🤔 AI Is Going Through Puberty. And That's a Concerning Moment.
Dario Amodei (the mind behind Anthropic) recently published an essay called "The Adolescence of Technology," and it offers a precise metaphor:
Modern technology — especially AI — is behaving like a teenager. Plenty of energy. Not enough self-awareness or understanding of consequences.
"We are very close to 'powerful AI': systems that can autonomously set sub-goals, act independently, work for months, and make decisions faster than humans. And yes, this may happen not 'someday' — but within the next few years."
Here are five risks Amodei identifies:
1. Autonomy without understanding consequences AI does exactly what you asked. The problem: we don't always understand what we're asking. The system optimizes the formal goal — not the intended one. Classic: "maximize the result" with no understanding of what gets broken along the way. (Sound familiar to anyone who's used Lovable? It fulfills the next prompt — but in its own interpretation, often removing things you didn't ask it to touch 🤷)
2. Amplifying small actors — the negative scenario Before, doing something dangerous required money, a team, and infrastructure. Now — a laptop, a model, and a bad mood. One person or small group can now have capabilities previously available only to states or corporations. Models have content filters, but with enough ingenuity, a human can combine information to synthesize something harmful.
3. The perfect tool for control AI scales not just benefit — but pressure. Surveillance, manipulation, automated repression, social scoring — all of it suddenly becomes cheap and efficient. Technology designed to "help" easily becomes technology of suppression. And since it's simple and cheap, it will be deployed.
4. The economy can't keep up AI boosts productivity — but unevenly. Some people become 10× more effective; others become "redundant by algorithm." And you need 1,000× fewer of the first group than the second. Where are you in that equation? Sharp inequality growth, devaluation of professions, and social instability arrive faster than we'll understand them.
5. Social side effects AI affects not just the labor market, but how we think, communicate, and trust. We're not just delegating tasks — we're delegating judgment. We're rapidly losing the habit of doubting, verifying, and deciding for ourselves. And what about children who may never even internalize the concept of "analyze and decide"?
The conclusion without moralizing:
The adolescence of AI is dangerous not because of malicious intent — but because of AI overestimating its own capabilities and us underestimating them.
❤