- See more Daily Mail on Google - save us as a Preferred Source
Earlier this month, artificial intelligence (AI) researcher Jacob Coxon resigned from Anthropic after just four months.
asserted
Coxon → see → months
In an announcement on X, he stated: 'The people building AI earnestly believe that it could kill us all by the end of the decade.'
uncertain
it → state → decade
A senior member of Anthropic's staff, Evan Hubinger, actually agreed with Coxon, adding he personally thinks the chance of this happening in the next decade is more than 10%.
Understandably, these statements made waves.
asserted
statements → agree → waves
But how exactly might AI kill us all?
uncertain
AI → kill → us
There's no shortage of fantastical scenarios, and most of them involve the concept of 'superintelligent' AI – that is, AI that's more capable than humans.
asserted
that → be → humans
I've distilled these scenarios down to the top five, ordering them roughly from most vague to most precise.
asserted
I → distil → precise
And I'd argue the list is also ordered from least probable to most probable.
asserted
list → argue → probable
We'll never know
AI doomers often justify their concerns by means of an annoying catch-22 paradox: how can we possibly imagine what a superintelligence might do to take out less intelligent beings like us?
uncertain
superintelligence → know → us
We'd have to be superintelligent to predict what a superintelligence would be able to do.
asserted
superintelligence → have → what
It's like asking your family dog to imagine thermonuclear war.
asserted
It → ask → war
Current AI models are really good at solving particular problems, but that's not the same as being more intelligent than a human in all domains.
asserted
that → solve → domains
However, AI did recently solve one of the seven most challenging maths problems known.
asserted
AI → solve → problems
It's apparently closing in on others, which might leave you feeling less optimistic here.
uncertain
which → close → you
What if AI got into the nuclear command and control chain and started a nuclear war?
asserted
AI → get → war
A superintelligent AI would likely be extraordinarily competent at achieving its goals.
asserted
AI → achieve → goals
A classic example of such indifference comes from Oxford philosopher Nick Bostrom's imagined superintelligent AI that's been designed to optimise paperclip production.
asserted
that → come → production
To produce its preferred form of office supplies, it quickly converts all available matter – including humans, planets and stars – into paperclips.
asserted
it → produce → paperclips
What we have here is the perfect execution of improperly specified objectives.
asserted
have → specify → objectives
The AI doesn't hate humanity; it simply recognises we're composed of atoms that could be better utilised for paperclips.
uncertain
that → hate → paperclips
The good news here is that this scenario confuses intelligence with power.
asserted
scenario → confuse → power
A superintelligent AI doesn't necessarily have the power to achieve its goals.
asserted
AI → have → goals
Turning the planet into paperclip factories would require planning permissions.
asserted
Turning → turn → permissions
Even if it got the permissions, building too many paperclip factories would lead to inevitable public outcry.
asserted
building → get → outcry
Interest groups would block the proceedings in the courts.
asserted
groups → block → courts
Environmental activists would block the bulldozers.
asserted
activists → block → bulldozers
There's a lot of friction in the world that prevents even the very intelligent from imposing their will on the rest of us.
asserted
that → be → us
In fact, you could think of data centres as a current embodiment of the theoretical paperclip scenario.
uncertain
you → think → scenario
And humans are increasingly pushing back against turning the planet over to data centres.
asserted
humans → push → centres
Humans are increasingly pushing back against turning the planet over to data centres
3.
asserted
Humans → push → centres
Bioweapons
Humanity could be killed by a superintelligent AI making and releasing some dangerous new bioweapon into the atmosphere.
uncertain
AI → kill → atmosphere
This is, in fact, one outcome of the AI 2027 scenario by the AI Futures Project, a non-profit dedicated to forecasting the impacts of advanced AI.
asserted
This → dedicate → AI
This risk was made more concrete last month, when researchers at Stanford University announced they’d used a genetic language AI model to synthesise 16 new viruses.
asserted
they → make → viruses
Worryingly, they just sent the genetic sequences off to a mail-order lab and it sent the viruses back in test tubes.
asserted
it → send → tubes
The whole experiment cost a couple of hundred thousand dollars at most.
asserted
experiment → cost → dollars
The good news here is that it’s remarkably hard to kill everyone with a new virus.
asserted
it → ’ → virus
To do that, you need a virus that’s very transmissible, so it spreads far and wide.
asserted
it → do → virus
But it’s a rule of biology – viruses that spread easily are typically less fatal.
asserted
that → ’ → biology
By contrast, if a virus is very fatal, transmissibility tends to go down, as most people infected die before there’s time to spread the infection.
asserted
people → tend → infection
COVID killed less than 1% of humanity.
asserted
COVID → kill → humanity
The deadliest pandemic in recorded history was the Black Death, when the plague killed more than one-third of Europe’s population in the 13th century.
asserted
plague → kill → century
…and 16 more, not listed.
The Aporia
