agi timeline
- User
agi timeline
- Assistant
AGI Timeline Predictions: From Experts to Surveys
Artificial General Intelligence (AGI) represents a system capable of performing the full spectrum of cognitive tasks at a level comparable to or exceeding humans. The question of when we might achieve AGI has generated diverse predictions from experts, researchers, and organizations.
Expert Predictions
Current predictions from leading AI researchers and industry figures show considerable variation:
-
Short-term predictions (2024-2030):
- Daniel Kokotajlo: 15% chance in 2024[1]
- Metaculus 25% Weak AGI estimate: February 2025[1]
- David Shapiro: 2026[1]
- Elon Musk: before 2026 ("discover new physics or invent new technology")[1]
- Dario Amodei (Anthropic CEO): by 2026[6]
- Masayoshi Son: 2027 or 2028[6]
- Jensen Huang (NVIDIA CEO): 2028 (a computer that can complete intelligence tests "fairly competitive" to a normal human)[1]
- Louis Rosenberg: by 2030[6]
-
Mid-term predictions (2031-2045):
- Ray Kurzweil: 2032 (updated from his previous 2045 prediction)[6]
- Sam Altman (OpenAI CEO): within the next four or five years[1]
- Demis Hassabis (DeepMind): by 2035[6]
- Yoshua Bengio (Turing award winner): 2028 to 2043 (90% confidence interval)[1]
-
Longer-term predictions:
- Geoffrey Hinton (Turing award winner): 2028 to 2053[1]
- Jürgen Schmidhuber: by 2050[6]
Major Research Surveys
Surveys of AI researchers provide a more systematic view of expert opinion:
- 2023 Expert Survey (2,778 AI researchers): High-level machine intelligence estimated by 2040[6]
- 2022 Expert Survey (738 experts): 50% chance of high-level machine intelligence by 2059[6]
- 2019 Forecasting Survey (296 AI experts): 50% chance of machines surpassing humans in 90% of economically relevant tasks before 2060[6]
- 2019 AI Experts Survey (32 experts): 45% predicted AGI before 2060, 34% after 2060, and 21% predicted it would never occur[6]
- 2018 Survey (165 AI experts): 50% predicted AI performing 99% of human-paid tasks before 2068[6]
- 2017 Survey (352 AI experts): 50% chance of AGI by 2060, with significant regional differences (Asian respondents: 30 years, North Americans: 74 years)[6]
- 2012/2013 Survey (550 participants): AGI probably between 2040 and 2050 (50% chance), very likely (90% chance) by 2075[6]
- 2009 Survey (21 AI experts): around 2050[6]
Community Predictions
Metaculus prediction platform shows more optimistic timelines from its community:
- When will AI pass a long, informed, adversarial Turing test? 2029[6]
- When will the first Artificial General Intelligence be developed? 2035[6]
- When will the first weakly general AI system be announced? 2026[6]
- When will the first general AI system be announced? 2030[6]
Recent Developments Affecting Timelines
Several recent developments have led some experts to reconsider their AGI timelines:
- OpenAI's o3 model reportedly passed the ARC-AGI-1 benchmark with 88% (high-compute version) and 76% (low-compute version), compared to the 80% average human score[8]
- Microsoft researchers published an evaluation of GPT-4 in 2023, concluding it "could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence system"[7]
- Geoffrey Hinton, who previously thought AGI was 30-50 years away, now estimates systems smarter than humans could appear within 5-20 years[7]
- Demis Hassabis stated in 2023 that he sees "no reason why progress would slow down" and expects AGI within a decade or even a few years[7]
Differing Perspectives on AGI
Opinions on AGI timelines generally fall into four camps:
- Near-term emergence: AGI could emerge within years or decades
- Long-term project: AGI will take a century or more
- Impossibility thesis: AGI may never be built due to fundamental limitations
- Already here in early form: Some claim current frontier models like GPT-4 represent early versions of AGI[7]
Historical Context
It's worth noting that past AI predictions were often overly optimistic:
- Herbert A. Simon in 1965: "machines will be capable, within twenty years, of doing any work a man can do"[7]
- Japan's Fifth Generation Computer Project in the 1980s set a ten-year timeline for AGI-like goals that were never achieved[7]
- Geoff Hinton claimed in 2016 that we wouldn't need radiologists by 2021 or 2026, which didn't happen[6]
Deployment Considerations
Even after AGI is achieved, deployment will face additional challenges:
- Large-scale AGI deployment throughout the business sector will likely follow validated emergence by at least 6 months[8]
- Initial deployment will likely occur in controlled operational settings within select government-approved organizations[8]
- The first deployed AGIs will likely take the form of collaborative AI agents fine-tuned for specific domains[8]
The timeline for AGI remains highly uncertain and contested. While the past suggests caution about overly optimistic predictions, recent advances in AI capabilities have caused many experts to revise their timelines forward, making the coming decades a critical period for AGI development.
-
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.