Will AGI (Artificial General Intelligence) be achieved by 2030?
Delphi’s forecast
- Yesyour pick4.3%$0.04
- No95.8%$0.96
This resolves too far out for us to open a market on it. The odds stand. You can still create this market yourself.
Forecast reasoning
No AI system has matched humans across most cognitive work as of October 2026, and frontier models still fail at long-horizon autonomous tasks despite rapid progress on narrow benchmarks. Expert predictions on AGI timelines remain unreliable and contradictory, with the Research Sentinel Institute finding no progress toward general intelligence across seven tracked problem areas. The 4% probability reflects that achieving AGI in just over 3 years would require an unprecedented acceleration bey...
Settlement
Resolution criteria: - "Yes": Wins if, strictly after 2026-09-30T18:47:46.787Z and at or before 2029-12-31T23:00:00.000Z, at least one major AI research organization (such as OpenAI, Google DeepMind, Anthropic, or a recognized national or international standards body) has formally and publicly declared that an AI system has achieved AGI — meaning it has been officially designated as a system capable of performing any intellectual task that a human can perform across a broad range of domains at human level or above — AND this declaration has been independently corroborated by at least one additional credible scientific institution or peer-reviewed publication. - "No": Wins if no such formal, publicly corroborated declaration of AGI achievement is made strictly after 2026-09-30T18:47:46.787Z and at or before 2029-12-31T23:00:00.000Z. Clarifications: - The window for qualifying events is strictly after 2026-09-30T18:47:46.787Z and at or before 2029-12-31T23:00:00.000Z. A corporate marketing claim, product announcement, or self-declared milestone from a single organization, without independent scientific corroboration, does not qualify. - "AGI achieved" means a formal declaration that an AI system can autonomously perform the full breadth of cognitive tasks a human can perform across diverse domains at human-level performance or above — not narrow task superiority, economic productivity thresholds alone, or performance on a single benchmark.
Dec 31, 2029, 11:00 PM UTC