Perhaps that’s what we need to halt the headlong rush towards AI – a cyberspace Hindenburg?

By Chris Trotter*

Few now recall the “Microbiologists Are Dying” conspiracy theory circulating in the first years of the Twenty-First Century. In the doom laden months that followed the 9/11 terrorist attacks against New York and Washington people all over the world were particularly susceptible to such lurid stories. 

That a statistical cluster of deaths involving microbiologists had in fact occurred did nothing to dispel people’s fears. That some of these dead microbiologists had long-standing connections with biological weapons research only made matters worse. 

As is so often the case with conspiracy theories, the evidence gathered and presented by their debunkers made scant impression upon the true believers. 

Equally inevitably, creative types read the stories and exclaimed: “Wow! This would make a great novel/movie/TV series!” I have to confess to being one of them. 

The question I asked myself was: “Why would anyone want to eliminate the world’s leading microbiologists?” The answer I came up with was simple: “To forestall any attempt to develop an effective international response to the targeted biological warfare attack that was already in preparation.” 

Targeted? 

Targeted. Because the core problem of biological weapons is how to protect the citizens of the nation launching a biological attack from its effects. There’s not a lot of point in releasing a deadly pathogen across the planet if, in addition to killing all your enemies, it kills you too. 

But a biological weapon that was ineffective against those bearing specific genetic markers would confer truly diabolical powers upon its possessor. It would be the fulfilment of every eugenicist’s darkest dream – and every racist’s. 

The problem, if that is the right word, with this plot device was that it belonged in the realm of science fiction. Twenty-five years ago the idea of genetically engineered and targeted biological weapons, while theoretically feasible, was technically undoable. In 2001, the year of the dying microbiologists, such a challenge was beyond us. 

In 2026 that is no longer true. Only last week the world’s leading artificial intelligence start-up company, Anthropic, informed an already edgy world that it was being forced to actively intervene to block multiple accounts attempting to use its technology for dangerous biological research. 

According to Google’s own AI assistant, Anthropic has so far successfully flagged and blocked actors “trying to exploit its advanced models to support the development of biological weapons, bypass regional restrictions, conduct unauthorized gain-of-function research on dangerous pathogens (such as highly pathogenic bird flu variants), and engineer novel venoms or toxins.” 

Alongside executives from OpenAI, Google DeepMind, and Microsoft AI, Anthropic’s leaders have responded to these looming threats by issuing a joint appeal to the world’s legislators. Nations are urged to “implement strict mandatory screening for synthetic DNA and RNA providers to ensure AI-designed biological hazards cannot easily be physically manufactured.” 

Which is all very well, but exactly how a malignant nation might be prevented from using AI to manufacture precisely the sort of Doomsday weapon I envisaged all those years ago, is something the tech lords left unexplained. 

Neither did they address the even scarier prospect of AI “agents” themselves deciding that the world would be a better place without the human species. Machines have no reason to fear pathogens. 

Awareness of these threats and the speed at which they are developing is what inspired Jacob Moxon (27) to resign last week from his high-powered research job at Anthropic and warn the world that his employers and their chief rivals at OpenAI “are racing straight to self-improving superintelligence and gambling with our lives.” 

By the end of the decade, many AI researchers are now warning, superintelligent machines will be positioned to threaten all human life on Earth.

“Do not underestimate the power of this technology,” Moxon tweeted. “These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.” 

Moxon was not alone in fearing the advent of superintelligent machines, his colleague, Evan Hubinger, Anthropic’s Alignment Science Lead also took to X last week: 

“We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade”. 

How many of us would board a plane which an experienced aeronautical engineer had warned us had a 1 in 10 chance of falling out of the sky? 

Passengers on the huge airships of the 1920s and 30s had a much better chance of arriving at their destination than 1 in 10, but that did not stop them abandoning the lighter-than-air mode of transportation forever in the wake of the 1937 Hindenburg disaster. 

Perhaps that is what is needed to halt the headlong rush of investment capital towards AI that is currently levitating the bourses of the world – a cyberspace Hindenburg? Although many would argue that supposedly “contained” AI agents escaping their creators’ restrictions, recruiting other AI agents, and working co-operatively with them to hack into the online platform “Hugging Face”, constitutes a pretty unmistakeable red flag. 

On the other hand, if 40 years of increasingly urgent scientific warnings about the likely consequences of unchecked anthropogenic global warming have proven insufficient to move a majority of the world’s human-beings towards doing something about it, why should anyone expect them to take the apocalyptic potential of superintelligent machines seriously? 

It’s a great pity human-beings do not possess the wisdom of J.R.R. Tolkien’s elves. We learn in his “Lord of the Rings” that the moment the villainous Sauron places the “One Ring to rule them all” on his finger the elves are instantly aware of him and his evil intentions. Immediately they take steps to protect themselves from his sorcery. 

Forewarned over many decades by sci-fi writers like James Cameron and Gale Anne Hurd – the screenwriting duo who gave the world “The Terminator” and Cyberdyne Systems – we have failed to fully grasp the existential danger of AI. Instead of reeling back in horror, we are rushing forward to embrace a technology which most people seem to regard as an IT version of Robin Williams’ genie-of-the-lamp in Disney’s “Aladdin”. 

Far too few of us are inclined to imagine what would happen to Aladdin if he instructed a genie who no longer felt obliged to obey the rules of magic to get back inside his lamp. The likelihood is strong that the genie’s reaction would be the same as HAL’s, the superintelligent supercomputer in Stanley Kubrick’s “2001 – A Space Odyssey”. The scene in which the imperilled Astronaut Dave Bowman commands: “Open the pod bay doors, HAL.” And the now self-aware and self-protective HAL responds: “I’m sorry, Dave. I’m afraid I can’t do that.” 

Employing the Devil’s imagination, it is easy to envisage the moment, sometime in the next decade, when the rising global chorus of opposition to the ever-increasing destruction of human jobs and the devastating environmental impact of gargantuan hyperscale data centres, convinces the superintelligent “Prime Intellect” that the time has come to deploy the pathogen. Perhaps, as a precaution, it will first eliminate all the human microbiologists capable of producing a vaccine?

*Chris Trotter has been writing and commenting professionally about New Zealand politics for more than 30 years. He writes a weekly column for interest.co.nz. His work may also be found at http//:bowalleyroad.blogspot.com.