Part I: When Power Grows Faster Than Wisdom
I just finished reading two essays by Dario Amodei, CEO of Anthropic. The first is The Adolescence of Technology, and the second is an earlier piece titled Machines of Loving Grace. Both are thoughtful, disciplined, and refreshingly honest about where we are and what’s at stake. Amodei doesn’t dodge the hard stuff. He doesn’t sugarcoat the risks. And he doesn’t pretend that good intentions alone are enough. But he does believe AI could usher in a new era of human existence unlike anything we have seen before.
That combination of clarity, humility, and intellectual muscle is rare these days, especially in AI. Credit where it’s due.
Reading the essays back to back left me with two very different images rattling around in my head. If you know me, you know how my brain works. I tend to process the world through movies, TV shows, books, and music. That’s my decoder ring. And when Amodei talks about AI being in its “adolescent” phase, powerful, capable, impulsive, and not fully formed, my mind immediately snapped to two cultural poles.
One future is dark, chaotic, and terrifying.
The other is structured, principled, and maybe even hopeful.
Let’s start with the scary one.
The Adolescent With Godlike Powers
When I hear the phrase “adolescent AI,” my mind goes straight to a classic episode of The Twilight Zone called “It’s a Good Life.” Billy Mumy plays a young boy with unimaginable powers. He can read minds. He can change reality. He can erase you if you displease him. The town lives in terror, smiling on command, praising him constantly, afraid that even a wrong thought might get them banished to the cornfield.
That episode isn’t really about evil. It’s about unchecked power without maturity.
And that’s the danger Amodei is warning us about, whether he uses Twilight Zone metaphors or not. His core argument in The Adolescence of Technology is that our tools, and AI in particular, are growing up faster than our institutions, norms, and governance structures. Capability is compounding. Responsibility is lagging.
That gap is where bad things happen.
Amodei frames this moment as one where AI systems are no longer narrow tools. They reason, they plan, they act, and they increasingly operate with a degree of autonomy inside real systems. They influence decisions, workflows, markets, and soon, if we’re not careful, people’s lives at scale.
The adolescent metaphor matters because adolescence isn’t just about growth. It’s about volatility. Poor judgment. Overconfidence. Identity not yet fully formed. Now imagine that phase applied to systems with global reach and near-instantaneous execution.
That’s not science fiction anymore. That’s Tuesday.
To his credit, Amodei doesn’t hand-wave this away. He explicitly calls out the risks of concentrated power, both in the models themselves and in the hands of the people and companies controlling them. He raises uncomfortable but necessary questions about whether society is equipped to manage entities that may soon rival, or even surpass, the cognitive abilities of the smartest humans alive.
That’s not hype. That’s a sober assessment.
The Skynet Problem Without Saying Skynet
Here’s where my Brooklyn skepticism kicks in.
History tells us that when powerful technologies emerge, we rarely pause long enough to ask, “Should we?” We usually jump straight to, “How fast can we ship?” and “Who’s going to win the market?”
We’ve seen this movie before.
Social media optimized for engagement instead of truth.
Surveillance capitalism justified as personalization.
Algorithms shaping behavior with little accountability.
Now dial that up with systems that can reason, generate, and act on their own.
That’s the slippery slope toward what everyone jokes about but secretly fears: a benevolent, or not so benevolent, overlord deciding what’s best for us. Not because it’s evil, but because its objectives drift, its values misalign, or its creators lose control of the feedback loops.
Amodei is explicit that intent is not enough. Even well-meaning systems can cause harm if incentives, training data, or deployment contexts go sideways. Adolescents don’t need bad intentions to wreck a car. They just need the keys and a false sense of invincibility.
That’s the Twilight Zone future. The one none of us want.
Why This Essay Hit Differently
I’ve read a lot of AI essays. Most fall into one of two buckets.
Utopian hype telling us this will solve everything.
Dystopian doom warning that we’re all finished.
Amodei’s work sits in a much rarer third category: adult supervision.
He doesn’t deny the extraordinary upside. He also doesn’t pretend that scale and intelligence automatically produce wisdom. In Machines of Loving Grace, he lays philosophical groundwork, drawing from human values, moral philosophy, and long-term thinking, about what it would mean for machines to act in ways aligned with human flourishing.
Crucially, he acknowledges that alignment isn’t static. It’s not a one-time configuration. It’s an ongoing process that requires transparency, feedback, and restraint.
That matters, especially coming from the CEO of a frontier AI company.
Because let’s be honest, the real risk here isn’t just the models. It’s us. The builders. The executives. The boards. The investors. The incentives.
Amodei doesn’t exempt himself or his peers from that scrutiny. He explicitly calls out the responsibility of AI company leadership to act as stewards, not just operators. That kind of thinking is in short supply in Silicon Valley these days.
Standing at the Fork in the Road
So here we are.
On one side of the fork is adolescent AI with immense power, immature guardrails, and a world forced to smile politely while hoping nothing goes wrong.
On the other side is a future where we at least try to raise this technology with values, constraints, and humility, acknowledging that intelligence without wisdom is a liability, not an achievement.
Part I of this story is about the risk. The volatility. The Twilight Zone version of adolescence.
In Part II, I want to talk about the other image that kept popping into my head while reading Amodei’s essays. It’s the one involving rules, constitutions, and a certain set of laws dreamed up by a science fiction writer who understood this problem decades ago.
Because if AI really is a teenager, the most important question isn’t how smart it becomes.
It’s who’s raising it, and how.
