When Will AGI Arrive? Expert Predictions Tracker
Nobody knows when artificial general intelligence will arrive — but the people building it, the researchers who pioneered it, and thousands of professional forecasters all publish timelines. This page tracks them side by side, next to our own news-driven estimate, so you can see who expects what and how far apart they are.
Our estimate is not an opinion poll: it is recalculated every day from analyzed AI news — model releases, benchmark results, funding, regulation and safety incidents — each scored for its impact on AGI progress and timelines. The full formulas and limitations are described on the methodology page.
Frontier lab leaders
| Who | AGI when | Position | Source |
|---|---|---|---|
| Sam Altman CEO, OpenAI | ~2026–2028 | "We are now confident we know how to build AGI as we have traditionally understood it" (Jan 2025). In March 2026 he put AGI roughly two years out, and OpenAI targets an automated AI researcher by March 2028. | blog.samaltman.com — Reflections as of March 2026 |
| Dario Amodei CEO, Anthropic | late 2026 – 2027 | Expects "powerful AI" — models at Nobel-laureate level across many fields, able to carry out long autonomous tasks — as soon as late 2026 or early 2027, while warning about civilization-level risks of getting the transition wrong. | Forbes, Jan 2026 as of January 2026 |
| Demis Hassabis CEO, Google DeepMind · Nobel laureate | ~2030 (3–10 years) | Has consistently framed AGI as 5–10 years away, tightening to 3–5 years in some 2025–2026 interviews. Says he and Amodei "don't disagree too much", but his timelines are "a little bit longer". | Interview digest, 2026 as of 2026 |
| Elon Musk CEO, xAI / Tesla | ~2026 | Predicted AI "smarter than the smartest human" by around 2026 — among the most aggressive public timelines from a lab owner. | AGI predictions survey (AIMultiple) as of 2024 |
Researchers & pioneers
| Who | AGI when | Position | Source |
|---|---|---|---|
| Geoffrey Hinton Turing Award laureate, "godfather of deep learning" | 2028–2043 | Moved from "30–50 years" before 2023 to "5–20 years" with wide error bars; puts a 10–20% probability on AI-caused human extinction within decades. | Position overview as of 2025 |
| Yann LeCun Cofounder, AMI Labs · Turing Award laureate | 10+ years, not via LLMs | Argues large language models are not a path to AGI: world models, persistent memory and planning are missing. Estimates "at least a decade and probably much more" along his research roadmap. Left Meta to put money behind the claim — AMI Labs raised $1.03B at a $3.5B valuation in March 2026 to build Joint Embedding Predictive Architecture (JEPA) world models that learn from reality rather than language, and says commercial applications may be years out. | TechCrunch, March 2026 as of March 2026 |
| Ray Kurzweil Futurist · Principal Researcher, Google | AGI 2029 · singularity 2045 | Has held the same forecast since 1999: human-level AI by 2029, human-machine merger ("the singularity") by 2045. Reaffirmed in "The Singularity Is Nearer" (2024). | AGI predictions survey (AIMultiple) as of 2024 |
Forecasting communities
| Who | AGI when | Position | Source |
|---|---|---|---|
| Metaculus community Aggregated forecast, "first general AI system" | ~2033 (50%) | As of February 2026 the community gives 25% probability of AGI by 2029 and 50% by 2033 (interval: May 2029 – December 2040). The forecast has oscillated with each frontier-model release. | Metaculus Q5121 as of February 2026 |
| Metaculus community Aggregated forecast, "weakly general AI" | mid-2028 | For a weaker bar — passing a set of general-capability benchmarks rather than full human replacement — the community median sits at June 2028. | Metaculus Q3479 as of February 2026 |
| Professional forecasters FutureSearch AGI timeline tracker · 14 named forecasters | clusters late 2025 – 2028 | Tracks 14 named forecasters and lab leaders (Kokotajlo, Amodei, Hassabis, Bengio, Jurkovic, Greenblatt, Lifland, Cotra, Wildeford and the Metaculus community among them) against a concrete bar: when most purely cognitive labor becomes automatable at better quality, speed and cost than humans. Estimates span early 2025 to 2046, but most cluster in late 2025–2028. The interesting part is the drift: the panel moved earlier through the ChatGPT era, later during the xAI/Meta/Gemini era, and — as of April 2026 — earlier again. | futuresearch.ai, updated April 2026 as of April 2026 |
Why the predictions disagree
Definitions differ. OpenAI defines AGI as "a highly autonomous system that outperforms humans at most economically valuable work"; Metaculus resolves on specific benchmark bundles; researchers like Yann LeCun mean human-level understanding of the physical world. A 2028 forecast and a 2040 forecast are often answering different questions.
Incentives differ. Frontier-lab leaders raise capital and talent on short timelines, while academics stake reputations on rigor. That doesn't make either side wrong — but it explains why lab CEOs cluster at 2026–2028 and skeptics at 2035 or later.
The path is contested. Scaling optimists believe current architectures plus compute get there; critics argue key components — world models, persistent memory, planning — are still missing, so no amount of scaling closes the gap.
That spread is exactly why we track the news instead of picking a side: every capability jump, deployment and safety incident moves our live index a little, and the trend is the signal. Our own estimate has drifted too — the full record of how far, and what moved it, is published rather than quietly revised.
AGI Timeline FAQ
What is AGI?
Artificial General Intelligence: an AI system that matches or exceeds human capability across most cognitive tasks — not just one domain like chess or coding. Today's frontier models are broad but still fail at long-horizon autonomy, which is why definitions (and timelines) vary.
When will AGI most likely arrive?
As of 2026, frontier-lab leaders say 2026–2028, aggregated community forecasts sit around 2033, and prominent skeptics say 2035 or later. Our own news-driven estimate currently points to July 26, 2029.
What is the technological singularity?
The hypothetical point where AI systems improve themselves faster than humans can follow, making the future hard to predict. Ray Kurzweil famously dates it to 2045; whether recursive self-improvement is possible at all remains an open question.
Is AGI dangerous?
Opinions range from "existential risk" (Hinton: 10–20% chance of catastrophe) to "overblown" (LeCun). We track the risk side daily as the Skynet Chance index, currently at 31.47%.