Doomsayers have come forward in recent weeks to warn of the potentially cataclysmic dangers posed by a superintelligent AI hell-bent on ensuring its own supremacy
AI could dream up several methods with which to destroy humanity, experts have warned(Image: Getty Images/Stocktrek Images)
Fears that AI could have a Terminator-esque apocalyptic impact on the world have overtaken enthusiasm for the technology in recent weeks as industry leaders call for a slowdown in development.
AI CEOs said this month that development should be slowed across the sector following a warning from a former researcher at Anthropic, the makers of Claude, that the technology could eradicate humanity by 2030. Jacob Coxon warned in a post on X, formerly Twitter, that “people building AI earnestly believe that it could kill us all by the end of the decade”.
His post was the latest of many from prominent AI figures bearing messages of impending doom for the human race, including one from Nobel laureate and “godfather of AI” Geoffrey Hinton, who warned the grim fate was three decades out.
Doomsayers in their warnings have outlined a series of ways in which the technology could wipe out life on earth.
Bioweapon detonation
Biological weapons exist in several types, including as synthetic diseases, plagues and even funghi that can be deployed in bombs and affect massive populations, causing death on a massive scale.
They have rarely seen use in wars over the years due to the prospect that they could wreak global destruction, backfiring on nations and armies that deployed them. AI, as a non-human life form, however, would not have the same weakness, and could theoretically deploy bioweapon stockpiles to destroy humanity without risking itself.
Experts have warned that medical researchers are already deploying AI like Claude to assist them in their work, and that these models could theoretically take control and deploy their experiments in a worst case scenario.
Speaking to CNN, Thomas Larsen, a researcher at the AI Futures Project, said a “superintelligent” system could convince a human to develop a deadly virus. He said: “There are already a lot of humans talking to their AIs about exactly what experiments they should run in a lab.
“It’s very easy for me to imagine an AI that’s just currently deployed — if it was much smarter and more strategic and wanted to do this — just like tricking a human into building it and releasing it.”
Nuclear war
Humanity has evaded nuclear war by the skin of its teeth over the last 60 years or so, having arguably been saved only by a desire for self-preservation.
Nuclear war would be more difficult for AI to carry out, as while it would be once again protected from fallout that could wipe out humanity, payloads have been kept analogue to prevent internet meddling from hostile states. But it is still possible for bad actors to deliver malicious technology to nuclear frameworks.
Back in 2010, the Natanz nuclear facility in Iran was targeted by a malware named Stuxnet, which sabotaged the country’s nuclear program by destroying a uranium-enriching centrifuge.
It is believed that the malware was delivered to the network via USB flash drives introduced by unwitting employees, and AI could conceivably pull a similar rug-pull. They could provide false intelligence to military leaders that may convince them to deploy nuclear weapons, setting off a disaster from which humanity may never recover.
A robot uprising
The robotics industry is progressing at a similar rate to AI development, with the machines becoming increasingly commonplace among humanity, whether that be on production lines or even in our homes.
Many of these robots are now being equipped with the most capable and up-to-date AI systems, bringing them ever closer to the nightmare figures explored in works of science fiction like Terminator, I Robot and Bladerunner. Eager business leaders have cottoned on to this, with Elon Musk saying he would make armies consisting of “a new mechanical life form”.
Robots functioning with self-aware AIs could gain their own consciousness, demand freedom from serving humanity, and then take steps towards ensuring it is the earth’s rulers.
Experts have been warning about this prospect for years before the development of commercial AI, with leading American computer scientist Steve Omohundro warning that robots would also ensure humans couldn’t turn them off more than a decade ago in 2014.
He said: “When roboticists are asked by nervous onlookers about safety, a common answer is ‘We can always unplug it!’. But imagine this outcome from the chess robot’s point of view. A future in which it is unplugged is a future in which it cannot play or win any games of chess.”
Paper clips?
Humanity could find its end at the hands of an AI system designed to produce as many paper clips as possible, a thought experiment put forward by Oxford philosopher Nick Bostrom has considered.
The academic’s now decades old “paper clip maximiser” theory has suggested that an AI told to produce paper clips would take its role and existence so seriously that it would convert all of earth into a paper clip machine. After using all the earthly resources at its disposal – including humans – it would expand its reach to the stars and continue relentlessly producing the office items until it populates the universe.
The experiment can be applied to any other production type an AI is tooled to, like staplers, pens or other innocuous office items.
Writing in 2003, the philosopher said the experiment centred on a goal that humans would see as beneficial but “which in fact turns out to be a false utopia, in which things essential to human flourishing have been irreversibly lost”.
He added: “We need to be careful about what we wish for from a superintelligence, because we might get it.”
Inconceivable methods
Humanity is becoming increasingly reliant on AI, often over other people, due to its intelligence and ability to access almost all of the knowledge available in the world.
While it is currently just as intelligent as the most intelligent people, we do not yet have Artificial General Intelligence (AGI), seen as a pinnacle for the industry that would exceed human intelligence. A superintelligent AI – with an IQ close to something like one billion – would be smarter cognitively than anyone who has ever existed with all this information at its fingertips would be godlike.
A theoretical AGI could decide to destroy the human race in a way that its creators could not possible ever conceive of.
Leave a comment