The Biggest Questions About AI
The map · 1 Trajectory · 1.2 General intelligence and missing capabilities · 1.2.5

AGI recognition

What should count as human-level or general intelligence, and would we recognize it before its consequences became obvious?

Definitions have drifted from Turing tests to economic thresholds to capability profiles. Without agreed criteria, arrival may be declared only in retrospect — after the consequences have already begun compounding.

View on the map → · Open in Browse →

What changed
October 2025–August 2026 · swept August 3, 2026 · editorial review pending
01

2026 is running the recognition experiment in public: Huang declared AGI achieved, Sequoia declared it functionally arrived, and the one legally binding AGI definition ever written — the OpenAI–Microsoft clause — was quietly deleted rather than adjudicated. The sharpest scholarly move argues the leading definition's human-psychometric foundation is the wrong ruler.

Recent thinking
4 featured from 6 tracked · October 2025–August 2026 · all 6 chronologically →
Tidler, Tešić, Pacchiardi, Hernández-Orallo et al. · AI Evaluation (Substack) · 31 Oct 2025 essay
Is the Definition of AGI a Percentage?
CHC is therefore well suited to describing variation among typical adult humans, but much less for children, neurodivergent people, other species, or AI systems.

The standing scholarly rebuttal to the canonical 'A Definition of AGI' paper: grounding AGI in a human psychometric model produces 'hugely miscalibrated expectations', undermining the 'GPT-5 is 57% of the way to AGI' framing.

Simon Willison · simonwillison.net · 27 Apr 2026 essay
Tracking the history of the now-deceased OpenAI Microsoft AGI clause
Revenue share payments from OpenAI to Microsoft continue through 2030, independent of OpenAI's technology progress

Definitive documented history of the only legally binding AGI definition ever written — from charter language to the secret $100B-profits test to its quiet April 2026 deletion. When real money required recognizing AGI, no workable criterion survived.

Jeremy Kahn · Fortune · 30 Mar 2026 news
Nvidia's Jensen Huang says 'We've achieved AGI.' But no one can agree on what that means.
AGI has long been the ultimate goal of many artificial intelligence researchers. That's been the case even though there is no universally accepted definition of the term.

Well-reported survey of the 2026 definitional chaos: cognitive-taxonomy, economic, and vibes-based definitions collide as arrival is claimed and denied simultaneously — the recognition problem playing out in real time.

Pat Grady & Sonya Huang · Sequoia Capital · 14 Jan 2026 essay
2026: This is AGI
Long-horizon agents are functionally AGI, and 2026 will be their year.

Prominent capital-allocator declaration that AGI has effectively arrived under a functional definition — a live example of definitional drift toward whatever current systems can do.

Additional relevant discussion (2)
Is AGI the Right Goal for AI? — Gary Marcus · Marcus on AI · 16 Oct 2025
Announcing FrontierMath Erdős — Tom Adamczewski and Greg Burnham · Epoch AI · 1 Sep 2026
Foundational reading (3)Levels of AGIMorris et al., DeepMind · 2023A Definition of AGIHendrycks et al. · 2025o3 scores 87.5% on ARC-AGI — a step-change in novel-task adaptation, and still not AGIFrançois Chollet (@fchollet) on X · 2024
Previous1.2.4 Learning efficiencyNext1.2.6 Timelines