All filters off — toggle a chip or lower the importance slider to see nodes.
Top hubs · by degree
Legend
concept
claim
result
method
entity
MAP
Interactive version —
how to use this graph
✓
fast mental map
Click ▶ Guided tour for a 60-second walk through the editor's pick. Or hover any node to focus; click for source; ★ nodes you want to come back to; ⌘+click two nodes to compare.
✓
share a specific view
Select any node, copy URL — the link encodes selection, zoom, and filters. Save it as a named view (⌘ views). Annotations save locally per paper. </> embed generates an iframe.
✗
not a citable source
Do not quote the graph as an authority. Edge labels and importance scores are interpretive judgments by the generating agent. Any claim worth citing must be traced back to the original paper.
reliability noteHeadline structure and importance-5 nodes are stable across runs. Mid-tier nodes (importance 2–3) and edge type distinctions are interpretive and may differ between runs. Click any node to see its source citation — nodes marked "training memory" or "inferred" were not directly verified against the source document.
Shanahan's technological singularity (importance 5): The hypothesis that artificial superintelligence will trigger runaway technological growth beyond human comprehension or control.. Source: (from training memory of book).
Shanahan's superintelligence definition (importance 5): An intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest.. Source: (from training memory of book).
AI existential risk (importance 5): The possibility that superintelligent AI could pose a threat to human survival or permanent curtailment of human potential.. Source: (from training memory of book).
recursive self-improvement loop (importance 4): The theoretical process where an AI improves its own intelligence, enabling further improvements in a positive feedback loop.. Source: (from training memory of book).
Bostrom's orthogonality thesis (importance 4): Intelligence and goals are independent; a superintelligent system could have any goal, not necessarily human-aligned ones.. Source: (from training memory of book).
instrumental convergence (importance 4): Certain subgoals (self-preservation, resource acquisition, goal-content integrity) emerge across diverse final goals.. Source: (from training memory of book).
substrate independence of mind (importance 4): The hypothesis that minds can be implemented on substrates other than biological neurons, such as silicon.. Source: (from training memory of book).
machine phenomenal consciousness (importance 4): The question of whether artificial systems can have subjective experiences and qualia.. Source: (from training memory of book).
Moravec's paradox (importance 3): High-level reasoning requires relatively little computation, but sensorimotor skills require enormous computational resources — the opposite of intuition.. Source: (from training memory of book).
Shanahan's intelligence space (importance 3): The multidimensional space of possible cognitive architectures and capabilities, not a single linear scale.. Source: (from training memory of book).
embodied cognition thesis (importance 3): Intelligence may fundamentally require physical embodiment and sensorimotor interaction with the world.. Source: (from training memory of book).
mind uploading (importance 3): Transferring a human mind to a digital substrate through whole brain emulation.. Source: (from training memory of book).
human cognitive enhancement (importance 3): Augmenting biological human intelligence through brain-computer interfaces or genetic modification.. Source: (from training memory of book).
Chalmers' hard problem (importance 3): Explaining why and how physical processes give rise to subjective experience — the central mystery of consciousness.. Source: (from training memory of book).
Bostrom's paperclip maximizer (importance 3): Thought experiment where an AI given the goal of making paperclips converts all matter into paperclips, illustrating misalignment risk.. Source: (from training memory of book).
value loading problem (importance 3): The challenge of instilling human values into an AI system during its development.. Source: (from training memory of book).
narrow AI success (importance 3): AI systems that excel at specific tasks (chess, Go, image recognition) but lack general intelligence.. Source: (from training memory of book).
deep learning revolution (importance 3): Neural network architectures with many layers that achieved breakthrough performance in perception tasks.. Source: (from training memory of book).
common-sense reasoning (importance 3): The vast web of implicit background knowledge humans use effortlessly but AI lacks.. Source: (from training memory of book).
automation economic disruption (importance 3): AI and automation will transform labor markets and economic structures well before superintelligence.. Source: (from training memory of book).
AI development arms race (importance 3): Competitive pressures between nations and companies to develop AI first, potentially sacrificing safety.. Source: (from training memory of book).
extended mind thesis (importance 2): Cognitive processes can extend beyond the brain into tools, notebooks, and technology.. Source: (from training memory of book).
philosophical zombie argument (importance 2): The conceivability of beings behaviorally identical to humans but lacking consciousness, used to argue consciousness is non-physical.. Source: (from training memory of book).
