Sources & Methodology

1. Research Scope & Analytical Framework

The AI Situation Room is an independent, continuously updated observatory that tracks the global state of artificial intelligence through quantitative measurement. It synthesises heterogeneous data streams into composite indices designed for longitudinal monitoring rather than point-in-time snapshots.

Research Question

"How can we quantify the global state of AI across adoption, capability, geopolitics, market sentiment, and public interest—and present these dimensions as a single, interpretable dashboard?"

Analytical Pillars

The dashboard is organised around five orthogonal measurement dimensions:

• Adoption — Enterprise penetration, developer tooling uptake, industrial automation density
• Capability — Frontier model performance, benchmark progression, agentic and multimodal scores
• Geopolitics — National infrastructure, hardware manufacturing, IP registrations, research output
• Market — AI-focused ETF performance, hype-to-value ratios, venture sentiment
• Public Interest — Search trend data, chatbot traffic volume, framework popularity

Update Cadence

Data is updated through a combination of semi-automated API integrations (Arxiv, Google Trends) and manual curation via a restricted admin interface. The dashboard is not a real-time feed; update frequency varies by dataset from daily (ETF prices) to quarterly (robotics density). All timestamps reflect the date of last verified update, not the date of original publication by the source.

2. Composite Index Methodology

Five headline indices distil the dashboard's raw data into interpretable signals. Each index is constructed from weighted sub-components, documented below with their formulae, component definitions, source mappings, and weighting rationale.

2.1 World AI Adoption Index

A normalised composite (0–1 scale) capturing the breadth and depth of AI integration across society. The index deliberately weights enterprise adoption most heavily, reflecting the thesis that commercial deployment is the strongest near-term signal of systemic AI integration.

D — Developer Ecosystem Sub-components

2.2 Progress to AGI Index

An experimental metric estimating cumulative progress toward artificial general intelligence, expressed as a percentage. The index uses a geometric mean of three sub-dimensions to enforce the constraint that balanced progress across all fronts is required—excellence in one area cannot compensate for near-zero capability in another.

2.2b Estimated AGI Year (Time to AGI)

A forward-looking projection of when each AGI pillar reaches 100%, derived from the Progress to AGI index above. The estimated year is determined by the bottleneck—the slowest pillar to reach full capability. This is a moderate scenario using conservative growth assumptions.

Where r is the monthly exponential growth rate. The estimated year is then: current_year + months_to_100 / 12, taken from the pillar with the largest value (the bottleneck).

2.3 AI Bubble Index

A market-sentiment gauge measuring the ratio of hype momentum to fundamental value creation, modulated by a substance factor. Higher values indicate greater divergence between narrative enthusiasm and demonstrated economic utility.

The Hype Growth component (H) blends two signals: cumulative growth since the earliest recorded data and recent weighted year-over-year momentum. This captures both the total accumulated hype and whether it is still accelerating or cooling off.

Scale Interpretation

2.4 Global AI Power Index

A composite country-level ranking scoring 12 nations across five strategic pillars. Each pillar is normalised to a 0–100 scale using min-max normalisation within the observed dataset, then weighted and summed.

2.5 Intelligence Doubling Time

An exponential-growth metric measuring how many months it takes the frontier AI intelligence score to double. Analogous to Moore's Law for transistors, this index tracks the pace of AI capability improvement rather than absolute capability. A declining doubling time indicates accelerating progress.

3. Data Sources Registry

The following table catalogues every dataset consumed by the dashboard, including provenance, collection method, temporal coverage, and known constraints. All data is stored as JSON exports from a SQLite database.

3.1 Model Rankings

3.2 Geopolitics

3.3 AI Public Interest

3.4 ETF Prices

3.5 Enterprise AI Adoption

3.6 AI Chatbot Website Traffic

3.7 Agentic Frameworks

3.8 Robotics & Physical Automation

3.9 Key Metrics (Computed)

4. Data Collection & Processing Pipeline

Data flows through a four-stage pipeline from source acquisition to frontend rendering. No stage is fully automated; human verification is required at each checkpoint to ensure data integrity.

4.1 Acquisition

• API Integration — Arxiv API for paper counts; Google Trends data export for search interest time series
• Manual Curation — Admin panel (authenticated Flask route) for direct data entry of ETF prices, model rankings, geopolitical indicators, and website traffic figures
• Secondary Sources — Deloitte reports, IFR publications, and SimilarWeb dashboards are manually reviewed and transcribed

4.2 Validation

• Range checks on numerical values (e.g., interest scores 0–100, prices > 0)
• Temporal consistency (new entries must be more recent than existing records)
• Cross-source verification for key figures where multiple sources exist
• Manual review of all entries before database commit

4.3 Storage

All validated data is stored in a SQLite database (ai_situation_room.db) with 9 normalised tables. The schema enforces primary keys, non-null constraints on required fields, and foreign-key-like consistency for country codes.

