superintelligence
/ˌsuː.pər.ɪnˈtel.ɪ.dʒəns/
RareA hypothetical AI that far exceeds the cognitive abilities of any human across every domain.
Often treated as a more extreme endpoint than AGI. In serious discourse refers to specific scenarios analyzed by Nick Bostrom. Casually used to describe any AI that feels impossibly capable, sometimes sarcastically.
| # | Example | What it means | Source |
|---|---|---|---|
| 01 | Bostrom's book on superintelligence genuinely changed how I think about AI risk. Read it before you dismiss this stuff. | The speaker credits the book with shifting their view on existential AI risk. | reddit, 2023 |
| 02 | People keep saying current LLMs are superintelligence. They are autocomplete. Not the same. | Pushing back on overclaiming about the capabilities of existing AI systems. | x, 2024 |
| 03 | A misaligned superintelligence would not be evil. It would just have goals that do not include you. | Clarifying that the risk is not malice but indifference from a system optimizing for something else. | reddit, 2023 |
| 04 | Building superintelligence before solving alignment is like building a nuclear reactor and figuring out containment later. | Arguing that capability without safety is dangerous by analogy. | x, 2024 |
Use it when
- Describing the hypothetical scenario where AI capability far exceeds human-level intelligence.
- Discussing the risk that a superintelligent system would be uncontrollable.
- Referencing Bostrom's analysis of superintelligence scenarios.
Do not use it when
- You just mean a very capable AI model that impresses you.
- You mean AGI in the sense of human-level general intelligence.Use “agi” instead.
Superintelligence vs AGI
superintelligence
Superintelligence far exceeds human-level intelligence across all domains, potentially incomprehensibly so.
agi
AGI matches human-level general intelligence but does not necessarily exceed it completely.
Origin and context
Concept developed and popularized by Oxford philosopher Nick Bostrom, particularly in his 2014 book 'Superintelligence: Paths, Dangers, Strategies'. Earlier roots in work by I.J. Good on the intelligence explosion concept from the 1960s.
- First seen
- 1960s
- Confidence
- high
Cite this page
Use this stable URL to reference the entry.
https://slang.fyi/superintelligence