Appendix Table 2: All AI Progress Questions

Question nameCategorySourceDate askedIncluded in Table 1?Included in post?
AI wins IMO gold medalCapabilitiesXPT2022YesYes
MATH dataset benchmarkCapabilitiesXPT2022YesYes
MMLU benchmarkCapabilitiesXPT2022YesYes
QuALITY benchmarkCapabilitiesXPT2022YesYes
Date of first publicly known advanced AICapabilitiesXPT2022NoNo
Nick Bostrom affirms existence of AGI by 2030CapabilitiesXPT2022NoNo
ImageNet classification training efficiencyCapabilitiesXPT2022NoNo
GPT revenue stays under $1bAdoptionXPT2022YesNo
Cost of compute of largest AI experimentAdoptionXPT2022YesYes
US computer R&D spendingAdoptionXPT2022YesYes
Maximum compute used in an AI experimentAdoptionXPT2022YesNo
Lowest price of 1 GFLOPSAdoptionXPT2022YesYes
Largest number of parameters in an ML modelAdoptionXPT2022NoNo
US GDP from software and information servicesImpactXPT2022YesYes
OECD labor force participation rateImpactXPT2022YesYes
Probability of GDP growth over 15%ImpactXPT2022NoNo
Negative public opinion of AIImpactXPT2022YesYes
AI catastrophic riskImpactXPT2022NoNo
AI extinction riskImpactXPT2022NoNo
NYT bestsellers written by AIImpactXPT2022NoNo
AI out-forecasts a team of superforecastersCapabilitiesRoots of Disagreement on AI RiskApr–May 2023YesYes
AI solves a famous open math problemCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
AI causes 10,000+ deaths or $10b+ in damageImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
Two AI-caused catastrophes of escalating severityImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
Global corporate AI investment falls below 25% of its 2021 levelAdoptionRoots of Disagreement on AI RiskApr–May 2023NoNo
Evidence of AI self-replication and shutdown evasionCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
US real GDP growth averages 4% or more (2023–2030)ImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
100x drop in compute for state-of-the-art AI within one monthCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
US implements 4+ of Muehlhauser's 12 AI policy ideasGovernanceRoots of Disagreement on AI RiskApr–May 2023NoNo
US requires testing of AI systems before deploymentGovernanceRoots of Disagreement on AI RiskApr–May 2023NoNo
Enforced six-month international pause on AI developmentGovernanceRoots of Disagreement on AI RiskApr–May 2023NoNo
LLM that verifiably never describes violenceCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
Unprompted AI writes and deploys improved AI softwareCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
Peer-reviewed demonstrations of convergent instrumental subgoalsCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
A top-four frontier AI lab is based in a non-democracyGovernanceRoots of Disagreement on AI RiskApr–May 2023NoYes
AI solely authors top-journal review articles and top-10 appsCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
Frontier AI model sustains deception of human expertsCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
IT and communications exceed 70% of the S&P 500ImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
AI autonomously influences a democratic election resultImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
Global real GDP exceeds 115% of any prior year's peakImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
Two AI-caused catastrophes of escalating severity (platform version)ImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
US implements 5+ of Muehlhauser's 12 AI policy ideasGovernanceRoots of Disagreement on AI RiskApr–May 2023NoNo
