The End of the World. Again.

It is the end of the world again.

The Fermi paradox asks why, in a galaxy this old and this full of planets, we haven’t met anyone.

One possible answer is the Great Filter: the idea that somewhere between lifeless matter and an advanced civilisation, there is a hurdle almost nobody gets past.

One darker interpretation is that intelligent civilisations eventually become technologically capable enough to destroy themselves.

Nuclear weapons, perhaps, or biotechnology. Or artificial intelligence.

AI has certainly brought the question closer to home.

In September, Anthropic researcher Evan Hubinger said he believed there was a greater than 10% chance that advanced AI could kill all humans within the next decade. Geoffrey Hinton, one of the pioneers of modern AI, later said that estimate was not unreasonable.

That is quite a claim.

But public discourse is rarely that careful. Humans love absolute terms.

Safe or dangerous.

Saviour or Armageddon.

At the turn of the millennium, we expected the Y2K bug to cause aircraft to fall out of the sky at midnight.

Not some computer systems failing. Aircraft, falling.

Absolute language makes complicated things easier to understand.

Perhaps these good or bad framings simply ask the wrong questions.

Take the smartphone.

I can now run Barnaby Cecil securely from almost anywhere in the world. I have more information in my pocket than I could have imagined when I was doing homework from the family Encyclopaedia Britannica.

Yet there is genuine concern about the relationship between social media and teenage mental health.

Even there, the research is more nuanced than the headlines. 

A 2025 review of 32 studies found associations between social-media use and adolescent anxiety, particularly around problematic use, but also significant variation across the evidence.

So perhaps the answer is not good or bad.

It is both.

Nuclear technology is the same.

Splitting the atom gave humanity Hiroshima and Nagasaki. It also gave us a major source of low-carbon electricity.

The technology was not moral or immoral; what mattered was what people did with it.

AI is dual-use in the same way.

It can help find a flaw in a computer system and attack it. It can also help find the same vulnerability and fix it first.

Our jobs are where the absolutes become louder.

Elon Musk has suggested that AI could eventually make human labour unnecessary.

The current evidence is more measured.

The International Labour Organisation estimates around one in four jobs globally has some exposure to generative AI, but expects transformation rather than outright replacement in most cases.

The IMF puts the figure at around 60% of jobs in advanced economies may be exposed to AI, with some workers becoming more productive and others facing reduced demand for parts of their role.

That feels much more plausible than the sudden disappearance of human work.

Some jobs will go. Some will shrink. Many will change. New ones will appear.

That has happened with almost every major technological shift.

None of this means we should dismiss the risks.

AI is advancing quickly. Governments are already testing frontier systems for cyber, biological and other potentially dangerous capabilities.

There will be failures: a bank outage, an airport closure, a disrupted hospital system.

Some of those failures will be serious.

But something going badly wrong is not the same as the end of civilisation.

This is where financial planning offers a useful lesson.

A good financial plan does not work because the adviser successfully predicts every event for the next 30 years.

They cannot.

Markets crash. Governments change tax rules. Businesses fail. People get ill. Families change.

A good plan works because it is resilient enough to cope when the prediction is wrong.

AI deserves the same approach.

Use it where it makes us better.

Set limits when the downsides matters.

Keep human judgement in the loop, build resilient systems and adapt when reality turns out differently from what we expected. Because it almost certainly will.

I don't know what AI will look like in ten years. Nobody does.

I am reasonably confident about what matters in the meantime.

Look after yourself.

Look after your family.

Look after the people who depend on you.

Use the tools available to you without surrendering your judgement to them and when the next apparently existential problem arrives, work out what needs doing and deal with it.

That is not a guarantee that everything will be fine.

It is something rather more useful: a plan.

‍