Welcome to The Widening Gyre
By Ned Desmond
This is the first edition of the Widening Gyre, a biweekly newsletter designed to help the readers of First Things understand the juggernaut of technological advances, notably in AI, shaping the world and our lives. I come at the effort as a journalist, which was my work in formative years, mostly abroad in South Asia and Japan. I am by nature and training skeptical of what I hear and read; at the same time, I believe that a good approximation of what’s real and true is usually within reach.
I am not a technologist, but my last thirty years have revolved around Silicon Valley, where I have worked as a journalist, investor, executive, and founder. I am also a Roman Catholic, which equips me with a certain outsider’s clarity, and in 2021 I wrote First Things’ first in-depth article on AI, “The Threat of Artificial Intelligence,” as well as two pieces this year, “The Twisted Consequences of Chatbots” and “Magnifica Humanitas, a Missed Opportunity?”
I borrowed the newsletter’s name from the opening stanza in W. B. Yeats’ 1919 poem “The Second Coming”:
“Turning and turning in the widening gyre
The falcon cannot hear the falconer . . . ”
Given the horrors of that era, Yeats had every reason to be gloomy. He subscribed to an esoteric view that human history moved in two-thousand-year cycles, and that his days marked the end of one, when “things fall apart; the centre cannot hold.”
The imagery of widening gyres and falcons deaf to the falconer is, to my mind, a cautionary metaphor for the technology-driven times we inhabit. Unlike Yeats, however, I don’t hold that any future is inevitable or preordained.
I believe God gave us tremendous capacity, and the freedom to shape our lives and the world. We can choose to think and act, for good or ill, or to be passive and let events sweep us away. This newsletter’s aim is to help thoughtful readers shape their perspective on technology, so they can exert authority and control over technology’s advances, whether at home or in the world at large. Both matter a lot.
The topic is vast and fast-moving, of course, and it’s natural to feel that the hour is late. Important AI experts openly fret that the race to build AGI (artificial general intelligence) will set free technological forces beyond anyone’s control. And we already know that chatbots have unleashed social chaos. At the same time, other classes of AI are making remarkable breakthroughs in science and medicine, which, though often hyped, are nonetheless astonishing. To help cover this ground, I will rely heavily on many great journalists and thinkers, while adding my own brief commentaries.
My education in AI began at TechCrunch, a media business that focused on startups. In collaboration with MIT and Berkeley, I produced several in-person technology events focused on robotics and AI, which is where I got to know a number of leading figures in those fields. I also went in-depth on Berkeley professor Stuart Russell’s 2019 book, Human Compatible: Artificial Intelligence and the Problem of Control, which remains one of the most important works on AI in recent years.
In 2020, I started a nonprofit, Sight Tech Global, dedicated to surfacing the remarkable AI technologists and product designers building to assist people who are blind or losing their vision. My wife’s vision loss was the inspiration. The annual global event rallied thousands of top AI experts and blind technology users, who were sometimes one in the same, such as Microsoft’s Saqib Shaikh, Meta’s Matt King, Berkeley’s Josh Miele, and APH’s Greg Stilson.
For the past six years, I have worked at SOSV, an early-stage venture capital firm, where I have helped founders translate their deeply technical work into language that is comprehensible to nonspecialists. Many of those companies deploy various types of AI speed to develop new cancer therapeutics, make robots safer and more capable, or discover more resilient crop hybrids. This year, I have also worked as an advisor to Molecule.one, a startup applying AI to synthesis in organic chemistry. I closed out college organic chemistry fifty years ago, but I have helped the founders tell their story.
The Hugging Face Incident
Enough about your newsletter author. From The Widening Gyre’s standpoint, the big issue at the moment is the clouds gathering over the AI boom.
Let’s start with what spiked P(Doom) sentiment in July: the Hugging Face incident. In July, OpenAI researchers unintentionally let an experimental AI slip onto the internet that then systematically hacked the servers at Hugging Face, a research hub for AI projects. The AI carried out seventeen thousand distinct actions and gained complete access to Hugging Face’s systems before Hugging Face’s security team pushed the hyperactive interloper out with the help of another AI (incidentally a Chinese one) that could match the rapid-fire moves, as detailed in this colorful New Yorker article. In the following days, both Anthropic and Meta announced that their AIs had committed similar transgressions, as if to assure their investors and customers that they were no less badass.
AI safety researchers, the folks at top AI firms charged with keeping the leash on AI, saw the incident as a harbinger of worse to come, and at month’s end some 1,300 signed a three paragraph petition, “Pacing the Frontier,” that called for Washington “to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.” The petition recalled a 2023 initiative calling for a six-month “pause” in advanced (usually called “frontier”) AI development.
