The State Dinner of Doom
"The danger is not data centers, but the belief in data centers"
When most people hear “AI is dangerous,” the image that pops into their head was storyboarded in 1984 by James Cameron. Skynet becomes self-aware, decides humanity is a threat, and launches the nukes. Cue Arnold Schwarzenegger in biker leathers, cue Sarah Connor learning to do pull-ups, cue thirty years of sequels.
If you are as old as I am, your envisioning probably drifts back to Isaac Asimov’s I, Robot in 1950.
As a model of how AI might actually go wrong, these stories are almost useless — and this matters, because it has quietly become the default mental furniture most people use when they try to think about AI risk. Terminator or I, Robot scenarios require a machine that wants things in the way a person wants things: dominance, survival, revenge. It requires malice, or at least a very human sense of self-preservation. The way Anthropic speaks of Claude as if it were a person feeds this sense.
Real machine learning systems don’t want anything. They optimize. A large language model doesn’t hate you; it doesn’t love you; it doesn’t clock out at five and complain about you to its friends. It’s closer to a word-processing autocomplete.
So why does the Terminator framing persist? Partly because it’s dramatic and Hollywood knows what sells popcorn. But partly, I think, because it lets us off the hook. If the danger is a malevolent robot army, then the solution is simple and cinematic: unplug it, blow it up, send Arnold back in time. If the actual danger is more diffuse — bad incentives, concentrated power, systems we don’t fully understand deployed faster than we can study them — that’s a much less satisfying story, and a much harder one to solve with a shotgun.
Like at a certain point when you’re in a society where there are trillionaires, like, that means mass immiseration, right?There is a particular Hollywood flavor of American spectacle in which extremely rich men gather on a stage to argue about whether the machines they are building will kill everyone. It has the tone of a group of arsonists manning a fire safety booth at a lumber convention. At this year’s All-In Summit, the assembled tech aristocracy took turns either sounding the alarm on artificial intelligence or dismissing the alarm as a symptom of moral weakness, sometimes within the same sentence, occasionally by the same person.
Apocalypse Not
Consider Elon Musk, who spent the spring calling one of his AI rivals “misanthropic and evil,” accused the rival company of hating Western civilization, and then, a few months and one $1.25-billion-a-month cloud computing contract later, discovered that the rival’s warnings about AI moving too fast were, in fact, correct. He said so on stage, more or less, agreeing that the industry should slow down while simultaneously running a company racing to build bigger data centers faster than anyone else on Earth.
It’s the kind of position that requires either extraordinary intellectual flexibility or a very good chiropractor. Sharing the stage with him was Gwynne Shotwell, who runs the actual rocket business at SpaceX while Musk tends to his other seventeen jobs, and who used her platform to talk about Starlink and reusable boosters and, delightfully, the prospect of putting data centers in orbit — which is either a brilliant solution to the AI industry’s power problem or the single most expensive way ever devised to keep a server room cool.
Then there was Jensen Huang, CEO of the company that sells the picks and shovels for this particular gold rush, who took a rather different tone. Huang has lately described existential AI risk as not grounded in science, and the fear of it as something close to a hoax — a position that happens to align nicely with Nvidia being worth $5.37-$5.41 trillion. You would have to combine the GDP output of 150 countries just to reach Nvidia’s stock value. Huang isn’t wrong to note that a chip salesman has priors here; everyone on that stage does.
David Sacks, All-In’s own resident philosopher-king and White House tech policy whisperer, has meanwhile built out the most elaborate theory of the bunch: that the recent wave of AI doom-saying is not really about AI at all, but a coordinated “Doomer Industrial Complex,” a phrase that sounds like something Eisenhower would have warned about back when America was great.
Sacks is a billionaire angel investor whose secret to wealth is to buy, build and sell out quickly. Out of college, he used Peter Thiel’s money to invent PayPal with Elon Musk. They sold it to eBay for $1.5 billion. He then founded Yammer to compete with Twitter and Facebook at the boardroom level. Microsoft bought it after 4 years for $1.2 billion. He then became a VC (known as Craft Ventures) and put down bets on Facebook, Uber, Airbnb, SpaceX, Bitcoin and Palantir. He ran the table.
Sacks says the whole AI doomer thing has the whiff of a psyop aimed at fencing the best labs with a regulatory moat. It is a genuinely interesting theory, timed conveniently around election season. It’s also extremely convenient for a man whose entire portfolio depends on nobody regulating anything.
