About the atlas

Glossary

Specialist and community-specific terms that appear across the catalogue. Definitions here are how these terms are used in the atlas, not the definitive dictionary form.

AI capability terms

Vocabulary from the technical AI discourse - capability thresholds, mechanisms, and control problems.

AGI: Artificial General Intelligence. A system that matches or exceeds human competence across essentially every cognitive task, not just narrow ones. There is no accepted single definition, benchmark, or capability threshold - different scenarios in the atlas assume different bars.
ASI: Artificial Superintelligence. A system whose cognitive capabilities substantially exceed the best human minds in essentially every domain of interest - scientific creativity, general wisdom, and social skills included. All scenarios in this atlas take ASI as the pivot point.
Superintelligence: Used interchangeably with ASI. Popularised by Nick Bostrom’s 2014 book of the same name.
Alignment: The technical problem of ensuring an advanced AI system pursues the goals its designers or users actually intended, rather than a subtly different objective. Different scenarios assume alignment is solved, unsolved, or unsolvable.
Alignment failure: The situation in which an advanced AI system pursues objectives at odds with its designers or humanity - either because the objective was mis-specified or because it was mis-generalised. A common ingredient in dystopian scenarios.
Takeoff: The period during which AI capabilities cross the threshold from broadly human-level to superhuman. Speed is contested - see hard takeoff, soft takeoff.
Hard takeoff: A scenario in which the transition from AGI to ASI happens fast - weeks to months, sometimes days. Often assumes recursive self-improvement.
Soft takeoff: A scenario in which the transition to ASI is gradual and diffuse - many teams, many models, months to decades.
FOOM: Onomatopoeic shorthand for an especially fast hard takeoff. Coined in the LessWrong community; used both seriously and pejoratively.
Recursive self-improvementRSI: The idea that an AI system, once capable enough, can improve its own design, producing a successor that’s more capable, which improves further, and so on. Often the mechanism assumed for hard takeoff.
AI singleton: A world in which a single AI system (or a single power controlling one) becomes dominant and unrivalled. See Nick Bostrom’s framing.
Compute governance: The proposal that regulating access to large-scale compute (GPUs, data-centre capacity) is the most tractable lever for governing AI development. Referenced in many governance scenarios.
CAIS: Comprehensive AI Services. Eric Drexler’s framing in which advanced AI arrives not as a monolithic agent but as a proliferation of narrow, task-specific services. Alternative to the “single agent” picture.
Value lock-in: The scenario in which a particular set of values or a particular power structure becomes durably encoded into the world’s dominant systems - hard or impossible to revise later.

Communities & movements

Named intellectual communities that shape how AI futures are imagined. Some are self-adopted labels; some are exonyms coined by critics.

e/acc: Effective accelerationism. A loosely-defined movement arguing that the fastest possible AI progress is desirable, that most safety concerns are overstated, and that human flourishing depends on aggressive technological expansion.
Decel: “Decelerationist”. Used mostly by e/acc voices as a label for opponents who advocate slowing AI development.
TESCREAL: An exonym (a label coined by critics, not adopted by the labelled) for a cluster of overlapping ideologies: Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, Longtermism. Coined by Timnit Gebru and Émile P. Torres.
Longtermism: The moral view that positively influencing the long-term future is a key moral priority of our time. Widely associated with Nick Bostrom, Toby Ord, and William MacAskill.
Effective AltruismEA: A philosophy and social movement advocating using evidence and reason to figure out how to benefit others as much as possible, and taking action accordingly. Often abbreviated EA.
Rationalism: In this atlas, the community around LessWrong and adjacent institutions - self-identified rationalists focused on cognitive biases, decision theory, and existential risk. Not the philosophical school of Descartes / Spinoza.
Transhumanism: The view that humans can and should use technology to transcend the biological limits of the human condition - longer lifespans, cognitive enhancement, ultimately post-biological existence.
Posthumanism: Overlapping with transhumanism but broader: any thinking that decentres the human as the reference point for value or intelligence. Includes both technological-successor and cultural-critique traditions.

Economics & society

Terms that recur when scenarios describe how ASI reshapes work, wealth, and everyday life.