Nothing came of the 2023 call for a “pause,” and the latest collective burst of conscience was oddly timed. Three days afterward was the deadline for federal agencies to complete frameworks for Trump’s Executive Order 14409, “Promoting Advanced Artificial Intelligence Innovation and Security, ” which is a voluntary regime for frontier models to seek government review before wide release. Many of the firms employing the “Pacing” signatories, not to mention their investors, had lobbied hard to water down the executive order, which is why compliance is basically by-your-leave. A cynic might conclude that the whole affair was a bit of “safety” theater.
Should we be worried? Yes, for certain, given that Silicon Valley’s frontier AIs will stop at nothing to build ever more powerful AIs, regardless of the risks. There is one figure, however, who can credibly stand against all that, and he did not sign the “Pacing” petition, nor does he reside in Silicon Valley. That is Nobel laureate Demis Hassabis, perhaps the most consequential figure in the AI world. (Sebastian Mallaby’s Infinity Machine: Demis Hassabis, Deep Mind and the Quest for Superintelligence is a great biography, and at the same time the best book on AI yet.)
The cofounder of DeepMind (acquired in 2014 by Google) and now chief scientist at Alphabet, London-resident Hassabis has neither cheered on the “accelerationist” camp in Silicon Valley, nor a laissez-faire approach to AI, and he has long argued that the best application of AI is human health. He won his Nobel for his work at DeepMind on AlphaFold, an AI-driven system that solved a fifty-year-old effort to predict 3D protein structures, a huge breakthrough in biology.
In mid-July, before the Hugging Face break-in surfaced, Hassabis published on X what may be the most lucid and compelling statement yet on the problem and what’s to be done. Noting that we are in the “foothills of the singularity,” and AGI’s “impact will be unprecedented, perhaps 10x of the Industrial Revolution at 10x the speed,” he warned:
“Urgent action is needed to address risks that might arise as we get closer to AGI. We’ve already seen the challenges frontier models pose for cybersecurity, and other threats including nuclear and bio risks may soon emerge as capabilities continue to advance. On the horizon, we will need robust safeguards to maintain control of increasingly agentic, recursively self-improving systems—and tackle unknown issues that will only become clearer over time.”
What Hassabis proposes far exceeds the current Trump executive order: a public-private entity Frontier-AI Standards Body, modeled on the financial industry’s FINRA, and tasked to establish “assessment protocols”
“. . . that Frontier Models would be required to pass … to be deployed in the US market. Labs would also work with the Standards Body to address any critical post-release vulnerabilities. Model assessments should include rigorous scientific evaluations of capabilities in cybersecurity, biological threats and other high-risk domains. Specific agentic AI tests could look for attempts to bypass safety guardrails or signs of deception, and ensure best practices, such as digitally watermarking AI-generated images and generating human-readable output tokens to understand model reasoning.”
To push a framework like that into place would be a huge undertaking, which Hassabis appears to recognize. He recently stepped back at DeepMind so that he can “put his full attention on actively shaping the future of AGI . . . work that is vitally important to Alphabet and humanity,” as his boss, Sundar Pichai, put it. If there is a falconer in the field of AGI, it’s Hassabis.
This bears watching.
The Data Center Dilemma
The AGI singularity may still be in the foothills, but the Great American Data Center Rebellion is already here, and hard on its heels is the threat of a U.S. financial meltdown stemming from unprecedented AI and data center investment (not to mention record U.S. debt).
As Bloomberg reported, “At least 75 data center projects, worth about $130 billion combined, were blocked or delayed by local opposition during the first three months of this year, according to research group Data Center Watch, approaching the total value of projects stalled by local resistance in all of 2025.” According to a HeatMap News poll, 75 percent of Americans now oppose construction of a data center near where they live, up from 42 percent a year ago. The resistance is overwhelmingly bipartisan, and Republican and Democrat governors alike are now straight-arming data center projects wherever they can.
What happened? The best answer to date is a twenty-six-year-old independent journalist Jasmine Sun’s widely discussed account, “No data centers in my backyard.” It recounts her trip into the heartland to visit several places where the fight over data centers is fiercest. At one level, there are very real fears that these vast data centers will deplete water resources and drive up energy rates. But far more injurious, Sun found is:
“... the way that AI companies engage with communities—forcing NDAs, dangling billion-dollar promises, pushing environmental externalities far away from AI’s wealthy user base—resembles a classic story about dark money in politics. Your politicians are being bought by billionaires and gigantic corporations to screw you over. It’s not Skynet: it’s the oligarchy.
