Metaculus hosts an AI 2027 Tournament with 19 forecasting questions derived directly from the AI 2027 scenario. The tournament launched in June 2025 and tracks crowd predictions on technology, economics, politics, and societal impact through 2028 and beyond.

This page compares the AI 2027 scenario’s claims, the Metaculus crowd forecasts, and our independent tracker assessments.

Forecast freshness: The crowd values below are the original March 2026 snapshot — not refreshed. A check on November 7, 2026 did not establish current forecasts or forecaster counts. That check date is not the date of a community forecast. The snapshot records what this comparison previously cited, not a verified forecast timestamp; it must not be read as today’s crowd view.

How to Read This Comparison

Different measurement systems: AI 2027 makes narrative claims (“X will happen by Y”). Metaculus uses probability percentages or numerical estimates. Our tracker uses status labels (confirmed, on-track, behind, etc.). They are fundamentally different — probabilities capture uncertainty; status labels capture current evidence.

Tracker statuses describe current evidence, not probabilities; see /methodology/ §4.


Side-by-Side: Key Tournament Questions

The individual question rows have no confirmed direct dossier mapping. “Not tracked by this site” refers to that exact question, not to the broader topic. The category scorecard below links related dossiers without treating them as resolutions of these questions.

Technology & Capabilities

Neuralese Recurrence

SourceAssessment
AI 2027Non-text reasoning (“neuralese”) deployed in early 2027 — a key capability unlock
Metaculus crowdMarch 2026 snapshot — not refreshed: median ~Jan 2028 for public release (Q38598); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)Oct 2026 (more aggressive than crowd)

Analysis: The historical snapshot placed public release later than the scenario’s early-2027 deployment. Public release and internal deployment are different tests; this row does not resolve the related tracker dossier or establish the current crowd timeline.

VideoGameBench Performance

SourceAssessment
AI 2027Implies rapid capability gains across diverse tasks
Metaculus crowdMarch 2026 snapshot — not refreshed: median ~58% by Jan 2028 (range: 34%–83%) (Q38609); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)80% (more aggressive than crowd)

Analysis: The historical snapshot forecast strong but incomplete performance on a selected benchmark subset, not general task competence. No change since that snapshot has been verified here.

AI-Authored Scientific Papers

SourceAssessment
AI 2027AI increasingly contributes to research, approaching researcher parity
Metaculus crowdMarch 2026 snapshot — not refreshed: ~78% chance of an AI-authored paper at NeurIPS/ICML/ICLR before 2028 (Q38403); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)84%

Analysis: The historical snapshot assigned a high probability to conference authorship, a narrower institutional milestone than autonomous research capability. It is not a direct tracker confirmation or a current probability.

OpenAI Self-Improvement Risk Levels

SourceAssessment
AI 2027AI models gain enough capability to contribute to AI R&D, creating recursive loops
Metaculus crowdMarch 2026 snapshot — not refreshed: “High” median ~May 2027; “Critical” much later (Q38787); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)“High” by May 2027; “Critical” by July 2032

Analysis: Current group medians could not be verified. The March 2026 snapshot is historical context only and cannot support a claim of current agreement with the tracker.


Economics & Compute

Total AI Compute Capacity (Jan 2028)

SourceAssessment
AI 2027Massive compute scaling — trillions in infrastructure investment
Metaculus crowdMarch 2026 snapshot — not refreshed: No numerical estimate was recorded in the original comparison for capacity on January 1, 2028 (Q38419); forecaster count not verified.
Our trackerNot tracked by this site

Analysis: This question measures installed compute capacity in H100 equivalents, not annual investment, hardware ownership value or the size of a single training run. Without a verified numerical forecast, this row cannot support a quantitative comparison.

Compute Concentration

SourceAssessment
AI 2027”OpenBrain” (based on OpenAI/Google) dominates; others 3–9 months behind
Metaculus crowdMarch 2026 snapshot — not refreshed: No numerical estimate was recorded in the original comparison for the largest organisation’s share of compute on January 1, 2028 (Q38393); forecaster count not verified.
Our trackerNot tracked by this site

Analysis: This question concerns one organisation’s share of compute, not the size of its capability lead over competitors. We do not treat these as interchangeable tests; no numerical comparison is available here.


Geopolitics & Governance

US-China AI Agreement

SourceAssessment
AI 2027No meaningful cooperation — instead an accelerating race
Metaculus crowdMarch 2026 snapshot — not refreshed: ~4–5% chance of a formal US-China agreement before 2029 (Q38418); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)4%

Analysis: The historical snapshot assigned a low probability to a formal agreement; the current probability is unverified. The tracker does not assign a probability to this question, so there is no three-way numerical agreement.

