Fears of Extinction Events, with and without Plausible Stories to Back Them Up
Are human beings (and maybe all other living things) going to be killed off ... soon? And if so, what will cause the catastrophe to come about? My working title for this column was "Constitutional Crises, Debt Crises, and Skynet," but for reasons that will soon become clear, that framing ended up being too narrow. We humans have spent centuries conjuring up various ways in which the end might arrive, proving ourselves to be quite imaginative in coming up with an impressive range of end-times scenarios.
Anyone who has been even casually following the news recently will have seen the sudden wall-to-wall coverage of apocalyptic warnings by various insiders from the world of Artificial Intelligence, or AI. (That label is still a misnomer, the accurate term being Large Language Models, but I remain unwilling to tilt at that particular windmill.) Semantics aside, the story that has understandably caught everyone's attention comes from a software engineer who resigned from one of the AI companies, announcing that there was a decent chance that AI would kill all of humanity soon.
Some reports are saying "within a decade," whereas others say "by the end of the decade," which is a fairly large difference, given that the current decade will end in three years, three months, and fourteen days, but of course all of that is entirely speculative and not meant to be precise. Soon, or soon-ish, is the message. The point is that now, politicians in both parties are freaking out, whereas Donald Trump has decided to embrace AI fully and claim that he alone can protect us from any problems posed by a technology about which he knows absolutely (and maybe even literally) nothing. I know that we all feel better now.
In my recent attempts on Dorf on Law and Verdict to debunk the hype around the ridiculous "$40 trillion US government debt will doom us all" news cycle (perpetrated by debt scolds and their allies at The New York Times), I expanded on a point that Paul Krugman raised about the doomsday predictions: What, exactly, is the mechanism? That is, a convincing panic-inducing story cannot merely be that, say, "government debt will cause a bond market crisis that will kill us all."
As I put it in that Verdict column, "when we are told that we must at all costs avoid a purportedly horrible fate, we should at least demand to know exactly how likely that fate is and how bad it would be." And we cannot know any of that without having at least a decent guess as to what the steps are that will take us from A to World War Z.
Speaking of which, it is obviously not only AI and government debt that have grabbed people's attention as possible end-of-days stories. For decades, zombie movies have captured our imagination, with variations that are fully mystical (undead zombies) as well as those that at least purport to have a cause (rage zombies), with the unjustly maligned movie "World War Z" falling into the latter category. And long before zombie movies came along, there have been stories about vampires and other beings that might or might not entirely wipe out humanity.
Actually, once I started to think about the many ways in which humans have imagined their own erasure, I felt overwhelmed by the ubiquity of it all. Meteors crashing into Earth, alien invasions (of the "Men in Black" sort, not the Stephen Miller fascistic sort), and on and on have always been with us.
Is the AI story just another example of "Jaws," the 1975 movie thought to be the first true summer blockbuster, which scared the living hell out of people, wrongly causing them to believe that shark attacks are an important thing for everyone to fear?
To be clear, any Armageddon story can be contrived with some kind of mechanism underlying it. Christian evangelists have said for as long as anyone can remember, for example, that God will inflict his wrath on people who tolerate "the gays." Mechanism: (1) Humans defend LGBTQ+ civil rights, because all people deserve dignity, (2) God gets angry, (3) The All Powerful One ends the world that He created, most likely in a fire of damnation.
On a more prosaic level, the oft-invoked scary crisis regarding US government debt does have a story that not only might one day happen but indeed has happened a few times in history, with a government issuing so much debt that the financial markets lose all confidence in the country in question, selling off financial assets and thus destroying the economy by making assets worthless.
As I noted in my Verdict column, however, that story is not enough to cause reasonable people to start digging fallout shelters. (Oh right, nuclear war is another favorite disaster scenario!) I quoted Krugman's response to the possibility of a Greek-style crisis happening in the United States, which (unlike Greece) issues debt in its own currency: "As far as I can tell—and I’ve studied the issue at length—there are no historical examples in which this happened.'"
Perhaps I should call myself on the carpet here as well, not in predicting a government-debt-fueled financial crisis but rather a constitutional crisis coupled with a monetary crisis. The disaster scenario in my writing (frequently co-authored with Professor Dorf) over the years concerns Republicans' hostage-taking via the debt ceiling. I/we have noted the inevitable constitutional crisis that would follow from the President being forced to choose among only illegal options, putting him at unavoidable risk of impeachment and removal from office.