Many people living near data center sites don’t feel that they got a choice. In San Francisco we see how desperate the industry is for compute—atoms still do not scale as seamlessly as bits. That tremendous financial pressure is materializing as fast tracks, fat checks, and limited time for public input. It looks like Big Tech cheating the system, engaging in cronyism and bribery, throwing money bags at politicians and hoping the dollar signs substitute for a positive case.
It turns out that politics is more than an optimization game; voter interests are neither stable nor reducible to dollar amounts. It is also about thymos: the desire to be “recognized as a human being… with a certain worth or dignity,” who can sacrifice material survival in service of higher ideals. Through Fukuyama’s lens, elections and town halls and social media debates matter not because they are maximally efficient preference aggregators, but because they are ‘a stage for the expression of thymos’—a place where people make their values heard.
A backroom data center deal, then, is the negation of thymos. It is corrosive not because it is uneconomical or will make residents worse off in every case, but because it suggests that the demos’s buy-in was never important at all.”
Part of what makes Sun’s account so good is the absence of the victim-oppressor cant that many journalists would reflexively bring to the story. Her reporting paints a very human picture, complete with everyone’s flaws and stupidities. It’s encouraging at one level to see subsidiarity alive and well, and discouraging at the same time because many of these often depressed communities would benefit greatly—at least in theory—from the investment and the employment.
Sun also brings an articulate self-awareness, as a recent Stanford graduate and San Francisco resident, about how the world she normally inhabits skewed her assumptions:
“Back in San Francisco, it is striking how little the AI industry understands its own unpopularity—the reasons that people resist. They ask if it’s a problem of benefits not being generous enough, or else a big misinformation campaign. (By the end of the trip, my friends and I felt slightly ashamed of the caricatures we originally believed in. In general, one should be cautious about assuming your opponents are just reading fake news.)”
That comment made me think about the splendid, almost otherworldly campus Meta built in my home town of Menlo Park, California. The West Campus features Frank Gehry–designed buildings and more than twelve astonishing acres (yes, really!) of rooftop gardens planted with mature oak trees, among other pleasant pastoral distractions. Meta, like all Silicon Valley tech companies, knows what it takes to land the best engineers. It’s a shame the tech titans did not take their fellow humans outside the valley similarly well into account, instead of putting the plow to local sensibilities. That said, the data center champions are improving their game, as Sun points out, and state governments are stepping in to help local officials negotiate better deals.
The big tech companies are nonetheless paying the price as delayed construction slows the frontier-AI juggernaut, contributing to their precarious financial outlook, and at the same time ups the risks facing teetering U.S. financial markets. I have always wondered when Silicon Valley might blow itself up; let’s hope it doesn’t take the rest of us with it, one way or another.
Recommended Reading
“Experts are warning: our AI arms race is putting humanity at risk,” Stuart Russell in The Guardian. The Berkeley professor is among the most credible independent experts on AI. Many of the founders and technologists working in the field are his former students.
“Elon’s Favorite Sci-Fi Prophet Was Wrong,” Niall Ferguson in the Free Press. The many possible futures AI might inspire are the stuff of sci-fi writers. Ferguson punctures Musk’s enthusiasm for the Culture novels of Iain M. Banks and nominates a more reality-based writer, who is also my favorite.
“The A.I. Giants Weren’t Prepared For This,” the Ezra Klein Show (podcast) from the New York Times. Klein talks with Jasmine Sun about her reporting on the data center rebellion.
“The Enemy Within,” Nathan Pinkoski in First Things. This in-depth review of Jacob Siegel’s The Information State: Politics in the Age of Total Control delivers on Pinkoski’s opening accolade that Siegel’s is “one of the most important books of the twenty-first century...a virtuoso performance [that] harmonizes political theory, investigative journalism, and contemporary history to demonstrate how Americans are now governed.”
“The Rise of the Catholic Anti-AI Movement,” Jonah McKeown in the National Catholic Register. An interesting survey of Catholics speaking out against AI, especially against engaging with chatbots.
“A Drone Killed Three Ukrainians. It Was Guided Entirely by A.I.,” Andrew E. Kramer in the New York Times. AI-driven warfare is here, and most of the theoretical constraints on “autonomous warfare” have already been broken. That’s what happens in war, but is it any worse than indiscriminate ballistic missile strikes?
“Meta Reaches $18 Billion Settlement With 48 States Over Child-Safety Claims,” Meghan Bobrowsky in the Wall Street Journal. Well, that seems like a lot of money, and Meta faces more claims. Did Meta anticipate this would just be the tolerable price they had to pay on the way to building a $1.5 trillion company?
