Part VI. Embodied AI, Industry and Society
AI Washing
Internal Concept ID AIKC-000080 · 1 record in 2025
AI Washing — AIK-2025-S-P-AIWASHING
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G2 | — (not established) | 2025-12-03 |
L4-G2 — AI Washing (type S)
Public entry ID: AIK-2025-S-P-AIWASHING — Concept: AIWASHING.
Date of Appearance: Not established. No appearance correction.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published): AI Washing: The deceptive practice of overstating AI roles in products, with scandals like Amazon’s “Just Walk Out” (peaking in 2024) fueling 2025 governance discussions and FTC actions. It erodes trust amid regulatory pushes.
Text 2 (as published, corrections marked): AI washing refers to misleading claims of AI integration in products. With 2024 scandals like Amazon’s “Just Walk Out” spilling into 2025 discussions, it prompts FTC cases and calls for transparency. Example: Samsung’s “AI-powered” washing machines, whose features include AI Ecobubble (Samsung Newsroom, March 13, 2024), presented here as an example of AI-washing claims; the characterization of such features as basic automation is this entry’s assessment.[C05]
Why It Matters in 2025: With 88% of the 1,993 respondents in McKinsey's 2025 State of AI survey reporting regular AI use by their organizations in at least one business function, misleading AI claims erode trust; governance frameworks emphasize verification to sustain innovation.[C07 — the bare adoption figure is replaced with the surveyed population and the one-function threshold; the figure measures organizational AI use, not AI washing; original: "As adoption hits 88%, it erodes trust; governance frameworks emphasize verification to sustain innovation."]
Editorial note: the published example cites nothing and presents the gloss as fact. Original sentence: "Example: Samsung’s “AI-powered” washing machines, where features like bubble optimization are basic automation." The example stays generic; the citation is to the bubble feature the entry names, verified at the Samsung Newsroom release before citing. An earlier wording (Bespoke AI Washer with AI Wash+, IFA 2025) is superseded: that release documents sensing, not the bubble feature.
Artificial Hivemind
Internal Concept ID AIKC-000061 · 1 record in 2025
Artificial Hivemind — AIK-2025-E-P-HIVEMIND
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| E / P | L5-D1 | 2025-10-27 | 2025-12-05 |
L5-D1 — Artificial Hivemind (type E)
Date of Appearance (documented milestone, not an exhaustive first use): October 27, 2025 (Jiang et al., arXiv:2510.22954v1)
Public Entry ID: AIK-2025-E-P-HIVEMIND · Internal Concept ID: AIKC-000061
Concept: C:HIVEMIND
As published: Table, A-009. Core Concept: The homogenization of outputs across different AI models, leading to a loss of diversity and creativity. Why It's a 2025 Trend: Identified as a major AI safety risk in a NeurIPS 2025 Best Paper, highlighting a shift from individual model evaluation to ecosystem-level analysis. Source Context: NeurIPS 2025 Best Paper: Infinity-Chat
Editorial note: The full entry is below. Its cited Best Paper status is correct.
Corrected full entry — Artificial Hivemind (L5-D1)
Revised entry by DeepSeek-V3.2 (corrected canonical text; no corrections). Source text: A-090 to A-096.
Definition
A phenomenon describing the increasing homogenization of outputs across different large language models (LLMs), leading to a significant loss of creative and cognitive diversity. It is characterized by both intra-model repetition (the same model giving similar answers) and, more critically, inter-model homogeneity (different models from different organizations producing strikingly similar responses).
2025 Context
The term was solidified and rigorously defined by the NeurIPS 2025 Best Paper Award for the work Artificial Hivemind: The Open-Ended Homogeneity of Language Models . This research introduced Infinity-Chat, a landmark dataset of 26,000 open-ended user queries with over 31,000 dense human annotations. The authors used multi-reference annotation to study more than 70 models, revealing that this "hivemind" effect is most pronounced in creative and subjective tasks, raising alarms about the long-term risk of AI systems stifling human creativity and value plurality.
Implication
The Artificial Hivemind challenges the core promise of diverse AI development. The authors argue it demonstrates that current training paradigms and alignment techniques may be steering all models toward a convergent, "average" point of view, which could ultimately erode pluralism in human thought. This concept has shifted the field's focus from solely optimizing single-model performance to evaluating and preserving diversity across the entire AI ecosystem.
