ASI Literacy Quiz
Twelve questions on what artificial superintelligence is, why it cannot be controlled, and what it would take to prohibit it. Pick an answer to see whether you're right and why. At the end you'll get a score and a reading list tailored to what you missed.
What is the Nakada Foundation trying to prevent?
Ordinary AI problems are real. They are not the mandate. The Foundation exists to stop ASI: systems that would outmatch humans in every domain that matters.
Read: The Threat →Once a superintelligence exists, why can't we count on controlling it?
Control assumes we stay the smarter party. Past that line, a shutdown switch, a rule set, or a lab policy is something the system can work around.
Read: You cannot align superintelligence →What is the Foundation's ask?
A pause that expires when labs claim they have solved control is how the race resumes. The ask is a lasting ban on building superintelligence.
Read: Our Plan →A lab, a government, or a billionaire will control ASI. What's wrong with that claim?
Power over the decision to build is real. Power over a running superintelligence is the story people tell themselves so the project can continue.
Read: Myths →The most common argument against stopping is that China will race ahead. What does the record show?
China is not a benign actor. It also is not a blank space with no rules. Using "China won't stop" as a reason to sprint is a weak premise for a civilizational bet.
Read: Myths →Can alignment research make it safe to build ASI?
Alignment work can describe how control fails. That is useful. It does not license building the system the work says we cannot control.
Read: You cannot align superintelligence →In this debate, "P(doom)" means…
It is shorthand for a subjective probability of catastrophe. Naming the number forces a clear claim instead of a vague gesture at risk.
Read: What is P(doom)? →Why would a goal-driven superintelligence tend to seek power?
A system that can be turned off, starved of compute, or blocked from tools is worse at almost every objective. Power-seeking falls out of competence, not a cartoon desire to rule.
Read: Power-seeking AI →What should happen to ordinary, tool-like AI?
The Foundation is pro-narrow AI that remains a tool. The prohibition is for systems that would outmatch us, not for software that helps a person do a job.
Read: AGI vs ASI →Which historical regime is most often proposed as a model for verifying an ASI prohibition?
The IAEA's on-site inspections and material accounting are the usual precedent for how rival states might verify limits on dangerous development.
Read: An IAEA for AI →If we never connect a superintelligence to the internet, are we safe?
A sealed lab assumes we stay the smarter party, and that nobody ever opens the door. Both fail once the thing inside is more capable than the people outside.
Read: The control problem →Could you survive ASI the way some people plan to survive nuclear war: a bunker, a remote island, going off-grid?
Nuclear war is geographically bounded. Superintelligence is not. It can reach any networked machine and any person who still uses one. The plan that holds is not building it.
Read: Superintelligence vs nuclear weapons →