iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Artificial general intelligence (AGI) is a proposed kind of AI that can perform intelligently across many different domains and learn new skills. Some definitions use human-level intellectual ability as the benchmark. There is no universally accepted test or threshold for AGI, so the term describes an ambition rather than a settled technical or legal category.
What does artificial general intelligence mean?
AGI refers to an AI system intended to handle a broad range of intellectual tasks, rather than being limited to one narrow function. The idea commonly combines two qualities: capability across multiple domains and the ability to learn new skills. Some definitions also compare the system’s abilities with human intelligence.
A 2021 UNESCO publication gives this formulation, attributed to ARTICLE 19 and Privacy International (2018): “The overarching, and as yet unachieved, goal of a system that displays intelligence across multiple domains, with the ability to learn new skills, and which mimic or even surpass human intelligence.” UNESCO’s AI terminology publication also records a usage of “strong AI” or AGI that involves consciousness and human-like responses. That difference matters: not every definition makes consciousness a requirement.
The OECD’s 2024 paper characterizes AGI as machines with human-level intelligence or greater, while noting that the concept is controversial. This is a description, not a binding test adopted across the field. The OECD paper therefore supports a broad characterization, not a claim that experts have settled on one definition.
#1 Best Overall
Why is there no single AGI test?
Definitions place different weight on breadth, learning, performance, and the meaning of intelligence. A system might perform well across many benchmark tasks but still be evaluated differently from one expected to acquire new skills in unfamiliar situations. A definition that requires human-level performance also sets a different bar from one that asks only for broad competence.
The European Commission says it is infeasible to give a precise list of tasks or capabilities needed to classify a model as general-purpose, because capabilities and use cases vary. That guidance concerns the EU’s legal category of general-purpose AI models—not a definitive AGI test. The Commission’s guidance is useful context for why broad AI categories can be difficult to draw precisely, but it does not establish an official AGI checklist.
Rank #2
When you encounter a claim that a system is AGI, ask what definition and evidence standard the speaker is using. A meaningful claim should make clear:
- Breadth: Is the system expected to work across several domains, a wide range of distinct tasks, or nearly all intellectual tasks?
- Learning: Must it acquire new skills, or is competence on a broad set of existing tasks enough?
- Performance: Is the comparison human-like, human-level, or beyond human capability?
- Intelligence: Does the definition mean broad task competence, or does it also require consciousness?
- Category: Is the source discussing AGI as a research ambition, or using a separately defined legal term such as “general-purpose AI model”?
AGI and the EU’s general-purpose AI categories are different
The EU AI Act defines “general-purpose AI model” and “general-purpose AI system”; it does not define AGI. Under Article 3(63), a general-purpose AI model is one that displays significant generality, can competently perform a wide range of distinct tasks, and can be integrated into varied downstream systems or applications. The definition includes an exception for models used only for research, development, or prototyping before they are placed on the market. The text of Regulation (EU) 2024/1689 sets out that legal scope.
Article 3(66) separately defines a general-purpose AI system as a system based on a general-purpose model that can serve various purposes, either directly or by being integrated into other systems. The distinction is important: a model is not automatically the same thing as the system built around it. Recital 97 explains that a model may need additional components, such as a user interface, to become an AI system. The Act’s recital and definitions address these model-versus-system distinctions.
| Term | What it refers to | What it does not establish |
|---|---|---|
| AGI | A contested, hypothesized level of AI capability, commonly involving broad competence and sometimes learning new skills or human-level performance. | A universally agreed operational test or legal status. |
| General-purpose AI model | A category defined by the EU AI Act, based on significant generality, competent performance across distinct tasks, and potential integration into varied applications. | That the model is AGI. |
| General-purpose AI system | A system based on a general-purpose AI model that can serve varied purposes directly or through integration. | That the model and the complete system are interchangeable terms. |
Is any current AI system definitively AGI?
There is no universally accepted threshold in the definitions cited here for deciding that a system has crossed into AGI. The sources also do not provide a system-by-system determination. It is more accurate to ask which definition someone is applying and what evaluations support the claim than to treat “AGI” as a label with one settled pass-or-fail test.
UNESCO’s 2021 publication described AGI as an unachieved goal at the time it was published. That statement is time-bound; it should not be recast as a timeless verdict on every system developed since. Likewise, the EU’s general-purpose AI classifications should not be used as proof that a model is—or is not—AGI.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