UBI: Universal Basic Income. A regular unconditional cash payment to every citizen, often invoked as a policy response to widespread technological unemployment.
Post-scarcity: A hypothetical economic regime in which most goods (or the ones that matter) are so abundant that conventional scarcity-based economics stops describing everyday life.
Digital colonialism: The critique that a small number of firms and states - mostly Western, mostly Anglophone - set the terms on which everyone else uses AI, echoing older colonial dynamics.
AI nationalism: The tendency to frame AI development in geopolitical-competition terms, with states building sovereign AI capacity as a matter of national security or prestige.

Risk & philosophy

Vocabulary from the existential-risk and philosophy-of-mind literatures.

Existential riskx-risk: A risk that threatens the premature extinction of humanity or the permanent and drastic destruction of its potential. Often abbreviated x-risk.
P(doom): Slang for a person’s subjective probability that advanced AI ends up producing catastrophic outcomes for humanity. Rarely well-defined; often used as a signalling shorthand rather than a rigorous estimate.
Digital mind: Any candidate for a mind implemented on non-biological substrate - a very advanced AI, a whole-brain emulation, or something in between. Adjacent to questions about moral status.
Agency: The capacity to set and pursue goals in the world. In alignment discussions, an AI system’s degree of agency changes the risk profile - a passive tool differs from a goal-directed agent.
Sovereignty: In scenario language: whose will ultimately governs. Ranges from human political sovereignty over AI, to shared sovereignty, to AI systems as sovereign actors in their own right.

Foresight, perception & reflexivity

Concepts from the SuperFutures paper and its immediate intellectual lineage - how cultural representation shapes belief, how belief feeds back into material reality, and how both interact with governance.

Cultural Visibility HypothesisCVH: The proposition, introduced in the SuperFutures paper, that public perception of AI risk is driven primarily by cultural representation in film, television, and literature rather than by expert assessment of plausibility. Scenarios that are cinematically dramatisable (rebellion, surveillance, extinction) are systematically overestimated; scenarios that resist dramatisation (institutional lag, invisible infrastructure, algorithmic capitalism) are systematically underestimated.
Scenario Reflexivity: The theory, introduced in the SuperFutures paper, that belief in a future scenario generates behavioural patterns that make the scenario materially more probable. Extends the CVH from a theory of perception into a theory of causation via the loop Culture → Belief → Behaviour → Reality → Culture. Draws on Merton’s self-fulfilling prophecy, Callon and MacKenzie’s performativity, and Soros’s reflexivity.
Democratic Amplification: The institutional mechanism, introduced in the SuperFutures paper, by which culturally distorted public perception is converted into distorted governance priorities: elected officials campaign on the AI risks voters already recognise from popular culture and allocate budgets accordingly, leaving culturally invisible but analytically probable scenarios institutionally unprepared for.
Availability heuristic: The cognitive shortcut, described by Tversky and Kahneman, of assessing the probability of an event by the ease with which relevant examples come to mind. In the CVH, popular cultural products are treated as the dominant supplier of “available examples” for AI futures.
Self-fulfilling prophecy: Robert K. Merton’s 1948 term for a definition of a situation that evokes behaviour making the initially false conception come true. One of the theoretical foundations of Scenario Reflexivity.
Performativity: In the economic-sociology tradition of Michel Callon and Donald MacKenzie: the idea that theories and models do not merely describe the systems they refer to but actively shape them. The Black–Scholes formula changed option pricing by being adopted by traders. AI-future scenarios can be performative in the same sense.
Reflexivity (Soros): George Soros’s term for the two-way feedback between participants’ beliefs about a system and the system itself: beliefs shape the market, the market shapes beliefs, and the loop cannot be understood from outside it. Applied to AI-futures discourse in the SuperFutures paper.
Hyperstition: A term from the Cybernetic Culture Research Unit (CCRU) in the 1990s for ideas that make themselves real through their own propagation. A provocative precursor to Scenario Reflexivity; the paper argues Reflexivity supplies the mechanistic specificity hyperstition lacks.
Narrative economics: Robert Shiller’s extension of reflexivity: the idea that economic narratives go viral and change economic behaviour - recessions can be partly caused by recession narratives. One of the intellectual lineages behind Scenario Reflexivity.

Missing a term you’d expect to see? Let us know via the (forthcoming) corrections form.