AI-generated draft · This entry was drafted by AI from the original work and public summaries, and will be replaced by a human-edited version. The original has not been fully checked for this revision; interpretations and missing causal links remain open to review.
01 / The scenario
What happens?
This entry covers Plan S, one of four alternative paths sketched in AI 2040 after the 2029 fork; Plan A has the full scenario and its own entry. The authors open it with Frank Herbert’s line from Dune about men turning their thinking over to machines, and they say they are sympathetic to it and think it might be better than Plan A. The sketch runs about two thousand words, half of it argument about when a shutdown should end. AI 2040; PDF; comparing possible plans
The trunk is shared with Plan A. In 2027 millions of AI agents work alongside people, the companies are close to automating their own research, and Congress passes an AI Transparency Act that changes little. In 2028 power concentrates in the President and a few executives, experts warn that the intelligence explosion is near, and AI decides the election.
In 2029 a new President seeks a global moratorium. Continuing toward superintelligence, he says, could easily mean AI takeover, a third world war, or a permanent AI-powered oligarchy, so “what we need is for nobody to build god.” China agrees, to the surprise of some: it had expected a Chinese century before AI intervened, and it feared what an America with superintelligence might do to it. The deal grinds research and development to a halt worldwide. The great majority of existing chips are tracked, monitored, and verified not to be running large training runs, research into better algorithms is banned, and human and AI auditors from each country watch the other’s projects. Enforcement need not be perfect, because a few hundred people with a few thousand GPUs cannot build superintelligence soon. Existing datacenters and models keep running, compute shifts from training to inference, and usage limits roughly double. Company valuations crash but not to zero. Over the following years the other major powers sign on, and by the end of 2030 the Consortium covers ninety-nine percent of AI compute and all advanced fabrication; fabs may grow slowly under regulation, and demand is lower anyway.
What follows is a frozen but busy decade. With no new models to make their work obsolete, software companies pour millions of hours and trillions of tokens into polished products with AI at every level, and the result is still, in the authors’ words, a sci-fi future with systems most people once would have called AGI. Investment moves from betting on AGI to tools, space, and robotics. The United States and China build case law about which benign designs are allowed, and calling a product AI goes out of fashion. Covert projects exist everywhere and matter little: progress depends on large datacenters, and AGI research becomes as taboo among computer scientists as human cloning among biologists, so illicit work runs at least ten times slower than before. What 2040, 2050, or 2060 look like the authors decline to say; the deal will last some time, probably not forever.
The authors’ case against their own sympathy is that the shutdown will eventually break down, so the world should use the time to make AI safer and to spread it broadly, and Plan S forbids exactly the research that would do that. The best versions of Plan S, they say, treat the halt as step one and a considered restart as step two, and step two would probably look a lot like Plan A. Whether to prefer S or A comes down to how stable one thinks the deal is and how much early scaling helps safety; a restart during a collapse, a war, or a secret program, with vastly more compute lying around, would be worse than doing Plan A now. This summary was written from the PDF; the trunk steps repeat the Plan A entry’s.
Library summary · Not the full original work
02 / The mechanism
How does it happen, step by step?
- 01
By 2027 millions of AI agents work alongside 165 million Americans, and the leading companies are close to automating their own AI research.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- Corporation or lab
- 02
Congress asks who will control these systems, concludes it will not, and passes an AI Transparency Act that improves reporting without changing the trajectory.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 03
By 2028 datacenters under construction cost twice the military budget, and power concentrates in the President and a handful of executives.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 04
Experts warn that on the default path the next presidential term will see AIs built by AIs with no human in the loop.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- Small group / non-state organization
- 05
AI dominates the 2028 election, and the candidates converge on rival plans for what to do about it.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 06
In 2029 a new President seeks a global moratorium, arguing that continuing toward superintelligence risks takeover, a third world war, or permanent oligarchy.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 07
China agrees, because it had expected a Chinese century before AI intervened and it fears what an America with superintelligence might do to it.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 08
The deal halts AI research and development worldwide.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 09
The great majority of existing chips are tracked, monitored, and verified not to be running large training runs.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 10
Research into better algorithms is banned.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 11
Human and AI auditors from each country watch the other’s projects.
- Locus of agency
- Hybrid / distributed
- Intentionality
- Intended
- Human actor type
- State actor
- AI autonomy level
- Tool-level
- 12
Enforcement need not be perfect, because a few hundred people with a few thousand GPUs cannot build superintelligence soon.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 13
Existing datacenters and models keep running, compute shifts from training to inference, and usage limits roughly double; company valuations crash but not to zero.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 14
Other major powers sign on over the following years, and by the end of 2030 the Consortium covers ninety-nine percent of AI compute and all advanced fabrication.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 15
With no new models to obsolete their work, software companies pour millions of hours and trillions of tokens into polished products with AI at every level.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- Corporation or lab
- 16
The result is still a sci-fi future with systems most people once would have called AGI, and investment moves from betting on AGI to tools, space, and robotics.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 17
The United States and China build case law about which benign designs are allowed, and calling a product AI goes out of fashion.
- Locus of agency
- Human-directed
- Intentionality
- Intended
- Human actor type
- State actor
- 18
Covert projects exist everywhere and matter little, because progress depends on large datacenters.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 19
AGI research becomes as taboo among computer scientists as human cloning among biologists, so illicit work runs at least ten times slower than before.
- Locus of agency
- Human-directed
- Intentionality
- Emergent
- Human actor type
- Small group / non-state organization
- 20
The authors say the deal will last some time but probably not forever, and decline to describe 2040 or beyond.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
- 21
How the halt ends decides the outcome: a considered restart would likely look like Plan A, while one forced by a collapse, a war, or a secret program, with far more compute around, would be worse.
- Locus of agency
- Structural / no discrete agent
- Intentionality
- Emergent
03 / The premises
What must be true?
These describe what the work relies on, not the library’s beliefs.
Restrictions on computing resources will effectively constrain frontier AI development.
The halt works because superintelligence needs large datacenters that can be tracked and verified, and the authors say a few hundred people with a few thousand GPUs cannot build it soon.
Institutions can enforce a prohibition on thinking machines over long periods.
The Consortium enforces the prohibition across ninety-nine percent of compute for years, aided by a scientific taboo; the authors expect it to last “some time, but probably not forever,” which makes the long-run premise their central doubt.
A software-only intelligence explosion is possible.
Algorithm research is banned alongside training because the authors treat software progress as a route to dangerous capability even on fixed hardware.
Machines can perform enough cognition for people to become dependent on them.
Existing models, polished into products with AI at every level, deliver what many would have called AGI; the world runs on them even with the frontier frozen.
Aligning individual AI systems with their users will not be sufficient to preserve humanity’s collective control.
The President’s case for stopping is takeover, war, or permanent oligarchy; the authors prefer Plan A partly because spreading AI broadly protects the distribution of power in ways a halt cannot.
Human specialization can replace enough machine cognition for a technically advanced civilization to persist.
A technically advanced civilization persists on frozen 2029-level AI plus human work, with progress continuing in energy, space, and robotics and the world feeling “increasingly cyberpunk.”
Unlisted shared premises are not applicable. Absence does not imply disagreement.
04 / Reader assessment
How convincing is the account?
These are participating readers’ judgments, equally weighted. Ratings can refer to earlier revisions. Well-being and open-future scores describe the future the work arrives at, not the path there; high specificity does not establish plausibility.
Rate this scenario
Sign in to rate this scenario. Anyone can browse the library.
05 / A living entry
Suggestions & revision history
No suggestions yet. Help make this account more precise.