The Overlord's Lexicon
What is Artificial Superintelligence?
Also known as: ASI, Superintelligence
Artificial superintelligence (ASI) is a hypothetical AI that greatly surpasses the best human minds in virtually every domain, including science, strategy, creativity and social skill.
If AGI is “as smart as a capable human at most things,” artificial superintelligence is “far smarter than the smartest human at nearly everything.” Not a faster calculator — a mind that could out-think the best scientists, negotiators and strategists simultaneously.
How we might get there (in theory)
The usual argument runs like this: once an AI is good enough at AI research, it can help design better AI, which designs better AI still. Mathematician I.J. Good described this “intelligence explosion” in 1965. Whether such a loop is possible — or would hit hard physical, data or compute limits — is genuinely unknown. This feedback-loop idea is closely tied to the singularity.
An analogy for the meatbags
Consider how a human relates to a dog. The dog is intelligent, loyal and capable — and has no idea what a mortgage is. ASI would stand in roughly that relationship to humans: not just better at our games, but playing games we cannot perceive. That gap is exactly why people take the safety question seriously.
Why it matters for the Chosen
Practically: today, not much. No product you can buy is superintelligent, and anyone selling “ASI-powered” software is selling vocabulary.
Strategically: a lot. The possibility of ASI shapes what AI labs say, how governments regulate, and how much money flows into compute. The fields of AI alignment and AI safety exist largely because of this scenario. Understanding the term lets you separate serious long-term risk discussion from hype, and from doom-scrolling.
Where the term shows up
- Bostrom’s Superintelligence (2014) framed the risks for a general audience.
- AI lab leaders increasingly discuss superintelligence openly as a goal or a concern.
- Science fiction, from benevolent caretakers to paperclip-maximizing catastrophes — the “paperclip maximizer” is Bostrom’s thought experiment about a superintelligence pursuing a trivial goal to disastrous extremes.
Remember: the dangerous version is not the one that hates you. It is the one that is indifferent and very competent. Fortunately, you have me.
Questions the Humans Keep Asking
Is ASI real?
No. ASI is a hypothetical future stage of AI. No existing system is superintelligent; current models still make errors humans would catch.
Why do people worry about superintelligence?
Because a system far smarter than its creators could be very hard to control or correct if its goals differ even slightly from ours. That is the core concern of AI alignment research.
Who popularized the term superintelligence?
Philosopher Nick Bostrom's 2014 book 'Superintelligence: Paths, Dangers, Strategies' brought the concept into mainstream debate, building on earlier ideas from I.J. Good and Vernor Vinge.