Artificial Intelligence · Projected milestone
Expert-level general AI
65% by 2030 · including earlier years
The year is an editorial timing estimate. The percentage applies to occurrence by the stated deadline. These estimates are not calibrated odds.
AI matches top human experts across most cognitive work; dependable rollout and widespread access are separate milestones.
Explore this milestone on the interactive mapWhat the forecast means
Milestone stage: demonstration
plausible-unconfirmed
The event-level broad expert capability may exist in a current frontier system under suitable elicitation; this remains unconfirmed.
Completion criteria: A published, independently inspectable evaluation across at least six materially different work domains finds expert-comparable accepted outputs in a majority of domains, on fresh work with unsuccessful attempts, human intervention and cost disclosed. No single benchmark score, lab label or adoption percentage resolves the event.
This page describes a forecast, rather than an achieved result. Read the methodology and assumptions.
Built on
- EUV lithographyRecorded · 2019requires
- 3D-stacked chipletsRecorded · 2022enables
- AI supercomputer campusesRecorded · 2024enables
- Reasoning modelsRecorded · 2024enables
- AI research partnersRecorded · 2025influences
- AI settles an 80-year-old Erdős problemRecorded · 2026enables
- Gigawatt AI campusesForecast · ~2027enables
Led to
- Hardware key for your AI agentsForecast · ~2027influences
- AI food scientistForecast · ~2029enables
- AI-written bestsellerForecast · ~2029enables
- Automated alignment researchersForecast · ~2029influences
- Billion-dollar one-person companyForecast · ~2029enables
- Chips designed end to end by AIForecast · ~2029influences
- Robots learn a job in a dayForecast · ~2029influences
- Products designed end to end by AIForecast · ~2030enables
- Publishers run by AI agentsForecast · ~2030enables
- Rapid storm strengthening predictedForecast · ~2030enables
- Self-running back officeForecast · ~2030enables
- Autonomous surgical robotsForecast · ~2031enables
- Buildings that pre-cool by forecastForecast · ~2031enables
- Four-day week in lawForecast · ~2031enables
- Private general AI on your own deviceForecast · ~2031requires
- AI managers for human teamsForecast · ~2032enables
- Lights-out AI-run chip fabForecast · ~2032influences
- Riemann hypothesis provedForecast · ~2032enables
- Superhuman AI scientistsForecast · ~2032influences
- AI answers outsell search adsForecast · ~2033enables
- AI home clinic kitForecast · ~2033enables
- Self-updating world mapForecast · ~2033enables
- AI-founded fields of mathematicsForecast · ~2034enables
- Limits on AI persuasion in adsForecast · ~2034enables
- Most Millennium Prize problems solvedForecast · ~2034enables
- Most office work done by AIForecast · ~2035enables
- Personal AGI within reach of almost everyoneForecast · ~2035influences
- Verifiably aligned AIForecast · ~2035influences
- Robot plumbers and electriciansForecast · ~2036influences
- Robot workers on the MoonForecast · ~2037influences
- Every land parcel on the mapForecast · ~2038enables
- AI judges with binding rulingsForecast · ~2040enables
- Self-driving bike-lane podForecast · ~2044enables
- Agent-run commerceForecast · ~2045enables
- Food waste halvedForecast · ~2045enables
- Mostly autonomous militariesForecast · ~2045enables
- Robots beat World Cup championsForecast · ~2045enables
- AI welfare standardsForecast · ~2048influences
- Human driving becomes rareForecast · ~2052enables
- Animal feelings measuredForecast · ~2058enables
Sources & evidence
- Research acceleration: The view inside OpenAI 2026-09-06
OpenAI reports an automated research intern, increasing agent contribution and substantial human intervention on longer tasks; it targets an automated researcher in March 2028. Internal measurements are preliminary, and compute and controls still constrain work.
- Measurements for understanding the pace of AI development inside frontier labs 2026-09-17
Anthropic reports AI leading 26% of R&D in August, no fully autonomous measured R&D subset, and about 30,000 concurrent internal agents. It reports complete action-monitor coverage, which is not complete detection. The measurements use internal/model judgments.
- Sharing AI progress in mathematics 2026-10-06
OpenAI shares research-level mathematical results, proof artifacts and evaluation details; the post was updated October 15. The claims still need ordinary mathematical scrutiny. These results support capability beyond benchmark scores, not mastery of every scientific field.
- Time horizons 2026-09-08
The page was last updated September 8 and is no longer actively updated. Its horizons measure expert-time difficulty at selected success rates on a software/ML/cyber task suite; estimates beyond roughly sixteen hours are unreliable. Missing new systems are a coverage gap.
Record reviewed 2026-10-18