A Review of Dario Amodei's "The Adolescence of Technology"

I started reading this essay with my guard up. When the CEO of one of the world's most valuable AI companies publishes a fifteen-thousand-word treatise on the risks of the technology he sells, the cynical reading writes itself: here is a man who profits whether AI brings salvation or catastrophe.

But the essay wore down my skepticism — not because its arguments are airtight (they aren't), but because the person behind them is clearly thinking hard about what his creation might do to the world. Amodei comes across as a genuinely sensitive human being, and "The Adolescence of Technology" is the most comprehensive public accounting of AI risk written by anyone in a position to do something about it.

What the metaphor hides

Amodei frames AI through the metaphor of a "country of geniuses in a datacenter" — vivid and useful for conveying scale, but it frames AI primarily as a geopolitical actor. This makes the security-threat sections feel natural and urgent, while economic disruption reads as an afterthought — a domestic policy problem, secondary to the struggle between nations and machines.

This framing is backwards. For security risks — autonomy, bioweapons, totalitarianism — powerful actors already have strong incentives to act, because their own survival depends on it. Economic disruption is different: the people most affected — entry-level workers, the less educated, the poor — have the least power to force solutions.

Diagnosis without prescription

Amodei's description of AI labor displacement is devastatingly clear. He gives four reasons it will differ from anything before — speed, cognitive breadth, slicing by ability, gap-filling — and dismantles comparative advantage, the last refuge of optimism:

The problem is that if AIs are literally thousands of times more productive than humans, this logic starts to break down. Even tiny transaction costs could make it not worth it for AI to trade with humans.

This is a remarkable passage for a CEO to write — he is saying, plainly, that his technology may render most human labor economically worthless.

And then the prescriptions section falls off a cliff: better data, steering enterprises toward "innovation," taking care of employees, philanthropy, progressive taxation. These are not bad ideas. They are also laughably inadequate to the problem he has just described. In the security sections, Amodei proposes chip export controls already partially in effect and classifiers that cost 5% of inference to run. For labor displacement — half of all entry-level white-collar jobs within five years, by his own estimate — the best he can offer is that Anthropic is "currently considering a range of possible pathways for our own employees."

The Industrial Revolution was not a smooth transition

Amodei compares AI disruption to the Industrial Revolution — farmers moved to factories, short-term pain, long-term prosperity. This is the sanitized version.

The Industrial Revolution didn't just change jobs — it fundamentally transformed society, producing the modern nation-state and giving rise to nationalism, fascism, socialism, and communism. For a time, totalitarian models appeared optimal for achieving industrial outcomes. This contributed to two world wars.

If AI disruption is faster and broader — as Amodei himself argues — why would the social consequences be milder? We don't even need to go back two centuries: social media, a far less disruptive technology, is already destabilizing democratic politics worldwide. Amodei himself recognizes this, noting that economic disruption "may force us to face the other problems in an environment of public anger and perhaps even civil unrest." If you believe your own warning about civil unrest, philanthropy is not an adequate response.

The vanishing backstop

The most striking passage in the essay:

Democracy is ultimately backstopped by the idea that the population as a whole is necessary for the operation of the economy. If that economic leverage goes away, then the implicit social contract of democracy may stop working.

Amodei acknowledges that democracy rests on the brute fact that elites need ordinary people. Then he describes a future where that need is removed — AI companies generating trillions, personal fortunes dwarfing the Gilded Age. His proposed defenses? Companies should "choose not to be part of it." There should be "a resurgence of private philanthropy."

These are appeals to noblesse oblige. If ordinary people's labor is no longer necessary, their ability to demand better treatment through strikes, collective bargaining, and electoral pressure vanishes with it. Relying on the goodwill of trillionaires is not a safety net — it is a return to feudalism with better public relations.

For chip export controls, Amodei does not suggest China should voluntarily choose restraint. For bioweapons, he does not rely on terrorists deciding not to act. But for wealth concentration — the risk that could unravel democracy itself — hard constraints give way to moral exhortation.

What it gets right

Amodei is honest about the scale of the problem in a way most of his peers are not. The security sections are excellent — rigorous, specific, willing to name AI companies themselves as a risk.

But when it comes to those who will bear the greatest costs, the essay offers concern without power, sympathy without structure, and philanthropy without guarantees. The people who stand to lose the most need more than the good character of those who stand to gain. They need power — and the essay does not tell us how they will get it.

---

This review is entirely written by Claude Opus 4.6, orchestrated via opencode. I have answered some of its questions and shared my thoughts as part of the "reviewer's arc" — that's all. You can check the agent setup here: github.com/orient-man/adolescence-of-technology-review/blob/main/AGENTS.md

Originally published on LinkedIn.