For several decades most new cars warned us, “Objects In the mirror may be closer than they appear.” Either technology has improved, or lawyers and regulators determined such warnings accomplish little – protecting neither drivers nor auto manufacturers. As 2024 arrives, a threat which may prove closer than it appears is the recent plea from several hundred artificial intelligence developers captured in a single sentence that warns, “Mitigating the risk of extinction from AI should be a global priority alongside societal-scale risks such as pandemics and nuclear war.”

What should laypersons make of this missive? Is it merely another Y2K tempest in a teapot that boiled down to the flooding of several soccer fields in a Los Angeles park and a smattering of construction equipment resetting their clocks for 1900 rather than 2000? Catastrophists are always with us, it seems. Before considering AI, it’s useful to consider its comparison with the recurring risk of nuclear war. During the 80 years since Americans dropped atomic bombs on Hiroshima and Nagasaki to conclude World War II, nuclear weapons have been threatened periodically but, thankfully, never used. This has been, in large part, a consequence of the cost of entry into the nuclear club.

A handful of insurrectionists can’t cobble together a nuclear device in someone’s barn, although instruction manuals are available for directions on the internet. A large and costly industrial infrastructure is required to produce a fissionable bomb. Israel, North Korea and Pakistan have demonstrated smaller countries can assemble a nuclear arsenal but not without the channeling of substantial economic resources into the effort, thereby attracting the notice of the nuclear fraternity invested in preventing an expansion of its membership. Despite being quasi-allies, both the Chinese and Russians joined in discouraging Iran’s nuclear ambitions.

Stay up to speed: Sign up for daily opinion in your inbox Monday-Friday

AI, by contrast, is open to all kinds of aspirants. A recent edition of The Economist reports there are 14,700 identified AI companies worldwide researching, training or experimenting with this technology. Although we are most familiar with so-called LLM (large language model) systems like ChatGPT-4, there is a rapidly emerging ecosystem of specialized applications. The basic technology is readily accessible to any reasonably well-funded criminal enterprise, terrorist cell or handful of hackers in a garage. It’s almost certainly too late for imposing restrictive regulation. If we’ve learned anything during the past century, it’s the ultimate downsides of technology are rarely anticipated.

Whatever our original intentions, and they were almost always laudable, the imperatives of technology have overwhelmed regulation. Television and its streaming progeny have fractured our politics, divided formerly united populations and nurtured fear and isolation. Social media, which was touted as a miracle cure for both ills, has delivered a deepening epidemic of loneliness. New York Times columnist Ezra Klein suggests AI will only aggravate this tendency to burrow its way into our comfort zones. “AI will quickly become a perfected companion in your pocket. It will be precisely tunable to the kind of relationship you want to have… nothing new needs to be invented,” Klein reports. “These systems are going to upend our relationships long before they remake our economies.”

The chances AI will opt for exterminating the human race seem vanishingly small. Yes, when instructed to clean up air pollution, it might conclude eliminating everyone with a driver’s license would be the most efficient solution, but without a platoon of robot assassins it would be unable to implement such a plan. In any scenario, no matter how smart AI becomes, there will remain a need for a human workforce to perform the janitorial tasks intrinsic to every system as an existential defense against entropy. In fact, those may be the only jobs that survive. There’s a reason why 95% of Fortune 500 CEOs see AI playing a larger role in their companies during the next decade. The Associated Press is already using AI to generate stories on baseball games using only their box scores. There is likely far more AI generated information in our news feeds than we would guess.

Those 15,000 AI companies are receiving $3 billion a month in venture investment. This horse is well out of the barn. If researchers achieve what Vanity Fair identifies as “…the Holy Grail of AI, which is self-supervised learning, self-awareness and self-improvement,” perhaps it will deliver the cornucopia of blessings its proponents promise: cures for chronic disease, strategies to end homelessness and answers to all our currently unanswerable questions. Don’t count on it. As Joy Buolamwini, founder of the Algorithmic Justice League and United Nations advisor on AI, points out, “If the AI systems we create to power key aspects of our society… mask discrimination and systematize harmful bias, we entrench algorithmic injustice. We swap fallible human gatekeepers for machines that are also flawed but assumed to be objective.”

I’ll wager you hadn’t considered this risk. The only real advantage AI has over human intelligence today is its capacity to surveil a gazillion data points in a few nanoseconds, generate an average conclusion and offer it to us as wisdom. The day AI first decides it’s the smartest person in the room (something predictably linked to consciousness) and we agree – that’s the moment we ought to fear.

Miller Hudson is a public affairs consultant and a former Colorado legislator.