AGI timeline predictions: when experts, surveys and forecasters say AGI will happen

Key Findings

• The Metaculus community median for the first general AI system to be devised, tested and announced was September 24, 2030 on October 29, 2026, with an interquartile range of March 2028 to August 2036. (Metaculus: Date of Artificial General Intelligence)
• The same Metaculus AGI forecast moved from 2056 in February 2022 to 2038 in December 2022 and 2031 by December 2023, a 25-year shift in under two years. (Metaculus: Date of Artificial General Intelligence)
• In AI Impacts' December 2024 survey of 1,580 AI researchers, the aggregate forecast gave a 50% chance of high-level machine intelligence by 2042 and of full automation of labor by 2098. (AI Impacts: ESPAI 2024)
• The AI Impacts surveys' 50% year for high-level machine intelligence fell from 2061 (2016 survey) to 2059 (2022), 2047 (2023) and 2042 (2024), according to the 2024 survey paper. (AI Impacts: ESPAI 2024)
• In the 2023 AI Impacts survey of 2,778 researchers, machines beating humans at every task had a 50% chance by 2047, but full automation of all occupations only by 2116. (Grace et al., Thousands of AI Authors on the Future of AI)
• The AI 2027 scenario, published April 3, 2025, has a superhuman coder in March 2027 and artificial superintelligence in December 2027 in its racing ending. (AI 2027)
• AI 2027 co-author Daniel Kokotajlo's AGI median moved from 2027 (held since 2022) to 2028 in February 2025 and to December 2030 by January 2026, according to the AI Futures Project. (AI Futures Project: how our timelines changed)
• In a March 6, 2025 submission to the White House OSTP, Anthropic said it expects powerful AI systems with Nobel-level capabilities to emerge in late 2026 or early 2027. (Anthropic: Recommendations to OSTP)
• At Google I/O on May 20, 2025, Sergey Brin said AGI would arrive just before 2030 and Demis Hassabis said just after, Axios reported. (Axios: Hassabis and Brin at Google I/O)
• Geoffrey Hinton said in March 2023 that general-purpose AI might be 20 years or less away, after previously expecting 20 to 50 years, CBS News reported. (CBS News: Geoffrey Hinton interview)

Figures

AGI and superintelligence timeline predictions, by source and date

Why AGI definitions differ, and why it changes the date

There is no agreed definition of AGI (see what counts as AGI), and the rows in this table use at least six different finish lines:

• AGI with no formal definition, as in most CEO interviews. Elon Musk, for example, defined it in April 2024 as smarter than the smartest human.
• "Powerful AI" (Anthropic): smarter than a Nobel Prize winner across most fields, able to work autonomously for long periods.
• HLMI (AI Impacts surveys): unaided machines can do every task better and more cheaply than human workers.
• FAOL (AI Impacts surveys): every occupation is fully automatable. The same researchers give dates 50 to 100 years later for this than for HLMI, even though it describes a similar end state.
• Metaculus resolution criteria: a specific bundle of tests, including an adversarial Turing test and robotic tasks for the "general AI" question, and a lighter bundle for "weakly general AI".
• Superhuman coder, superhuman AI researcher and ASI (AI 2027): milestones defined by what share of AI research work the system can do.

A lab CEO saying "AGI in 2027" and a survey saying "HLMI in 2042" are partly measuring different things. The field guide to AGI forecasts explains the framing effects in more detail.

How AGI forecasts shifted after 2022

Every type of forecast in this dataset moved earlier between 2022 and 2024, then mostly held steady or moved slightly later in 2025 and 2026.