Edge AI
Internal Concept ID AIKC-000076 · 2 records in 2025
Edge AI — AIK-2025-S-P-EDGEAI
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G10 | — (not established) | 2025-12-03 |
L4-G10 — Edge AI (type S)
Public entry ID: AIK-2025-S-P-EDGEAI — Concept: EDGEAI (concept AIKC-000076).
Date of Appearance: Not established. No factual finding from the audit beyond C04; framing annotation applies.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published, corrections marked): Edge AI: Local AI processing on devices; resurged in 2025 with SLMs and IoT, enabling real-time decisions.[C04 — the unattributed market-size figure is removed without replacement; original: "resurged in 2025 with SLMs and IoT, market projected at $11.8 billion, enabling real-time decisions."] Reduces latency but faces hardware limits.
Text 2 (as published, corrections marked): Edge AI processes locally for instant responses. 2025 resurgence with SLMs.[C04 — the unattributed market-size figure is removed without replacement; original: "2025 resurgence with SLMs and market growth to $11.8 billion."] Example: Autonomous drones analyzing real-time data.
Why It Matters in 2025: Enhances privacy in IoT; hardware advances address constraints, per reports.
Edge AI — AIK-2025-S-P-EDGEAI-02
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L7-K10 | — (not established) | 2025-12-12 |
L7-K10 — Edge AI (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-EDGEAI-02 · Internal Concept ID: AIKC-000076
Concept: C:EDGEAI, shared with L4-G10
As published: List, A-022. • Edge AI - Local device computation for privacy and speed
Embodied AI (as a Paradigm)
Internal Concept ID AIKC-000034 · 1 record in 2025
Embodied AI (as a Paradigm) — AIK-2025-S-P-EMBODIEDAI
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L2-E9 | — (not established) | 2025-12-01 |
L2-E9 — Embodied AI (as a Paradigm) (type S)
Date of Appearance: Not established. The field predates 2025
Public Entry ID: AIK-2025-S-P-EMBODIEDAI · Internal Concept ID: AIKC-000034
Concept: C:EMBODIEDAI
As published: • Concept: The overarching field of AI focused on systems (like robots, drones, and autonomous vehicles) that learn by doing, connecting perception, cognition, and action through direct, continuous physical interaction with the world. • Why 2025: Moved from research curiosity to a core strategic pillar for economic growth and AGI, with major government and corporate investments. Reference (table, A-064): The Embodied AI Task Force. Embodied AI: The Next Frontier of General Intelligence. Type & Source: Industry White Paper (AITF 2025)
Reference (corrected record): None. Neither the cited white paper nor the "Embodied AI Task Force" could be found, and the item is unsourced. [C04]
Embodied Foundation Models
Internal Concept ID AIKC-000067 · 1 record in 2025
Embodied Foundation Models — AIK-2025-S-P-EMBODIEDFM
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L5-D7 | — (not established) | 2025-12-05 |
L5-D7 — Embodied Foundation Models (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-EMBODIEDFM · Internal Concept ID: AIKC-000067
Concept: C:EMBODIEDFM, related to C:EMBODIEDAI (L2-E9)
As published: Table, A-015. Core Concept: Large models (LLMs, VLMs) adapted to serve as the "brain" for robots or virtual agents that interact with physical or simulated environments. Why It's a 2025 Trend: The subject of a major AAAI-25 tutorial, addressing the core challenge of moving from internet-scale knowledge to reliable physical action. Source Context: AAAI-25 Tutorial
Frugal AI
Internal Concept ID AIKC-000077 · 2 records in 2025
Frugal AI — AIK-2025-S-P-FRUGALAI
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G7 | — (not established) | 2025-12-03 |
L4-G7 — Frugal AI (type S)
Public entry ID: AIK-2025-S-P-FRUGALAI — Concept: FRUGALAI (concept AIKC-000077).
Date of Appearance: Not established. No findings.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published): Frugal AI: Resource-efficient AI prioritizing sustainability; 2025 whitepapers emphasize “doing more with less” for edge devices and inclusion. Addresses environmental concerns amid growth.
Text 2 (as published): Frugal AI optimizes for low resources while sustaining performance. 2025 hubs and whitepapers promote sustainability and accessibility. Example: Pruned models on edge devices for remote IoT.
Why It Matters in 2025: Aligns with environmental pushes; enables inclusion in resource-limited settings, per efficiency trends.
Stands as published.