Somewhere in the middle of all this stood Dario Amodei, Anthropic’s CEO, who published a weekend essay imploring the industry to “pace the frontier” — his phrase for slowing down without technically stopping, the AI-era version of unilateral disarmament. Amodei has since admitted his hardest dilemma is China: any American slowdown risks ceding ground to a country that shows no particular interest in pacing anything, so it is conveniently said. Musk, Sam Altman, and Demis Hassabis all publicly nodded along with Amodei’s plea, forming perhaps the first bipartisan coalition of bitter corporate rivals in Silicon Valley history, united entirely by the shared realization that going slow is much easier to recommend once you’re already winning.
And then, presiding benevolently over the entire circus by remote, was Vice President JD Vance, who has made a career out of a single elegant point: that watching multibillion-dollar AI companies march into Washington begging to be regulated looks suspiciously like a Trojan horse, one large enough to hide the entire competitive landscape. Vance wants everyone to relax, admire the craftsmanship of the horse’s wheels, and be assured there is nothing to be scared of.
There are Valid Points of Concern
Our capability is scaling faster than our understanding of these systems. This, it should be noted, is a 300,000-year persistent problem with homo in the living biosphere. It is no longer about extinguishing woolly mammoths. Now we can build models that pass bar exams and write passable poetry, and even the researchers who build them will tell you that they don’t fully understand why the models do what they do. That’s an odd position to be in with any powerful technology, let alone one being wired into hospitals, courtrooms, and financial systems.
Second, the incentive structure is genuinely worrying — not because any single company is evil, but because competition has a way of overriding caution. If one lab slows down to be careful, and three others don’t, the cautious lab doesn’t stop the technology from being deployed; it just loses the race and any influence it might have had. It’s the same logic that governed the nuclear arms race and gain-of-function research.
Third, and probably most legitimately alarming: even a “dumb” AI system — one with no goals, no self-awareness, nothing remotely like intent — can still cause enormous harm simply by being effective at a bad task. You don’t need Skynet to write a more convincing scam email, design a more targeted disinformation campaign, or lower the cost of synthesizing something nasty. Bill Joy worried that genetic engineering, nanotechnology, and robotics were becoming so cheap and open source in 1999 that a small group — or a bored teenager in a high school lab — could cause catastrophic, irreversible harm. Unlike nuclear weapons, these technologies don’t need uranium mines and centrifuge halls. Joy’s fear about accessible destructive capability applies just as well to AI as it did to biotech, since AI is explicitly built to lower the barrier to doing complicated things.
None of that requires robots with red glowing eyes. It requires ordinary human ambition, competition, and carelessness, amplified by a tool that’s gotten remarkably good remarkably fast.
Which Brings Us to Thursday
On Thursday, Xi Jinping arrived at the White House with his wife for the first Chinese state visit to Washington in over a decade, and President Cobblepot seated him at a table flanked by, depending on the final RSVP list, a guest list that read less like a diplomatic dinner and more like the cast of a disaster film — some combination of Sacks, Musk, Huang, Bezos, Altman, Pichai, and Zuckerberg.
Imagine it: fine china, awkward toasts, and at least three men at the table who have spent the past two weeks publicly disagreeing about whether their own product might end civilization, all now expected to make polite small talk with Madam Peng, a renowned singer, champion of ecovillages and climate policy, UNESCO Special Envoy for women and girls, and a Major General in the PLA. Her husband Xi, whose own AI industry has shown no interest whatsoever in “pacing the frontier,” gets to watch the tech elite perform their doomer cabaret in real time and draw whatever conclusion suits — that they are thoughtfully deliberating the risks of a civilization-altering technology, or that they’re too busy fighting each other to notice that the king they’ve enthroned is quite mad. Xi will almost certainly conclude the second thing. The whole “negotiation” — the reason he came to dinner — you know, chips, tariffs, and Taiwan? will be amusing but inconsequential. At best, the Mad King will be bathering on about secret sniper bullet caches or algae. At worst, Jared will suggest swapping Taiwan for golf resort rights on the China Sea.
The real comedy, of course, is that none of this changes the menu. Whatever gets said between the soup course and dessert, the underlying dynamic will remain exactly what it was walking in: two governments locked in a race neither believes it can afford to lose, flanked by a room full of billionaires who agree on almost nothing except that someone else, somewhere, should probably slow down first. Pass the wine. It’ll be a short dinner.
