zero, it is mostly caused by humans, and reducing it is worth the effort. Even a 1% chance of losing everything is a risk that justifies serious attention.
Why there is reason for hope
Unlike an asteroid or the slow brightening of the Sun, most of these risks are within our control. History offers real examples: arms control treaties cut nuclear arsenals from about 70,000 warheads in the mid-1980s to around 12,000 today, and the global effort to fix the Y2K bug and the ban on ozone-destroying chemicals show that coordinated action works.
The most effective responses that experts point to are AI safety research and sensible regulation, better biosecurity and pandemic preparedness, renewed arms control, and international cooperation. See the Doomsday Clock for how scientists track these risks year by year.
Frequently asked questions
What is the probability of human extinction?
Estimates range from about 1% to about 19% this century, depending on who is asked and how the question is defined. Toby Ord’s widely cited figure is about 1 in 6 for an existential catastrophe in the next 100 years.
What is the biggest threat to humanity?
In Toby Ord’s estimates, unaligned artificial intelligence is the largest single risk, followed by engineered pandemics. Other experts rank nuclear war or biological threats higher. Nearly all agree that human-made risks outweigh natural ones.
Will humans go extinct by 2100?
Very probably not. Even the more pessimistic estimates put the chance below 20%. But most experts think the risk is high enough to deserve far more attention than it gets.
How long will humans survive?
Nobody knows. If we avoid self-inflicted catastrophes, the main natural limit is the Sun, which will make Earth uninhabitable in about 1 billion years. See the full scientific timeline.
Is climate change an extinction risk?
Ord estimates the chance that climate change alone causes an existential catastrophe at about 1 in 1,000. It is far more likely to cause enormous harm without ending humanity, and it makes other risks, such as conflict, worse.
Sources and further reading
- Toby Ord, The Precipice: Existential Risk and the Future of Humanity
- Forecasting Research Institute: Existential Risk Persuasion Tournament
- Martin Rees, Can we prevent the end of the world? (TED)
- Bulletin of the Atomic Scientists: Doomsday Clock
- Federation of American Scientists: Status of world nuclear forces
Related articles
Or start from the beginning: all predictions about the end of the world.
We summarise the research and scripture in plain English and link to the originals. If anything here is inaccurate or has changed, please get in touch.