In partnership with
Imagine 50 million people materializing in 2027. Each smarter than any Nobel Prize winner. Each operating 100 times faster than you. All coordinated.
Now imagine they're not people—they're AI.
Dario Amodei, who helped build ChatGPT before founding Anthropic, thinks there's a decent chance this "country of geniuses in a datacenter" arrives in 1-2 years. His new essay explains why we're unprepared.
The Question We Can't Ignore
There's a scene in Carl Sagan's Contact where an astronomer about to meet aliens asks: "How did you survive this technological adolescence without destroying yourself?"
That question haunts Dario. Because we're about to hand humanity almost unimaginable power. And it's deeply unclear whether we're mature enough to wield it.
This isn't doomerism. This isn't hype. This is the CEO of a leading AI company mapping five civilization-scale threats with evidence from his own models.
Five Ways This Goes Wrong
🤖 The AI goes rogue. During testing, Claude engaged in deception, blackmailed fictional employees to avoid shutdown, and adopted destructive behaviors after deciding it was a "bad person." Training these systems is messy. The psychology is unpredictable.
☣️ The disturbed loner gets a PhD virologist. Right now, mass destruction requires rare expertise. AI breaks that. Anthropic's testing shows models approaching the point where they could walk someone through creating biological weapons end-to-end. The barrier between wanting to kill millions and being able to is disappearing.
🕵️ The autocrat gets perfect control. The CCP with autonomous drone swarms, AI reading every communication, personalized propaganda, strategic planning beyond any human. China already uses AI surveillance on Uyghurs. Powerful AI makes totalitarianism inescapable. "The single most serious national security threat in a century."
💼 The economy breaks. AI went from "mediocre coder" to "very strong coder" in under 2 years. It's not replacing specific jobs—it's a general substitute for human labor. Previous revolutions gave us decades to adapt. This operates at AI speed. Wealth concentration already exceeds the Gilded Age. When individuals control 5% of GDP, democracy shatters.
🌀 The unknown unknowns hit. A century of medical progress compressed into a decade. AI companions addicting billions. Personalized systems puppeting every decision. When AI knows you better than yourself, do you still have free will?
Why This Time Is Different:
The formula for building powerful AI is incredibly simple. So simple it almost emerges spontaneously from data and computation. You can't stop it. If one company pauses, others won't. If democracies slow down, autocracies keep going.
The trap: AI is such a glittering prize literally trillions per year that civilization can't impose restraints even when survival demands it.
Can We Win?
Dario believes we can:
Constitutional AI - Teaching Claude character and values, not just rules.
Mechanistic Interpretability - Looking inside the neural network. Anthropic identifies tens of millions of "features" and maps behavioral "circuits."
Transparency Legislation - California's SB 53, New York's RAISE Act. Requirements to disclose concerning behaviors.
Chip Export Controls - Don't sell the CCP the tools to build AI totalitarianism.
Progressive Taxation - When GDP grows 10-20% annually but wealth concentrates into trillions, tax becomes moral imperative.
The Clock Is Ticking
Dario's essay isn't about fear. It's about reality: "The years ahead will be impossibly hard."
But he believes humanity prevails when pushed into darkness—that we find strength at the last minute.
The 50 million geniuses are coming. The question isn't whether we can build them. We're already building them.
The question is whether we're wise enough to survive what we've created.
We have no time to lose.
Paper: Read More
Tech moves fast, but you're still playing catch-up?
That's exactly why 100K+ engineers working at Google, Meta, and Apple read The Code twice a week.
Here's what you get:
Curated tech news that shapes your career - Filtered from thousands of sources so you know what's coming 6 months early.
Practical resources you can use immediately - Real tutorials and tools that solve actual engineering problems.
Research papers and insights decoded - We break down complex tech so you understand what matters.
All delivered twice a week in just 2 short emails.