LLM summaryLLM summaryBasic text summary used in search results, entity link tooltips, info boxes, and related page cards.crux content improve <id>ScheduleScheduleHow often the page should be refreshed. Drives the overdue tracking system.Set updateFrequency in frontmatterEntityEntityYAML entity definition with type, description, and related entries.Add entity YAML in data/entities/Edit historyEdit historyTracked changes from improve pipeline runs and manual edits.crux edit-log view <id>OverviewOverviewA ## Overview heading section that orients readers. Helps with search and AI summaries.
Tables0/ ~1TablesData tables for structured comparisons and reference material.Add data tables to the pageDiagrams0DiagramsVisual content — Mermaid diagrams, charts, or Squiggle estimate models.Add Mermaid diagrams or Squiggle modelsInt. links14/ ~3Int. linksLinks to other wiki pages. More internal links = better graph connectivity.Ext. links0/ ~1Ext. linksLinks to external websites, papers, and resources outside the wiki.Add links to external sourcesFootnotes0/ ~2FootnotesFootnote citations [^N] with source references at the bottom of the page.Add [^N] footnote citationsReferences0/ ~1ReferencesCurated external resources linked via <R> components or cited_by in YAML.Add <R> resource linksQuotes0QuotesSupporting quotes extracted from cited sources to back up page claims.crux citations extract-quotes <id>Accuracy0AccuracyCitations verified against their sources for factual accuracy.crux citations verify <id>
Issues1
StructureNo tables or diagrams - consider adding visual content
Intelligence Paradigms
Overview
This section explores different paradigms for building intelligent systems. Understanding architectural diversity helps anticipate which safety approaches may be needed and which risks are most relevant.
Current Paradigms
Dense TransformersConceptDense TransformersComprehensive analysis of dense transformers (GPT-4, Claude 3, Llama 3) as the dominant AI architecture (95%+ of frontier models), with training costs reaching \$100M-500M per run and 2.5x annual c...Quality: 58/100
SSM/MambaCapabilityState-Space Models / MambaComprehensive analysis of state-space models (SSMs) like Mamba as transformer alternatives, documenting that Mamba-3B matches Transformer-6B perplexity with 5x throughput but lags on in-context lea...Quality: 54/100
State Space Models:
Alternative to attention mechanism
Linear scaling with sequence length
Emerging research area
Neuro-SymbolicCapabilityNeuro-Symbolic Hybrid SystemsComprehensive analysis of neuro-symbolic AI systems combining neural networks with formal reasoning, documenting AlphaProof's 2024 IMO silver medal (28/42 points) and 2025 gold medal achievements. ...Quality: 55/100
Hybrid neural-symbolic systems:
Combine learning with logical reasoning
Potentially more interpretable
Research stage
Scaffolding Approaches
Minimal ScaffoldingCapabilityMinimal ScaffoldingAnalyzes minimal scaffolding (basic AI chat interfaces) showing 38x performance gap vs agent systems on code tasks (1.96% → 75% on SWE-bench), declining market share from 80% (2023) to 35% (2025), ...Quality: 52/100
Foundation models with basic prompting
Light ScaffoldingCapabilityLight ScaffoldingLight scaffolding (RAG, function calling, simple chains) represents the current enterprise deployment standard with 92% Fortune 500 adoption, achieving 88-91% function calling accuracy and 18% RAG ...Quality: 53/100
Chain-of-thought, few-shot learning
Heavy ScaffoldingConceptHeavy Scaffolding / Agentic SystemsComprehensive analysis of multi-agent AI systems with extensive benchmarking data showing rapid capability growth (77.2% SWE-bench, 5.5x improvement 2023-2025) but persistent reliability challenges...Quality: 57/100
Extensive tool use, agentic frameworks
Alternative Pathways
World ModelsCapabilityWorld Models + PlanningComprehensive analysis of world models + planning architectures showing 10-500x sample efficiency gains over model-free RL (EfficientZero: 194% human performance with 100k vs 50M steps), but estima...Quality: 54/100
Learning predictive world representations
Provably SafeConceptProvable / Guaranteed Safe AIProvable Safe AI uses formal verification to provide mathematical safety guarantees, with UK's ARIA investing £59M through 2028. Current verification handles ~10^6 parameters while frontier models ...Quality: 64/100
Architectures with formal safety guarantees
Biological Approaches
Brain-Computer InterfacesCapabilityBrain-Computer InterfacesComprehensive analysis of BCIs concluding they are irrelevant for TAI timelines (<1% probability of dominance) due to fundamental bandwidth constraints—current best of 62 WPM vs. billions of operat...Quality: 49/100
Biological OrganoidCapabilityBiological / Organoid ComputingComprehensive analysis of biological/organoid computing showing current systems (DishBrain with ~800k neurons, Brainoware at 78% speech recognition) achieve 10^6-10^9x better energy efficiency than...Quality: 54/100
Genetic EnhancementCapabilityGenetic Enhancement / SelectionGenetic enhancement via embryo selection currently yields 2.5-6 IQ points per generation with 10% variance explained by polygenic scores, while theoretical iterated embryo selection could achieve 1...Quality: 51/100
Whole-Brain EmulationCapabilityWhole Brain EmulationComprehensive analysis of whole brain emulation finding <1% probability of arriving before AI-based TAI, with scanning speed (100,000x too slow for human brains) as the primary bottleneck despite r...Quality: 48/100
Novel/UnknownCapabilityNovel / Unknown ApproachesAnalyzes probability (1-15%) of novel AI paradigms emerging before transformative AI, systematically reviewing historical prediction failures (expert AGI timelines shifted 43 years in 4 years, 13 y...Quality: 53/100
Heavy Scaffolding / Agentic SystemsConceptHeavy Scaffolding / Agentic SystemsComprehensive analysis of multi-agent AI systems with extensive benchmarking data showing rapid capability growth (77.2% SWE-bench, 5.5x improvement 2023-2025) but persistent reliability challenges...Quality: 57/100Light ScaffoldingCapabilityLight ScaffoldingLight scaffolding (RAG, function calling, simple chains) represents the current enterprise deployment standard with 92% Fortune 500 adoption, achieving 88-91% function calling accuracy and 18% RAG ...Quality: 53/100Minimal ScaffoldingCapabilityMinimal ScaffoldingAnalyzes minimal scaffolding (basic AI chat interfaces) showing 38x performance gap vs agent systems on code tasks (1.96% → 75% on SWE-bench), declining market share from 80% (2023) to 35% (2025), ...Quality: 52/100Sparse / MoE TransformersCapabilitySparse / MoE TransformersComprehensive reference on Sparse/MoE transformer architectures covering key models (Mixtral, DeepSeek-V3, DBRX, Switch Transformer), efficiency gains (2-18x parameter efficiency ratios), and safet...Quality: 55/100Genetic Enhancement / SelectionCapabilityGenetic Enhancement / SelectionGenetic enhancement via embryo selection currently yields 2.5-6 IQ points per generation with 10% variance explained by polygenic scores, while theoretical iterated embryo selection could achieve 1...Quality: 51/100State-Space Models / MambaCapabilityState-Space Models / MambaComprehensive analysis of state-space models (SSMs) like Mamba as transformer alternatives, documenting that Mamba-3B matches Transformer-6B perplexity with 5x throughput but lags on in-context lea...Quality: 54/100