• Metaculus, general AI (Q5121): 2056 in February 2022, 2038 in December 2022 (weeks after ChatGPT), 2031 in December 2023, 2033 in May 2025 and September 2030 on October 29, 2026.
• Metaculus, weakly general AI (Q3479): 2042 in February 2022, November 2027 by December 2022, and August 2027 on October 29, 2026.
• AI Impacts surveys (HLMI, 50% year): 2061 in 2016, 2059 in 2022, 2047 in 2023 and 2042 in the survey run in December 2024. The 2024 paper calculates that the forecast horizon got about 3.4 years closer for each year that passed. Full automation of labor moved from 2164 (2022) to 2116 (2023) to 2098 (2024).
• Individual forecasters: Geoffrey Hinton went from 20 to 50 years to "20 years or less" in March 2023; Yoshua Bengio from "decades to centuries" to 5 to 20 years in August 2023; Ajeya Cotra from about 2050 to about 2040 in August 2022.
• Some later lengthening: the AI Futures Project's December 2025 model put full coding automation 3 to 5 years later than its April 2025 model, and Daniel Kokotajlo's AGI median moved from 2027 to December 2030. Their April 2026 update moved his automated-coder median back earlier, from late 2029 to mid-2028.

AI 2027: what the scenario predicts

AI 2027 is a scenario by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland and Romeo Dean of the AI Futures Project, published April 3, 2025. In its racing ending, a superhuman coder arrives in March 2027, a superhuman AI researcher in August 2027, a superintelligent AI researcher in November 2027 and artificial superintelligence in December 2027.

The scenario's dates were a modal path, not the authors' median forecast. Its timelines supplement gave all-things-considered superhuman-coder medians of 2030 (Eli Lifland), 2028 (Nikola Jurkovic) and 2033 (three FutureSearch forecasters). By January 2026 the authors' AGI medians were December 2030 (Kokotajlo) and January 2035 (Lifland).

Median predicted year by type of source

Medians are taken over the "Midpoint year" column of this dataset. Rows that give only an upper or lower bound ("by 2030", "as early as 2026") or no year have no midpoint and are left out. The milestones differ by row, so treat these as rough comparisons.

Methodology

Each row is one dated prediction from a named person, organization, survey, forecasting platform, scenario or model, verified on the cited page on October 29, 2026. Primary sources were used where they exist: the person's own essay or blog post, podcast transcripts, the company's policy submission, the survey paper and the scenario website. Where a statement is only available in a paywalled or video interview, the row cites reputable news coverage (Ars Technica, CNBC, Axios, Fortune, CNN, TechCrunch, The Verge, The Guardian, Politico, Business Insider, Decrypt, The Register, VentureBeat).

"Predicted year or range" keeps the source's own framing. When a source says "in five to 10 years", the row converts that to calendar years from the statement date and says so. "Midpoint year" is filled only when the source gives a single year, a median, or a two-sided range; it is the middle of the range (for example 2026.5 for "2026 or 2027"). Open-ended statements such as "by 2030", "20 years or less" or "as early as 2026" have no midpoint. No values were estimated beyond that arithmetic.

Metaculus values come from the recency-weighted community forecast history published on each question page (Q5121 for general AI, Q3479 for weakly general AI), read on October 29, 2026. The date in each Metaculus row is the date that forecast was current. The October 29, 2026 values match the medians displayed on the question pages that day.

AI Impacts survey figures are the aggregate (mixture) forecasts reported by the survey authors. The 2022 HLMI figure is 2059 on the AI Impacts results page and was reported as 2060 in the 2023 paper; the 2024 paper says small code and data-cleaning changes shifted it. Likewise, the 2023 FAOL figure is 2116 in the 2023 paper and 2115 when recomputed in the 2024 paper. The 2024 survey was fielded in December 2024 and published in September 2026.

AI Futures Project rows for Daniel Kokotajlo and Eli Lifland before 2026 come from the project's January 27, 2026 post, which lists each author's past AGI medians with dates and links to the original statements.

Sources disagree on one statement: for Elon Musk's January 22, 2026 Davos remarks, Decrypt and a published transcript report "2030 or 2031" for AI smarter than all of humanity, while Wired reported 2035. The table uses 2030-2031, which matches the transcript wording.

Excluded: statements whose wording could only be found in social media reposts or low-quality aggregator sites, and predictions that do not refer to a general or transformative capability milestone. Sergey Brin's and Demis Hassabis's May 2025 answers were to a before-or-after-2030 question, not a stated median.

