Meta's CEO just compared AI risk to medieval folklore while his company races to build the very thing he claims doesn't exist yet.
The Summary
- Mark Zuckerberg published a 7,000-word manifesto arguing that fears of "superintelligence" are overblown, comparing AI doom scenarios to worrying about dragons
- His core claim: current AI concerns are premature because we're nowhere near artificial general intelligence, and regulatory fear-mongering will only help incumbents
- The timing matters: Meta is spending $65 billion on AI infrastructure in 2025 while simultaneously arguing the thing they're building isn't dangerous enough to regulate
The Signal
Zuckerberg's dragon analogy does something clever and disingenuous at once. He's right that superintelligence doesn't exist today. He's also building the systems that might create it while arguing we shouldn't prepare for that possibility. It's like SpaceX saying "Mars colonies are science fiction" while laying the groundwork for exactly that.
The manifesto arrives as Meta pours unprecedented capital into AI development. The company is on track to spend more on compute infrastructure this year than most countries spend on their entire technology sectors. That's not the behavior of someone who thinks the technology is overhyped.
"You don't regulate dragons when you're still figuring out how to make fire."
But here's what Zuckerberg gets wrong about the dragon metaphor. Medieval people didn't worry about dragons because dragons were real threats. They worried about dragons because powerful, uncontrollable forces were real, and dragons were the best available mental model. Replace "dragon" with "famine" or "plague" or "invasion" and the concern makes perfect sense.
The modern equivalent isn't superintelligence as imagined in science fiction. It's the actual, observable pattern where AI systems become too complex for their creators to fully predict or control. We're already seeing this with large language models producing unexpected capabilities their designers didn't explicitly program.
Key contradictions in the manifesto:
- Claims AI risk is "premature" while racing to build more capable systems
- Argues against regulation while Meta benefits massively from regulatory capture via existing content moderation laws
- Dismisses long-term thinking while making multi-decade infrastructure bets
Meta's position makes strategic sense if you squint. Downplay existential risk, inflate near-term utility, and you create regulatory breathing room while competitors get hamstrung by safety protocols. It's the same playbook Big Tech used with privacy: "Don't regulate us now, we'll figure it out, trust us, regulation only helps the incumbents."
Except Meta is the incumbent. They have the compute, the data, the distribution. Zuckerberg positioning Meta as the scrappy underdog fighting against regulatory capture would be laughable if it weren't so effective as a political move.
The Implication
The dragon debate misses the point entirely. The question isn't whether superintelligence arrives next year or next century. The question is whether we build guardrails while we still understand the systems we're creating, or whether we wait until we've lost the ability to course-correct.
Watch what they build, not what they say about building it. Meta's infrastructure spending tells you everything you need to know about how seriously they take AI's potential. The manifesto is a PR document designed to buy time and space. The real signal is in the capex.