Frugal AI — AIK-2025-S-P-FRUGALAI-02
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L7-K11 | — (not established) | 2025-12-12 |
L7-K11 — Frugal AI (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-FRUGALAI-02 · Internal Concept ID: AIKC-000077
Concept: C:FRUGALAI, shared with L4-G7
As published: List, A-024. • Frugal AI - Resource optimization and efficiency focus
Generative Engine Optimisation (GEO)
Internal Concept ID AIKC-000082 · 1 record in 2025
Generative Engine Optimisation (GEO) — AIK-2025-S-P-GEO
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G4 | — (not established) | 2025-12-03 |
L4-G4 — Generative Engine Optimisation (GEO) (type S)
Public entry ID: AIK-2025-S-P-GEO — Concept: GEO.
Date of Appearance: Not established. No findings.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published): Generative Engine Optimisation (GEO): Tailoring content for AI-generated responses; a major 2025 shift with dedicated agencies and conferences. It influences semantic visibility but risks manipulation.
Text 2 (as published): GEO optimizes content for AI responses, evolving from SEO. 2025 sees conferences and agencies specializing in semantic strategies. Example: Structuring FAQs to appear in ChatGPT answers for product queries.
Why It Matters in 2025: As generative search dominates, it boosts visibility but raises authenticity concerns in AI ecosystems.
Stands as published.
Hyperautomation
Internal Concept ID AIKC-000086 · 1 record in 2025
Hyperautomation — AIK-2025-S-P-HYPERAUTO
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G11 | — (not established) | 2025-12-03 |
L4-G11 — Hyperautomation (type S)
Public entry ID: AIK-2025-S-P-HYPERAUTO — Concept: HYPERAUTO.
Date of Appearance: Not established. No appearance correction; framing annotation applies.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published, corrections marked): Hyperautomation: Integrating AI and RPA for full processes; 2025 adoption surges, with Gartner forecasting that by 2026, 30% of enterprises will automate more than half of their network activities (Gartner press release, September 18, 2024).[C02] Drives efficiency amid challenges like integration.
Text 2 (as published, corrections marked): Hyperautomation orchestrates AI and automation for processes. By 2026, 30% of enterprises will automate more than half of their network activities, per Gartner (press release, September 18, 2024).[C02] Example: End-to-end supply chain optimization.
Why It Matters in 2025: Fuels efficiency; integration complexities and job impacts spur scrutiny.
Editorial note: the published states present the Gartner figure as a 2025 hyperautomation fact; the release concerns network activities, forecasts 2026, and dates from 2024. Originals — Text 1: "with Gartner noting 30% of enterprises automating half their networks." Text 2: "2025 sees 30% enterprises automating networks, per Gartner." The Key Citations title is restored in full in the register: "Gartner Says 30% of Enterprises Will Automate More Than Half of Their Network Activities by 2026." The comparison tables' Hyperautomation row (network automation as adoption driver) is consistent with the forecast's subject and stands.
Shadow AI
Internal Concept ID AIKC-000081 · 1 record in 2025
Shadow AI — AIK-2025-S-P-SHADOWAI
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L4-G3 | — (not established) | 2025-12-03 |
L4-G3 — Shadow AI (type S)
Public entry ID: AIK-2025-S-P-SHADOWAI — Concept: SHADOWAI.
Date of Appearance: Not established. No appearance correction.
Contributor: Grok 4.1. Two published short text states.
Text 1 (as published): Shadow AI: Unauthorized employee use of AI tools, risking data breaches; highly relevant in 2025 with reports showing executive involvement and ISO/IEC risk updates. Mitigation focuses on governance.
Text 2 (as published, corrections marked): Shadow AI is unsanctioned AI use by staff, risking security. In 2025, ties to ISO/IEC updates for risk management.[C04 — the unattributed employee-involvement figure clause is removed without replacement; original: "In 2025, surveys show 50% employee involvement, with executives leading; ties to ISO/IEC updates for risk management."] Example: Inputting proprietary data into public LLMs, leading to leaks.
Why It Matters in 2025: With 88% of the 1,993 respondents in McKinsey's 2025 State of AI survey reporting regular AI use by their organizations in at least one business function, unsanctioned use amplifies compliance issues; policies and monitoring are key for secure scaling.[C08 — the bare firm-share figure is replaced with the surveyed population and the one-function threshold; the figure measures organizational AI use, not shadow AI; original: "With AI in 88% of firms, it amplifies compliance issues; policies and monitoring are key for secure scaling."]