AI as “Most Important Problem” (Gallup)

SourceAssessment
AI 2027Growing public concern, anti-AI protests, political salience
Metaculus crowdMarch 2026 snapshot — not refreshed: median ~19% by Jan 2028 (Q38407); forecaster count not verified.
Our trackerNot tracked by this site
Jan Wehner forecast (August 2025)3% (dramatically lower than crowd)

Analysis: The historical snapshot does not establish today’s crowd view. The category includes computers and technology, not only AI. Jan Wehner’s 3% is an August 2025 forecast, not a contemporaneous crowd comparison.


Category Scorecard

CategoryAI 2027 ScenarioMetaculus Crowd LeanOur Tracker Assessment
Coding automationVery aggressive (superhuman by Mar 2027)Moderate — expects progress but slowerCoding agents begin transforming professional work Confirmed (85%); SWE-bench-Verified score reaches 85% Confirmed (90%); Superhuman coder emerges Not Yet Testable (60%)
AI research accelerationCentral claim — recursive improvement by 2027Mixed measures; self-improvement dates are an unrefreshed March 2026 snapshotAI companies focus on AI-for-AI-research Confirmed (95%); AI R&D progress multiplier reaches 1.5× Emerging (75%); AI R&D progress multiplier reaches 2x Emerging (50%); AI R&D progress multiplier reaches 3× Not Yet Testable (40%); AI algorithmic progress multiplier reaches 4× (~2× overall progress) Not Yet Testable (35%)
Compute scalingMassive and rapidNo numerical forecast recorded in the March 2026 snapshot; not refreshedMassive AI infrastructure investment continues Confirmed (95%); Massive datacenter buildouts continue Confirmed (95%); 2×10²⁸ FLOP training run completed Not Yet Testable (50%)
Revenue/valuation$45B registered annual-revenue target, $2.5T–$3T valuation by end 2026Not directly tracked in tournamentLeading AI company reaches $45B annual revenue On Track (75%); Leading AI company reaches $3T valuation Behind (80%)
Public concernHigh — protests, political salience~19% (March 2026 snapshot — not refreshed)AI reshapes employment and intensifies public concern On Track (75%); Large-scale anti-AI protest (10,000+ people) Emerging (55%); A leading AI company’s net public approval falls to minus 35 points Not Yet Testable (75%)
Geopolitical raceIntense US-China competition, no cooperation~4–5% chance of agreement (March 2026 snapshot — not refreshed); not a tracker probabilityLeading US AI lab maintains a 3-9-month capability lead Behind (85%); Export controls impact Chinese AI compute Emerging (65%); Leading Chinese AI lab ~6 months behind US frontier On Track (75%); China nationalizes/centralizes AI research Emerging (65%)
Neuralese / novel architecturesKey unlock in early 2027Median ~Jan 2028 (March 2026 snapshot — not refreshed)Leading lab adopts high-bandwidth non-text reasoning (neuralese) internally Emerging (35%)
Safety / alignmentIncreasingly urgent, partially ignoredModerate concern acknowledgedFixing observed misconduct leaves failures in unfamiliar conditions Emerging (80%); AI systems behave differently when they recognize evaluation Emerging (75%); Weaker AI supervisors struggle to detect stronger systems’ misconduct Emerging (80%)

What Can Be Compared?

The historical neuralese snapshot placed public release later than the scenario’s early-2027 deployment, while the AI-authored-paper snapshot assigned a high probability to a narrower research milestone. Neither is a current crowd forecast or a direct grade of the scenario as a whole.

The historical agreement and public-salience figures do not establish today’s crowd view. We cannot verify a change since March 2026, and the tracker does not supply comparable probabilities for a claim of numerical agreement across all sources.

Tracker status labels evaluate evidence for specified claims; they are not crowd probabilities. The related category dossiers above remain useful context, but do not establish agreement on all eight tournament questions. None of the historical crowd rows should be used for a current timeline comparison.


Methodology Note

Metaculus forecasts change continuously. These figures are a March 2026 snapshot — not refreshed, not current community predictions. Their individual forecast timestamps and forecaster counts are unverified. The comparison includes an August 2025 analysis, so the March label does not establish that every underlying forecast was made in March. The two compute rows originally contained no numerical estimates, so none are supplied here. No forecaster count is presented as verified. A page’s retrieval date does not establish the date or freshness of its forecast.

The Jan Wehner forecasts are historical August 2025 context and have not been refreshed. Tracker statuses on this page match the linked dossiers as of 2026-11-09. For live forecasts and resolution criteria, open the linked questions or the Metaculus AI 2027 Tournament.

Our tracker assessments are evidence-based judgments updated as new information arrives. See our Methodology page for how we assign status labels and confidence scores.


Related:

The Metaculus tournament originally had 16 questions (as of August 2025) and has since expanded to 19. Not all questions map directly to our tracked predictions.