When will AGI arrive? A timeline roundup
"Image synthesis assisted by Reve 2.1, an AI partner within the Global Future Nexus ecosystem."
From DeepMind co-founder Shane Legg's 17-year-old prediction to Metaculus forecasters and expert surveys, the aggregate evidence points to a clear signal: the window for AGI is narrowing fast, with multiple independent methods now converging on the late 2020s as the critical threshold.
A Prediction That Refuses to Budge
Shane Legg, co-founder and Chief AGI Scientist at Google DeepMind, has held the same forecast since 2009: a 50% chance of achieving "minimal AGI"—an AI that can perform the typical cognitive tasks of a human—by 2028. As he put it in a December 2025 podcast: "I've held the same prediction for well over a decade: there's a 50% chance we'll see AGI by 2028" . In a field where goalposts shift with each breakthrough, Legg's consistency is remarkable. He defines minimal AGI as an AI that can complete the typical cognitive tasks a human can perform, while full AGI would encompass the full range of human cognitive capabilities, including inventing new theories and creating original artwork.
The Aggregate Forecasts
Beyond individual predictions, broader surveys tell a similar story. Aggregate forecasts, led by the nonprofit organisation AI Impacts, give at least a 50% chance of AI systems achieving several AGI milestones by 2028. These milestones include autonomously constructing a payment processing site from scratch, creating a song indistinguishable from a new song by a popular musician, and autonomously downloading and fine-tuning a large language model.
The same survey estimates a 10% chance of unaided machines outperforming humans in every possible task by 2027, rising to 50% by 2047. As MIT Technology Review observed, time horizons have shortened dramatically—from 50 years at the time of GPT-3's launch to just five years by the end of 2024. The consistency of this shift across independent methods strengthens the signal that AGI is approaching faster than many anticipated.
The Forecasting Platforms
Prediction markets and forecasting platforms, which aggregate the wisdom of thousands of participants, corroborate the trend. Metaculus, a reputation-based forecasting platform with over one million predictions, shows AGI probability at 25% by 2027 and 50% by 2031. In just four years, AGI predictions on Metaculus compressed from 50 years away to five.
Other platforms paint a similar picture. Kalshi contributors assigned a 40% probability to OpenAI achieving AGI by 2030, while Polymarket placed the probability of AGI by 2027 at 9%. Prediction markets put roughly 50% odds on AGI by 2028–2031 and 90% odds stretching toward mid-century.
The Industry Titans
The industry's leaders have aligned their public statements with these forecasts, though the spectrum of views remains wide.
Elon Musk of xAI has predicted "AGI in 2026". Dario Amodei of Anthropic has forecast 2026–2027 for powerful AI. Sam Altman of OpenAI has placed AGI around 2027–2028. Demis Hassabis of DeepMind has given a 50% probability by 2030. Yann LeCun of Meta has projected a timeline of 2030–2035.
At Davos 2026, Hassabis said he still sees about a 50% chance of AI systems that match all human cognitive capabilities by the end of the decade, while Amodei suggested the timeline could be shorter, driven by AI increasingly being used to build more powerful AI. Amodei has predicted that powerful AI systems will likely emerge in late 2026 or early 2027.
The Expert Divergence
Despite the convergence toward near-term timelines, significant disagreement remains. A RAND Corporation analysis published in March 2026 found that substantial disagreement persists even when definitions and information are held constant: people with similar training, working in the same organisations and looking at the same data, often reach very different conclusions about timelines and risk.
Some forecasts stretch far into the future. Daniel Kokotajlo of the AI Futures Project, who predicts a 50% chance of superintelligence by 2028, stands at one extreme. Others, including Eli Lifland, project AGI after 2045. The difference between the 90th and 10th percentiles in some surveys spans over a century.
The GFN Context: Preparing for a Range of Futures
For Global Future Nexus, the diversity of AGI timelines is not a reason to wait—it is a reason to act now. As the RAND analysis concludes, the policy question is not "when will AGI arrive?" but "how should we prepare for a range of possible AI futures?" . Effective strategy under such uncertainty requires three qualities: flexibility to pursue different objectives as circumstances evolve, adaptiveness to respond to unanticipated developments, and robustness to shocks.
The convergence of multiple independent forecasting methods on the late 2020s to early 2030s—Legg's 2028, Metaculus's 2031, the expert surveys' 2028 milestones—sends a clear signal. AGI is not a distant horizon. It is approaching faster than most institutions are prepared for. GFN's work on anticipatory governance, cross-species trust, and AGI identity frameworks provides the infrastructure for navigating this uncertainty—not by predicting the exact date, but by building the capacity to respond to whatever future arrives.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