ARC Evals finds AI able to self-replicate and evade shutdownCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
Robot assembles a scale-model car from written instructionsCapabilitiesRoots of Disagreement on AI RiskApr–May 2023NoNo
AI is the "most important problem" for 15%+ of AmericansImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
Twofold partisan gap over AI as a top US policy priorityImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
US–China intergovernmental AI safety organizationGovernanceRoots of Disagreement on AI RiskApr–May 2023NoNo
Developers shut down an AI system after power-seeking behaviorImpactRoots of Disagreement on AI RiskApr–May 2023NoNo
AI reliably writes bug-free code of 10,000+ linesCapabilitiesRoots of Disagreement on AI RiskApr–May 2023YesYes
VCT: probability that AI matches the median virologistCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
VCT: AI matches or outperforms expertsCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
AI outperforms experts on long-form biothreat questionsCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
Probability of large-scale human-caused epidemic conditional on AI capabilitiesImpactLLM BioriskLate 2024 / early 2025NoNo
Influenza synthesis: AI enables 10% of non-experts to synthesize influenza virusCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
Influenza synthesis: AI enables 50% of non-experts to synthesize influenza virusCapabilitiesLLM BioriskLate 2024 / early 2025NoNo
Bioweapon attack planning: AI enables plausible planCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
Dual-use DNA: AI enables 90% success rateCapabilitiesLLM BioriskLate 2024 / early 2025YesYes
AI Cyber: year AI solves >90% of tasks on CybenchCapabilitiesAI Cyber2025YesYes
AI Cyber: year AI enables 25% of moderately skilled individual hackers to find vulnerabilities and write elite exploitsCapabilitiesAI Cyber2025YesNo
AI Cyber: year AI enables novice hackers to match an elite team in an operational technology capture-the-flag-style competitionCapabilitiesAI Cyber2025YesYes
AI Cyber: AI-enabled actor first causes a $100m+ grid blackoutImpactAI Cyber2025YesYes
Active Site RCT: forecasting internet-only armCapabilitiesActive Site RCT2025NoNo
Active Site RCT: forecasting internet-plus-LLM armCapabilitiesActive Site RCT2025YesYes
State-of-the-art AI accuracy on FrontierMath benchmarkCapabilitiesLEAP W1Jun–Aug 2025YesYes
Autonomous vehicles: share of ride-hailing tripsAdoptionLEAP W1Jun–Aug 2025YesYes
Change in US jobs by occupational classCapabilitiesLEAP W1Jun–Aug 2025YesNo
General AI progressImpactLEAP W1Jun–Aug 2025NoNo
AI's impact on the Technological Richter ScaleImpactLEAP W1Jun–Aug 2025NoYes
AI solves or substantially assists a Millennium Prize ProblemCapabilitiesLEAP W2Aug–Sep 2025YesYes
Share of science publications that are AI-engagedAdoptionLEAP W2Aug–Sep 2025NoYes
Share of new drug sales from AI-discovered drugsAdoptionLEAP W2Aug–Sep 2025YesYes
Share of US electricity used to train and run AIAdoptionLEAP W2Aug–Sep 2025YesYes
Annual global private AI investmentAdoptionLEAP W3Sep–Oct 2025NoNo
Share of US work hours assisted by generative AIAdoptionLEAP W3Sep–Oct 2025YesYes
Benchmark gap: open- vs. closed-weight frontier modelsCapabilitiesLEAP W3Sep–Oct 2025YesYes
Share of US adults using AI for companionship dailyAdoptionLEAP W3Sep–Oct 2025YesYes
Annual recurring revenue of the top AI companyAdoptionEEAIOct 2025–Feb 2026YesYes
Change in GDPImpactEEAIOct 2025–Feb 2026NoNo
Change in labor productivityImpactEEAIOct 2025–Feb 2026NoNo
Change in total factor productivityImpactEEAIOct 2025–Feb 2026NoNo
