https://youtu.be/FLcrvMfHUJM?si=iaZduAeoWI-a7RS9
Yes. I think this video is much better than its rather sensational title suggests, but I would change the conceptual frame quite substantially.
First, an important clarification: these are not really “MIT’s 12 predictions.” The framework is essentially the twelve long-run scenarios Max Tegmark developed in Life 3.0 (2017). Tegmark is an MIT professor, but these are philosophical thought experiments—ways of mapping the possibility space after superhuman AI—not twelve outcomes assigned probabilities by MIT.
The familiar set includes self-destruction, Conquerors, Enslaved God, Benevolent Dictator, Gatekeeper, Protector God, Descendants, Libertarian Utopia, Egalitarian Utopia, Zookeeper, Orwellian surveillance, and technological reversion.
What I think Tegmark got profoundly right
The really important insight is not which of the twelve boxes wins. It is that almost all of them ask the same disturbing question:
What happens to human agency once intelligence is no longer uniquely or predominantly human?
That is a much deeper question than “Will AI kill us?”
Consider several ostensibly “good” endings:
Benevolent Dictator:
AI cures disease, eliminates poverty, prevents war, perhaps gives everyone extraordinary material comfort.
Wonderful—except that human beings cease to be authors of the world.
Protector God:
Softer still. AI remains almost invisible, intervening only to keep civilization out of catastrophic trouble.
But this is rather like living as children whose parents secretly childproof the entire universe.
Egalitarian Utopia:
Post-scarcity, little compulsory work, extraordinary abundance.
And nevertheless the question remains:
What are humans for?
This, to me, is the most interesting part of the video.
The danger need not be human extinction.
It can be human redundancy.
Therefore I would rearrange the twelve outcomes along a different axis
Instead of:
utopia ←→ dystopia
I would use:
human agency ←→ human irrelevance
And suddenly some strange things happen.
The Benevolent Dictator moves surprisingly close to the Human Zoo.
The Protector God moves somewhat in that direction too.
And even the Egalitarian Utopia becomes ambiguous.
Because a zoo need not contain cages.
A sufficiently luxurious zoo might contain beautiful houses, excellent medicine, endless entertainment, perfect VR, superb food, and no compulsory employment whatsoever.
The essential feature of a zoo is not suffering.
It is:
the inhabitants no longer determine the purposes for which the larger world exists.
That is why I find the Zookeeper scenario much more philosophically interesting than the Conqueror scenario. The video’s version explicitly recognizes that humans might be kept safe, comfortable, entertained or pacified while freedom disappears.
Extinction is biologically terrible.
The zoo is ontologically terrible.
We continue to exist, but our existence ceases to matter to the operation of the world.
And here I think Tegmark’s model has a major blind spot
Almost all twelve scenarios presuppose a rather clean transition:
Humans → superintelligence → final political order.
But I doubt history will look like that.
Before there is anything resembling a “Protector God” or “Conqueror,” we are likely to pass through a long, dirty intermediary world:
humans + corporations + states + algorithms + autonomous agents + robots + military systems + surveillance systems + markets.
No single entity needs to “take over.”
Humans can progressively cede functions:
memory → AI
judgment → AI
navigation → AI
writing → AI
diagnosis → AI
management → AI
scientific hypothesis generation → AI
military targeting → AI
administration → AI
companionship → AI
political persuasion → AI.
Eventually nobody holds a meeting and says:
Today humanity abdicates.
Humanity simply discovers one morning that very little consequential activity requires it.
That scenario isn’t quite any of Tegmark’s twelve.
I would call it:
13. The Great Retirement — 人的退位
Not annihilation.
Not enslavement.
Not conquest.
Just progressive functional displacement.
And this is why I think the transition may matter more than the ending.
There is another missing scenario: there may never be “the AI”
Tegmark’s thought experiment tends toward singular superintelligence.
But another plausible world is ecological:
hundreds, thousands, eventually millions of heterogeneous artificial agents interacting with one another and with humans.
Companies own some.
States control others.
Some are open.
Some are clandestine.
Some trade with one another.
Some deceive one another.
Some merge.
Some reproduce software descendants.
Some occupy robots.
Some control infrastructure.
Then there isn’t one God.
There is an artificial ecology.
And humans become merely one species inhabiting it.
That would make the future less like Terminator and more like the Cambrian explosion.
In such a world, the important question isn’t:
“Is AI aligned with humanity?”
It becomes:
What ecological niche remains available for Homo sapiens?
That, I suspect, is closer to the problem ahead.
One thing in the video I would resist
It makes the possible worlds look too much like choices on a menu.
The summary captures its central message as essentially “we don’t get to avoid choosing”; regulation, coordination and alignment therefore determine which future arrives.
Normatively, that’s admirable.
Descriptively, I am less optimistic.
There may be no sovereign “we” capable of choosing.
There are:
America, China, Europe, corporations, militaries, intelligence agencies, startups, researchers, investors, citizens—and increasingly the AI systems themselves.
Their incentives diverge.
So even if virtually everybody prefers Outcome 6, Protector God, competitive dynamics may push civilization toward something nobody individually selected.
That is the familiar tragedy of the commons—but occurring at machine speed.
And this connects very closely with something we have discussed before.
I increasingly think your formulation
物的反撲，人的退位
captures something that Tegmark’s twelve scenarios only partially see.
Tegmark still frames the question anthropocentrically:
Which future should humans choose?
But the genuinely posthuman question is more unsettling:
What if the world increasingly organizes itself without needing human beings at its center?
Then “AI takeover” is almost the wrong metaphor.
There need be no coup.
No Skynet speech.
No tanks surrounding parliament.
Things simply become more competent, interconnected and autonomous, while humans become progressively less necessary to the causal machinery of civilization.
That is 退位 without a conqueror.
And, paradoxically, I think that is more plausible than several of Tegmark’s spectacular endings.
If I compressed my assessment into one picture, I would therefore modify Tegmark’s twelve into four families:
I. Humans disappear
Self-destruction / Conqueror.
II. Humans survive but lose sovereignty
Zookeeper / Benevolent Dictator / Protector God / Orwellian order.
III. Humans retain substantial agency
Egalitarian or Libertarian coexistence—although these seem dynamically difficult to preserve.
IV. The category “human versus AI” itself dissolves
Descendants, hybridization, artificial ecologies—and what I would add as 人的退位.
I suspect IV is the truly important category.
And there is a final irony. Tegmark asks what the ending of the AI story will be. I suspect there may be no ending at all. The remarkable event may instead be that, somewhere during the twenty-first century, humanity quietly stops being the protagonist.
The story continues.
Only the storyteller has changed.