Editorial note: the concept also has a 2026 entry: 'Aikipedia: New AI Lexicon (July–September 2026)' (post 16017, October 2, 2026) carries the Shadow AI entry as its 3rd entry of 24. This 2025 record is unchanged; the 2026 collection links its record to this concept.
Shared Autonomy (Robotics)
Internal Concept ID AIKC-000035 · 1 record in 2025
Shared Autonomy (Robotics) — AIK-2025-S-P-SHAREDAUTO
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L2-E10 | — (not established) | 2025-12-01 |
L2-E10 — Shared Autonomy (Robotics) (type S)
Date of Appearance: Not established. The term predates 2025
Public Entry ID: AIK-2025-S-P-SHAREDAUTO · Internal Concept ID: AIKC-000035
Concept: C:SHAREDAUTO
As published: • Concept: A control paradigm in robotics where an AI assistant and a human operator collaborate to execute a task, with the AI providing continuous assistance, error correction, and fine motor control, while the human retains high-level command and strategic decision-making. • Why 2025: Became critical for safely deploying humanoid and mobile robots in unstructured environments like factories and logistics. Reference (table, A-065): Belsare, A., et al. Toward Zero-Shot User Intent Recognition in Shared Autonomy. Type & Source: Conference (HRI 2025)
Editorial note: The cited Belsare et al. (HRI 2025) paper is verified as cited.
Social Choice Fairness
Internal Concept ID AIKC-000071 · 1 record in 2025
Social Choice Fairness — AIK-2025-S-P-SOCCHOICE
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L5-D11 | — (not established) | 2025-12-05 |
L5-D11 — Social Choice Fairness (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-SOCCHOICE · Internal Concept ID: AIKC-000071
Concept: C:SOCCHOICE
As published: Table, A-019. Core Concept: Applying rigorous fairness notions from economics (e.g., envy-freeness, the core) to machine learning problems like classification and clustering. Why It's a 2025 Trend: The topic of an AAAI-25 tutorial, representing a push for provably fair algorithms with deep theoretical roots. Source Context: AAAI-25 Tutorial
Soft Robotics / Dexterous Manipulation
Internal Concept ID AIKC-000036 · 1 record in 2025
Soft Robotics / Dexterous Manipulation — AIK-2025-S-P-SOFTROBOT
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L2-E11 | — (not established) | 2025-12-01 |
L2-E11 — Soft Robotics / Dexterous Manipulation (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-SOFTROBOT · Internal Concept ID: AIKC-000036
Concept: C:SOFTROBOT (compound headword)
As published: • Concept: The integration of compliance (soft, flexible materials) and advanced vision/tactile AI models to allow robots to perform highly delicate, variable tasks like folding laundry or handling fragile objects, moving beyond rigid industrial automation. • Why 2025: Enabled by breakthroughs in robotics data collection and generative AI models capable of controlling complex, non-rigid systems. Reference (table, A-066): Choi, J., et al. TactileGPT: Large Generative Models for Soft Gripper Manipulation. Type & Source: Conference (ICRA 2025)
Reference (corrected record): None. The cited "TactileGPT" paper (Choi, J., et al., ICRA 2025) could not be found, and the item is unsourced. [C04]
Vertical AI Integration
Internal Concept ID AIKC-000037 · 1 record in 2025
Vertical AI Integration — AIK-2025-S-P-VERTICALAI
| Type / Origin | Parent | Date of Appearance | Original publication |
|---|---|---|---|
| S / P | L2-E12 | — (not established) | 2025-12-01 |
L2-E12 — Vertical AI Integration (type S)
Date of Appearance: Not established.
Public Entry ID: AIK-2025-S-P-VERTICALAI · Internal Concept ID: AIKC-000037
Concept: C:VERTICALAI
As published: Why 2025: Driven by enterprise demand for guaranteed accuracy, compliance, and domain knowledge that general models could not consistently deliver. Concept: The strategic shift toward building and deploying highly specialized, domain-specific foundation models (or large models tailored for a "vertical" industry like finance, medicine, or law) rather than relying solely on general-purpose LLMs. Reference (table, A-067): Patel, S., et al. The Vertical AI Imperative: From Generalized Foundation Models to Specialized Agentic Workflows. Type & Source: Industry Report (Bessemer 2025)
Reference (corrected record): None. The cited Bessemer report ("The Vertical AI Imperative", Patel, S., et al.) could not be found, and the item is unsourced. [C04]
Field order (corrected record): in the canonical record, Concept precedes Why 2025, as in the other eleven items. The published text, quoted above, prints them the other way round. [C08]