That would be plenty bad, but I/we never claimed that that alone would be an extinction event. We did, however, call out the people (including Krugman, as it happens) who were saying that the President could avoid all of that by having the Treasury embrace the Big Coin Option -- exploiting a supposed legal loophole to mint a platinum coin carrying any arbitrarily large nominal value. That, we said, would create a monetary crisis and thus destroy the global economy.
How? The mechanism there is: (1) A President suddenly says that he can create trillions of dollars at his whim, (2) people are confronted by the fact that money is a social delusion (necessary and useful, but still a shared delusion), (3) being forced to acknowledge the illusion makes people no longer think that their money is valuable, (4) everyone loses the trust on which any financial system works, (5) economic collapse.
I still believe everything in that particular disaster scenario, even though there are no examples of it happening before. But even though there has never been a Big Coin-triggered crisis, history is littered with examples of money losing value the second that people realize that other people no longer accept whatever they had been happily accepting as money. (This problem can happen with gold and other commodity moneys as easily as it can with government-printed currencies.)
What does any of this tell us about the sudden panic over AI? To their credit, quite a few people have been asking, in essence: "How, exactly, will this happen?" It is not enough merely to talk about Skynet and Terminators, with the possibility (which science fiction writers have been toying with forever) of the machines at some fateful point deciding that humans are the problem, or at least expendable.
In that sense, there is at least the germ of an idea of how the AI extinction event would work, with machines becoming capable of controlling systems and taking steps (whether "conscious decisions" or not) that kill people. But could happen here is exactly the same as in the government-debt crisis scenarios. The issue is whether it will happen in any plausible scenario.
How is that issue being addressed? Not particularly well.
In a recent BBC interview, for example, a co-founder of Anthropic offers little more than the warning that AI is potentially dangerous. Starting at the 1:12 market of the linked video, he says that "... the technology's got substantially more powerful, and those risks which were once academic or theorized, we can now see the early signs of them." The interviewer asks him to be more specific (asking about "swarms of AI taking over the internet" as one example), but the response was fairly unhelpful: "The basic risk is, how well do you fancy your chances at being able to deal with the risks of a system that is a hundred times smarter than the smartest person on the planet? I think that your chances of dealing with that and getting out in a safe way is pretty low."
Sure. But what could happen, exactly, and how? "[Y]ou need to control AI technology because you're making things that might be smarter than people. Things that are smarter than people could be really, really, really dangerous." Hmm. Later in the video (at the 3:35 mark), the closest we get is a reporter quoting other experts saying that "unchecked, the economic impact could be unprecedented white-collar unemployment."
I am genuinely fighting my urge to be snarky here, because the guy from Anthropic is not obviously and completely wrong. Yes, powerful things can get out of control, and he does say that what he is looking for is a policy response that requires "kill switches" to keep human beings in control. That is all to the good, but the skeptic in me cannot help but notice that the doomsday scenarios still sound an awful lot more like "We're all gonna die!!" than convincing hypotheticals of cause and effect.
Meanwhile, a colleague pointed out to me in a recent conversation that we have an actual doomsday scenario that is fully and obviously in motion as we speak, one for which the mechanisms have been laid out for decades but that has been allowed to grow worse every day. She was, of course, referring to the climate crisis.
And it is only fitting that Trump's intervention on the pro-AI side of the recent panic pointedly compared it to climate concerns. As the AP reported two days ago: "'AI taking over the World, destroying Humanity, and all other things bad, is a HOAX,' Trump posted on social media, later suggesting that those calling for guardrails on the technology had also warned about climate change, which he also called a hoax."
So that is the ignoramus's version of skepticism. Outside of that, however, there is nothing wrong with adjusting our assumptions and becoming more cautious when it comes to AI. When the doomsayers are able to be less vague, that should change the conversation. But will it? The Trump approach on the environment (which has been the Republican Party's approach for decades) simply rejects all bad possibilities out of hand, because it is inconvenient to them to admit the truth. They might already have decided to do the same with AI.
For everyone else, however, even a great deal of skepticism should not be confused with a stubborn refusal to look at evidence and arguments. On AI, I remain skeptical. It is surely more likely to kill us off than a meteor strike or the rise of the undead, but we have a lot of work to do to figure out the genuine dangers and thus the best responses.
- Neil H. Buchanan