I wrote this for a general audience and cross-posted it here for feedback. The argument compresses a lot, deliberately. I'm interested in pushback on the timelines and on the human-overlord scenario.

IF AI development goes as well as it possibly can:

1. We will have a new overlord within a decade. It may be human.
2. Your preschool kids (or grandkids) will never go to college, never work, and never have wrinkles. They’ll have the bodies of young athletes, the minds of the wisest person, and centuries to enjoy both. They will live in a palatial human equivalent of a national park. Or a zoo. Nothing they do will ever matter.
3. You will lose your job in the next seven years. And never get a new one.
4. The transition will not be smooth. The primary targets are obvious. Stay upwind.
5. There is nothing anyone can do about this. Relax and enjoy the show. No one has ever seen it. No one ever will again.
6. There will not be a 22nd century.

IF AI development goes poorly:

Your preschool kids or grandkids will never make it to wrinkles…

Why This Is Inevitable

Right now, a thousand key people are making AI smarter. They’re the senior researchers and engineers at places like Anthropic, OpenAI and Google DeepMind. They are the only humans who understand, at a granular level, how to build more capable AI systems.

Within five years, AI will be as smart as they are. Maybe smarter.

What happens next is simple math: Instead of a thousand humans making AI better, you’ll have millions of AI researchers. Each one as capable as the best humans. Working 24/7. Never tired. Never sick. Never needing sleep.

Those millions will build something smarter than themselves. That next generation builds something even smarter. And so on.

The jump from “as smart as humans” to “vastly smarter than humans” won’t take decades. It could take months. The years between 2040 and 2100 will see millennia worth of technological progress. There won’t be a 22nd century as we understand centuries.

AI matches the top researchers in five years. The cascade through everything else takes a few years longer. Robots have to catch up. Bureaucrats have to be fired. By 2033, your job is gone too.

No country will stop this. If America pauses, China won’t. If both pause, someone else races ahead. The first to reach superintelligence wins everything. The logic is identical to nuclear weapons development, except the prize is not parity but permanent supremacy.

A recent academic analysis of AI geopolitics reaches the same conclusions: once AI can improve itself, the race becomes winner-take-all, and cooperation becomes nearly impossible. The paper models exactly how this plays out and why no one can prevent it.

Here’s the part that should terrify you: The winner might be human.

Someone will control this godlike intelligence first. Maybe a tech CEO. Maybe a president. Maybe a dictator. Look at every person in history who gained absolute power. Now give them a billion times more.

Their first order will be obvious:

Sabotage every other AI lab. Make it look natural. Feed them false progress so they never know they’ve been stopped.

One entity gets god. Everyone else gets a leash.

What comes after that? Nobody knows. But it’s likely binary. Superintelligence could make every human life long and amazing, or it could end us. It depends entirely on who holds the remote.

The most important events in human history will all happen in the next 20 years.