Cite and Reuse This Dataset

The data is free to reuse under CC BY 4.0: copy it, chart it or build on it, with a link back to this page. Download the CSV or JSON.

Citation

agitracker.io (2026). AGI timeline predictions: when experts, surveys and forecasters say AGI will happen. Updated October 29, 2026. https://agitracker.io/data/agi-timeline-predictions/

HTML link

<a href="https://agitracker.io/data/agi-timeline-predictions/">AGI timeline predictions: when experts, surveys and forecasters say AGI will happen</a> (agitracker.io)

Frequently Asked Questions

When will AGI happen?

No one knows, and estimates depend on the definition. As of October 29, 2026, the Metaculus community forecast for the first general AI system is September 2030. AI researchers surveyed by AI Impacts in December 2024 gave a 50% chance of human-level machine intelligence by 2042. Lab leaders such as Dario Amodei, Elon Musk and Sam Altman have named dates between 2026 and 2030.

What is the AI 2027 prediction?

AI 2027 is a scenario from the AI Futures Project, published April 3, 2025. Its racing ending has a superhuman coder in March 2027 and artificial superintelligence in December 2027. The authors said 2027 was a plausible path, not their median: lead author Daniel Kokotajlo's AGI median was 2028 at publication and December 2030 by January 2026.

What do AI researchers predict for AGI?

In the largest recurring survey, run by AI Impacts, researchers who publish at top AI venues gave a 50% chance that machines outperform humans at every task by 2047 (2023 survey, 2,778 respondents) and by 2042 (December 2024 survey, 1,580 respondents). Asked instead when all occupations will be fully automatable, the same groups answered 2116 and 2098.

Have AGI predictions gotten earlier?

Yes, sharply between 2022 and 2024. The Metaculus general AI forecast moved from 2056 in February 2022 to 2031 by December 2023. The AI Impacts survey's 50% year for human-level machine intelligence moved from 2059 (2022) to 2042 (2024). Some forecasts lengthened slightly in 2025, and Metaculus sat at 2030 in October 2026.

What did Sam Altman predict about AGI?

In a Bloomberg Businessweek interview published January 5, 2025, Sam Altman said AGI would probably be developed during the U.S. presidential term that began that month. In September 2024 he wrote that superintelligence might arrive in a few thousand days, and in September 2025 he said he would be very surprised not to have extraordinarily capable models by 2030.

What is the difference between AGI, HLMI and superintelligence?

AGI usually means AI that matches humans across most cognitive work, but labs define it differently. HLMI, used in AI Impacts surveys, means unaided machines can do every task better and more cheaply than human workers. Superintelligence means AI far beyond the best humans at virtually all cognitive tasks. See what counts as AGI for five definitions compared.

Is the Metaculus AGI forecast reliable?

It is a crowd forecast from more than 1,800 forecasters, updated continuously, and forecasters are scored on resolved questions. Its main limitation is the resolution criteria: the question resolves on a specific set of tests, which may not match other definitions of AGI. The field guide to AGI forecasts covers the strengths and weaknesses of each method.

Research Using This Dataset

• AGI Is Always About Four Years Away: What 74 Predictions Show: Since 2023, AGI predictions put it a median 4.2 years out, and the predicted year keeps moving later. An analysis of 74 dated forecasts.

Related Datasets

• AI Model Release Timeline: Every Major LLM From GPT-1 to GPT-6 (2018-2026): Release dates, open-weight status, context windows, input types and launch API prices for 159 notable LLMs from 2018 to October 2026, each sourced to the lab.
• Biggest AI companies: most valuable private AI companies by latest valuation: The 45 most valuable private AI companies by latest round, each dated and sourced, plus the top public AI companies by market cap.

Version History

• October 29, 2026: First published with 74 records, each linked to its source.

Spotted an error or a missing record? Every row links to its source; corrections are made and dated on this page. See all datasets.