Labor force participation rateImpactEEAIOct 2025–Feb 2026NoNo
Unemployment rateImpactEEAIOct 2025–Feb 2026NoNo
Youth unemployment rateImpactEEAIOct 2025–Feb 2026NoNo
Share of job categoriesImpactEEAIOct 2025–Feb 2026NoNo
Change in employment by occupationImpactEEAIOct 2025–Feb 2026NoNo
Wealth inequalityImpactEEAIOct 2025–Feb 2026NoNo
Labor shareImpactEEAIOct 2025–Feb 2026NoNo
Median household incomeImpactEEAIOct 2025–Feb 2026NoNo
Life satisfactionImpactEEAIOct 2025–Feb 2026NoNo
Work hours assisted by Gen AIAdoptionEEAIOct 2025–Feb 2026NoNo
AI electricity consumptionAdoptionEEAIOct 2025–Feb 2026NoNo
State-of-the-art accuracy on LiveCodeBench Pro HardCapabilitiesLEAP W4Nov–Dec 2025YesYes
Valuation per employee at the top five AI companiesAdoptionLEAP W4Nov–Dec 2025YesNo
Entry-level share of new hires at top 15 tech companiesImpactLEAP W4Nov–Dec 2025NoYes
Year three $10b companies have five or fewer employeesImpactLEAP W4Nov–Dec 2025YesNo
Global hyperscale data center capacityAdoptionLEAP W4Nov–Dec 2025YesYes
AI-coauthored AI paper judged worthy of a Test-of-Time awardImpactLEAP W4Nov–Dec 2025NoNo
US–China frontier-model gap on the Epoch Capabilities IndexCapabilitiesLEAP W5Jan–Feb 2026YesYes
Manufacturing origin of frontier AI training chipsGovernanceLEAP W5Jan–Feb 2026NoYes
NATO authorization of fully autonomous cyber weaponsGovernanceLEAP W5Jan–Feb 2026NoYes
Regulatory fines on AI-generated contentImpactLEAP W5Jan–Feb 2026NoNo
ForecastBench: year AI beats superforecastersCapabilitiesLEAP W5Jan–Feb 2026YesNo
US real GDP growth (annualized)ImpactLEAP W6Feb–Mar 2026YesYes
US labor force participation rateImpactLEAP W6Feb–Mar 2026YesYes
Wealth inequality: top 10% share of US wealthImpactLEAP W6Feb–Mar 2026YesYes
Industrial robots in use globallyAdoptionLEAP W7Mar 2026NoYes
US labor shareImpactLEAP W7Mar 2026NoNo
Year a robot passes the Coffee Test in an unfamiliar homeCapabilitiesLEAP W7Mar 2026NoYes
Year a robot performs an appendectomy without physical human helpCapabilitiesLEAP W7Mar 2026NoNo
Amazon e-commerce sales per distribution employeeAdoptionLEAP W7Mar 2026NoNo
AI's impact on the Technological Richter Scale (second asking)ImpactLEAP W8Apr–May 2026NoYes
Year a majority of LEAP panelists agree AGI existsImpactLEAP W8Apr–May 2026NoNo
State-of-the-art METR 80% time horizonCapabilitiesLEAP W8Apr–May 2026YesYes
METR 80% time horizon: AI completes 8-hour tasksCapabilitiesLEAP W8Apr–May 2026YesNo
Rapid progress scenario timelinesImpactLEAP W8Apr–May 2026NoNo
AI's overall impact on the US over 20 yearsImpactLEAP W8Apr–May 2026NoNo
Year the US, UK, or EU requires pre-approval of frontier AI modelsGovernanceLEAP W9May–Jun 2026YesYes
Cybercrime losses reported to the FBIImpactLEAP W9May–Jun 2026NoYes
AI catastrophic riskImpactLEAP W9May–Jun 2026NoNo
First major AI global harm eventImpactLEAP W9May–Jun 2026NoNo
Top-10 economies rated "authoritarian" on the EIU Democracy IndexImpactLEAP W9May–Jun 2026NoNo
Deaths from non-eradicable diseasesImpactLEAP W10Jun–Jul 2026NoNo
US life expectancyImpactLEAP W10Jun–Jul 2026NoYes
Global real GDP growthImpactLEAP W10Jun–Jul 2026NoNo
Consumer value of generative AIAdoptionLEAP W10Jun–Jul 2026NoNo
Global average happinessImpactLEAP W10Jun–Jul 2026NoNo
Frontier AI company independenceImpactLEAP W11Jul–Aug 2026NoNo
Frontier AI company revenueAdoptionLEAP W11Jul–Aug 2026YesYes
AI-exposed equity performance: software sectorAdoptionLEAP W11Jul–Aug 2026NoNo
AI-exposed equity performance: semiconductor sectorAdoptionLEAP W11Jul–Aug 2026NoNo
Data center investmentAdoptionLEAP W11Jul–Aug 2026NoNo