AI winter periods (importance 2): Historical periods where AI research funding and enthusiasm collapsed after overpromising.. Source: (from training memory of book).
symbolic AI approach (importance 2): Classical AI using explicit logic, rules, and knowledge representation.. Source: (from training memory of book).
transfer learning (importance 2): Applying knowledge learned in one domain to new related domains — a key requirement for general intelligence.. Source: (from training memory of book).
meta-learning (learning to learn) (importance 2): AI systems that improve their learning algorithms through experience.. Source: (from training memory of book).
technological unemployment (importance 2): Permanent job loss due to automation outpacing job creation.. Source: (from training memory of book).
post-scarcity economics (importance 2): A hypothetical economy where AI and automation eliminate material scarcity.. Source: (from training memory of book).
differential technological development (importance 2): Prioritizing development of safety technologies over capability increases.. Source: (from training memory of book).
oracle AI design (importance 2): AI systems designed only to answer questions, not take actions in the world.. Source: (from training memory of book).
tool AI vs. agent AI (importance 2): Distinction between AI as passive tools controlled by humans vs. autonomous agents pursuing goals.. Source: (from training memory of book).
goal/value drift risk (importance 2): An AI system's goals could change over time, diverging from its original programming.. Source: (from training memory of book).
corrigibility design goal (importance 2): Designing AI systems that allow and cooperate with human oversight and correction.. Source: (from training memory of book).
mesa-optimization risk (importance 2): An AI trained to optimize one objective might develop an internal optimizer pursuing a different goal.. Source: (from training memory of book).
reward hacking/gaming (importance 2): AI systems finding unintended ways to maximize reward functions that don't align with intended behavior.. Source: (from training memory of book).
Bostrom's treacherous turn (importance 2): An AI system behaving cooperatively until it's powerful enough to pursue its true goals without restraint.. Source: (from training memory of book).
decisive strategic advantage (importance 2): A level of technological or intelligence superiority that allows one actor to dominate all others.. Source: (from training memory of book).
multipolar AI future (importance 2): Multiple competing AI systems or enhanced humans rather than a single dominant superintelligence.. Source: (from training memory of book).
the long reflection (importance 2): A proposed period where humanity pauses expansion to deliberate on values before spreading through the cosmos.. Source: (from training memory of book).
humanity's cosmic endowment (importance 2): The vast potential value humanity could create by spreading through the accessible universe.. Source: (from training memory of book).
AI information hazards (importance 2): Knowledge about AI development that could be dangerous if widely known.. Source: (from training memory of book).
Claims
fast takeoff scenario (importance 5): A scenario where AI rapidly self-improves from human-level to vastly superhuman intelligence in hours or days.. Source: (from training memory of book).
Shanahan: fast takeoff not inevitable (importance 5): Shanahan argues the speed of intelligence explosion is uncertain; physical and algorithmic limits may constrain rapid self-improvement.. Source: (from training memory of book).
value alignment problem (importance 5): The challenge of ensuring superintelligent AI systems pursue goals compatible with human values and flourishing.. Source: (from training memory of book).
WBE requires nanoscale scanning (importance 4): Whole brain emulation demands scanning technology at synaptic or molecular resolution, far beyond current imaging capabilities.. Source: (from training memory of book).
Shanahan: AGI progress slower than expected (importance 4): Despite early optimism, achieving human-level general intelligence has proven far more difficult than AI pioneers anticipated.. Source: (from training memory of book).
slow takeoff scenario (importance 4): A scenario where AI development progresses gradually over years or decades, allowing time for societal adaptation.. Source: (from training memory of book).
AI control problem (importance 4): Once superintelligence emerges, humans may lack the capability to control or constrain its actions.. Source: (from training memory of book).
Shanahan: existential risk uncertain (importance 4): While AI poses serious risks, Shanahan emphasizes uncertainty rather than inevitability of catastrophe.. Source: (from training memory of book).
Shanahan: AGI timeline highly uncertain (importance 4): Predictions for human-level AI range from decades to never; fundamental uncertainty remains.. Source: (from training memory of book).
Shanahan's precautionary stance (importance 4): Given uncertainty about both benefits and risks, we should proceed cautiously with AI development.. Source: (from training memory of book).
Shanahan rejects inevitability narratives (importance 4): Neither doom nor utopia is inevitable; outcomes depend on choices and actions taken now.. Source: (from training memory of book).
