What Evidence Would Prove That AGI Has Arrived?

**Possible standards**

- Broad benchmark performance

- Economic usefulness

- Autonomous research

- Transfer to unfamiliar tasks

- Reliable long-horizon planning

- Human-level judgement

**The definition problem**

AGI claims cannot be compared properly when each person uses a different threshold.

**Why AGI definitions differ**

Some people define AGI through human-level performance across cognitive tests. Others focus on economic work, autonomous scientific discovery, or the ability to learn unfamiliar tasks.

These definitions can produce very different timelines. A system may perform valuable work across many occupations while still failing simple reasoning tests.

**A stronger evidence package**

A credible AGI claim would likely require several forms of evidence rather than one benchmark:

- Broad performance across unfamiliar domains

- Reliable transfer to new tasks

- Long-horizon planning

- Learning from limited feedback

- Strong calibration

- Robust operation outside curated tests

- Meaningful economic usefulness

**Why the definition should come first**

The criteria should be stated before evaluating the system. Otherwise, supporters and critics can move the threshold after seeing the result.

AGI may also be a gradual economic and technical transition rather than one universally agreed announcement.

**Community question**

**What single piece of evidence would most strongly convince you that AGI had arrived?**

*This is independent WhatAI editorial coverage. AI Explained has not endorsed or sponsored this post.*