4.4 Export & Rendering

A Python export script (export_data.py) serialises each table to JSON in the data/ directory alongside a metadata.json manifest containing row counts and export timestamps. The frontend consumes these JSON files directly via fetch requests, with Chart.js handling visualisation.

5. Normalization & Weighting Procedures

5.1 Min-Max Normalization

Country-level indicators in the Global AI Power Index are normalised using min-max scaling within each pillar:

This maps each country's raw score to a 0–100 range within the observed dataset. The normalisation is relative, not absolute—a country scoring 100 is the best in the current sample, not at a theoretical maximum.

5.2 Percentage-Based Scaling

The World AI Adoption Index components are expressed as percentages or ratios before weighting. Where raw data is not natively percentage-based (e.g., GitHub stars), it is normalised against an estimated total population denominator.

5.3 Weight Selection

Composite index weights are determined through editorial judgement, not statistical optimisation. Weights reflect the research team's assessment of each component's relative importance to the phenomenon being measured. All weights are disclosed in Section 2 and should be interpreted as explicit analytical choices, not objective truths.

6. Limitations & Caveats

Transparency about limitations is essential for responsible interpretation. The following constraints should be considered when citing or acting upon dashboard outputs.

6.1 Data Freshness & Latency

• Updates are manual or semi-automated; the dashboard is not a real-time feed
• Some upstream sources (IFR, Deloitte) publish annually, creating inherent staleness
• ETF prices may lag by 1–3 trading days depending on manual update timing

6.2 Source Bias

• Google Trends — Uses the “Artificial Intelligence” topic for cross-language consistency; still overrepresents countries with high internet penetration and Google market share; underrepresents China (where Baidu dominates search)
• Arxiv — English-language bias; underrepresents Chinese-language research published on domestic platforms (CNKI, Wanfang)
• Lens.org — Patent filing-to-publication lag of 12–18 months; coverage varies by patent office
• SimilarWeb — Modelled estimates, not census data; mobile app and API usage undercounted

6.3 Methodological Limitations

• Composite index weights are editorial, not empirically derived
• The AGI Progress Index assumes three-dimensional decomposition; alternative frameworks (e.g., Levels of AGI by Morris et al., 2023) may yield different assessments
• The Bubble Index uses six AI-focused ETFs as a market proxy, which may not represent the full AI investment landscape
• Small sample sizes in some datasets (e.g., 4 robotics records, 5 frameworks) limit statistical robustness

6.4 Missing Dimensions

• Talent migration and brain drain between nations
• Regulatory maturity and AI governance frameworks
• Open-source contributions beyond star counts (commits, contributors, forks)
• Energy consumption and environmental impact of AI compute
• Private-sector AI deployment beyond survey data
• AI safety and alignment research as a separate track

6.5 Peer Review Status

This methodology has not been peer-reviewed or published in an academic venue. The indices are designed for monitoring and discussion, not for policy decisions or investment advice. Users are encouraged to examine the underlying data and form independent assessments.

7. Versioning & Change Log

The methodology is versioned to track analytical evolution. Breaking changes to index formulae or weight structures will increment the major version; data source additions or corrections increment the minor version.

8. Citation & Contact

Suggested Citation

Contact & Correspondence

For methodological inquiries, data corrections, or collaboration proposals:

• Email: aisituationroom@proton.me

Data Usage & Disclaimer

All data displayed on this dashboard is aggregated from publicly available sources cited above. The composite indices and analytical commentary represent independent editorial analysis and do not constitute financial advice, policy recommendations, or official benchmarking. Redistribution of aggregated data should credit the original sources and this dashboard.

9. Project Status & Contributing

The AI Situation Room is, at its current stage, a conceptual prototype. It was built with real data sources and genuine analytical intent, but it should be understood as an early-stage demonstration of what a comprehensive AI situation room could look like — not a production-grade intelligence platform. The indices, dashboards, and data pipelines presented here are functional proofs of concept designed to explore how disparate AI signals might be synthesised into a single, coherent observatory.

If you are interested in contributing — whether through data engineering, frontend development, analytical methodology, or domain expertise — the project welcomes collaboration. Reach out at aisituationroom@proton.me to discuss how you can help shape the next iteration.