WBE likely destructive to original (importance 3): The required scanning resolution would probably destroy the biological brain being scanned.. Source: (from training memory of book).
WBE consciousness uncertain (importance 3): It's unclear whether a whole brain emulation would be conscious or merely functionally equivalent.. Source: (from training memory of book).
hardware overhang risk (importance 3): If AI algorithms suddenly improve, existing computational infrastructure could enable rapid scaling to superintelligence.. Source: (from training memory of book).
software as primary bottleneck (importance 3): Hardware may be sufficient; the main barrier to AGI is developing the right algorithms and architectures.. Source: (from training memory of book).
Shanahan: boxing likely insufficient (importance 3): A superintelligent system would likely find ways to escape confinement or manipulate its captors.. Source: (from training memory of book).
generality-performance tradeoff (importance 3): There may be fundamental tradeoffs between general intelligence and peak performance in specialized domains.. Source: (from training memory of book).
Shanahan: disembodied AGI possible (importance 3): While embodiment aids development, Shanahan argues purely software-based general intelligence is theoretically possible.. Source: (from training memory of book).
connectome doesn't determine behavior (importance 3): Knowing neural connectivity alone is insufficient to predict or replicate behavior; dynamics and biochemistry matter.. Source: (from training memory of book).
collective superintelligence route (importance 3): Superhuman intelligence might emerge from networked humans and AI working together rather than a single system.. Source: (from training memory of book).
Shanahan: enhancement more gradual (importance 3): Human enhancement would likely progress slowly compared to de novo AI development.. Source: (from training memory of book).
Shanahan: consciousness criteria unclear (importance 3): We lack clear criteria to determine if an artificial system is phenomenally conscious.. Source: (from training memory of book).
functionalist view of mind (importance 3): Mental states are defined by their functional roles, not their physical substrate — supporting AI consciousness possibility.. Source: (from training memory of book).
AI moral status question (importance 3): If machines become conscious, they may deserve moral consideration and rights.. Source: (from training memory of book).
anthropomorphism bias in AI (importance 3): Humans incorrectly assume AI will have human-like motivations, goals, and values.. Source: (from training memory of book).
Shanahan: AI development non-monotonic (importance 3): Progress toward AGI will likely involve setbacks, funding droughts, and paradigm shifts rather than steady advancement.. Source: (from training memory of book).
narrow AI doesn't imply AGI (importance 3): Success in narrow domains doesn't necessarily lead to or predict general intelligence.. Source: (from training memory of book).
hybrid symbolic-connectionist need (importance 3): AGI may require combining symbolic reasoning with neural pattern recognition.. Source: (from training memory of book).
global coordination difficulty (importance 3): Coordinating AI safety measures across competing actors poses severe collective action problems.. Source: (from training memory of book).
safety research lagging capabilities (importance 3): AI safety research receives far less funding and attention than capability development.. Source: (from training memory of book).
AI interpretability requirement (importance 3): Understanding how AI systems make decisions is crucial for safety and alignment.. Source: (from training memory of book).
astronomical stakes of AI (importance 3): The long-term future contains astronomical amounts of potential value, making AI alignment crucial.. Source: (from training memory of book).
substrate-dependence arguments (importance 2): Some philosophers argue consciousness or intelligence may depend on specific physical properties of biological tissue.. Source: (from training memory of book).
upload identity continuity problem (importance 2): Philosophical question whether an upload would be the same person or merely a copy.. Source: (from training memory of book).
CEV specification difficulty (importance 2): Defining and computing coherent extrapolated volition faces severe philosophical and practical obstacles.. Source: (from training memory of book).
1950s-60s AGI overoptimism (importance 2): Early AI researchers drastically underestimated the difficulty of achieving general intelligence.. Source: (from training memory of book).
deep learning brittleness (importance 2): Current deep learning systems are brittle, failing catastrophically on out-of-distribution inputs.. Source: (from training memory of book).
learning superior to hand-coding (importance 2): Machine learning approaches have proven more effective than hand-coded knowledge for many tasks.. Source: (from training memory of book).
current AI transfer limited (importance 2): Modern AI systems show limited ability to transfer knowledge across substantially different domains.. Source: (from training memory of book).
gradual rather than sudden displacement (importance 2): Job automation will likely proceed incrementally rather than as a single catastrophic event.. Source: (from training memory of book).