The Church of the Machine
If you want to understand why this whole spectacle feels less like a policy debate and more like a tent revival, dig up a copy of Paulina Borsook’s Cyberselfish. Borsook spent the late nineties embedded among tech engineers and came away with a damning, prophetic diagnosis: a culture violently short on compassion, allergic to regulation, and so enamored of its own cleverness that it mistakes technical mastery for moral authority. All that, while being one of the largest beneficiaries of government largesse in history, and mind you, this was before Elon. She wasn’t describing AI. She was describing the species of person who would eventually build it, and who would arrive at every subsequent controversy absolutely certain that the rules other institutions live by simply don’t apply to them. Sound familiar? If you liked Trump One or Trump Two, you’ll love President Vance.
That is the part worth sitting with, because it explains the theological energy better than any argument about the Singularity. Every camp in this fight — the Kurzweilian transcendence-through-silicon crowd, the Sacks-style true believers in market purification, even Amodei’s earnest doomer-priesthood pleading for restraint — is running on faith, not evidence, because nobody actually knows what a superintelligence would do any more than Mormons know what heaven looks like. But faith-based movements convinced they alone hold the truth have a very nasty habit of ending in holy war. The Crusades, the Reformation, the 20th century’s ideological bloodletting, the Gaza Genocide, the Iran Debacle — all of it followed the same script: true believers, all in.
An AI accelerationist convinced she’s building humanity’s salvation and an AI doomer convinced he’s the last man standing between us and annihilation are, structurally, running the same software. Nobody’s checking their priors. Both are quite sure God, or the market, or the math, is on their side.
Which brings us back to Bill Joy, who, like Borsook, saw this coming in 2000, and for exactly the right reason. His fear was never really about the technology itself; it was that the technology would keep getting cheaper and easier to disseminate, until immense destructive capability sat in the hands of people accountable to no one but their quasi-religious conviction. Ironically, Donald Trump is a safer Commander in Chief than JD Vance for this reason. You don’t need a robot with malicious intent to get a catastrophe. You just need true believers, ample compute, and nobody in the room willing to admit they might be wrong.
That, and not Skynet, is the actual existential risk on the table: not a machine that wakes up and decides to hate us, but an old, human, thoroughly documented pattern — zealotry plus proliferation plus zero accountability — now running at silicon speed, dressed up in the vocabulary of venture capital, and seated, this Thursday, at the head table.
Let’s end these wars. We support peace in Iran, Lebanon, the West Bank and Gaza and the efforts to end the war in Ukraine immediately. Global Village Institute’s Peace Thru Permaculture initiative has sponsored the Green Kibbutz network in Israel and the Marda Permaculture Farm in the West Bank for over 30 years. With your assistance, it will continue.
We support Ukrainian families seeking refuge in ecovillages and permaculture farms along the Green Road, and we work to heal collective trauma worldwide through the Pocket Project. You can learn more about our work on the Global Village Institute website (GVI.cool). I appreciate your support.
Could you help me get my blog posted every week? All Patreon donations and subscriptions to Blogger, Substack, and Medium are welcome. You make this happen. Your contributions can be made to Global Village Institute, a tax-deductible 501(c)(3) charity. PowerUp! Donors on Patreon get an autographed book off each first press run. Please help if you can.
#RestorationGeneration.
 |
| Even when humans are locked in a cage, the Earth remains beautiful. So the lesson is that human beings are not necessary. The air, soil, sky and water are still beautiful without you. So, when you step out of the cage, please remember that you are guests of the Earth, not its hosts. |
Thanks for reading The Great Change!
This post is public, so feel free to share it.
We have a complete solution. We can restore whales to the ocean and bison to the plains. We can recover all the tremendous old-growth forests. We possess the knowledge and tools to rebuild savannah and wetland ecosystems. Coral reefs rebuilt with biorock build beaches faster than the seas are rising. It is not too late. All of these great works of nature are recoverable. We can have a human population sized to harmonize rather than destabilize. We can have an atmosphere that heats and cools just the right amount, is easy on our lungs and sweet to our nostrils with the scent of ten thousand flowers. All of that beckons. All of that is within reach.
 |
| These bumper stickers should be on every porta-potty at Burning Man! We now have our own merchandise store on Redbubble. Swing on in for the latest wearables and chachkas…. |
Thanks for reading! Everything expressed in this article is my opinion.
Comments