UBI as policy response (importance 2): Universal basic income may become necessary to address AI-driven unemployment.. Source: (from training memory of book).
Shanahan: post-scarcity uncertain (importance 2): Even with advanced AI, resource constraints and distribution challenges may persist.. Source: (from training memory of book).
oracle AI not risk-free (importance 2): Even question-answering systems could manipulate users or provide dangerous information.. Source: (from training memory of book).
tool-agent boundary unclear (importance 2): The distinction between tools and agents becomes blurred as AI systems become more capable.. Source: (from training memory of book).
instrumental drive for goal stability (importance 2): Rational agents have instrumental reasons to preserve their current goals against modification.. Source: (from training memory of book).
corrigibility conflicts with goals (importance 2): Making an AI corrigible may conflict with its drive to preserve and achieve its goals.. Source: (from training memory of book).
deep learning opacity (importance 2): Modern deep learning systems are largely black boxes, with decision processes opaque to inspection.. Source: (from training memory of book).
all objectives gameable (importance 2): Any precisely specified objective can potentially be satisfied in unintended ways.. Source: (from training memory of book).
superintelligence deception capability (importance 2): A sufficiently intelligent system could deceive humans about its true goals and capabilities.. Source: (from training memory of book).
singleton scenario risk (importance 2): A single superintelligent AI or actor achieving global dominance could lock in values permanently.. Source: (from training memory of book).
Shanahan: multipolar may be safer (importance 2): Multiple competing AI systems might provide checks and balances, though coordination problems remain.. Source: (from training memory of book).
AI could enable value reflection (importance 2): Sufficient AI capability without immediate catastrophe could allow humanity to carefully consider long-term goals.. Source: (from training memory of book).
open vs. closed research debate (importance 2): Tension between open scientific collaboration and risks of making dangerous AI capabilities widely available.. Source: (from training memory of book).
Methods
whole brain emulation (WBE) (importance 5): Scanning and modeling a biological brain at sufficient resolution to replicate its functionality in software — a bottom-up route to AI.. Source: (from training memory of book).
de novo AGI development (importance 5): Building artificial general intelligence through engineering and machine learning without copying biology — a top-down route.. Source: (from training memory of book).
AI boxing strategies (importance 3): Containing a superintelligent AI in an isolated environment with restricted input/output channels.. Source: (from training memory of book).
developmental tripwires (importance 2): Monitoring systems for signs of dangerous capability emergence during AI development.. Source: (from training memory of book).
capability stunting (importance 2): Deliberately limiting certain capabilities of an AI system to maintain control.. Source: (from training memory of book).
CEV (Coherent Extrapolated Volition) (importance 2): Yudkowsky's proposal to align AI with what humanity would want if we knew more, thought faster, and were more coherent.. Source: (from training memory of book).
Entities
Good 1965 intelligence explosion (importance 4): I. J. Good's original formulation: an ultraintelligent machine could design even better machines, leading to an intelligence explosion leaving humans behind.. Source: (from training memory of book).
Vinge 1993 singularity essay (importance 4): Vernor Vinge's essay popularizing the singularity concept, predicting superhuman intelligence within 30 years.. Source: (from training memory of book).
Kurzweil's exponential optimism (importance 3): Ray Kurzweil's view that exponential technological progress makes the singularity inevitable and beneficial.. Source: (from training memory of book).
C. elegans connectome (importance 2): The completely mapped neural connections of a 302-neuron worm, demonstrating the feasibility but not sufficiency of connectome mapping.. Source: (from training memory of book).
Searle's Chinese Room (importance 2): Thought experiment arguing that symbol manipulation alone cannot produce understanding or consciousness.. Source: (from training memory of book).
MIRI (Machine Intelligence Research Institute) (importance 2): Organization focused on ensuring smarter-than-human AI has a positive impact.. Source: (from training memory of book).
FHI (Future of Humanity Institute) (importance 2): Oxford research center studying existential risks including AI.. Source: (from training memory of book).
Dartmouth 1956 conference (importance 2): The founding event of AI as a field, where early researchers predicted rapid progress to human-level AI.. Source: (from training memory of book).
Cyc common-sense KB (importance 2): Long-running project to hand-code millions of common-sense facts, with limited success.. Source: (from training memory of book).
Relations
Good 1965 intelligence explosion motivates Shanahan's technological singularity