AGI refers to a hypothetical artificial intelligence system capable of understanding, learning, and applying knowledge in any intellectual task that a human being can perform, with the same flexibility and adaptability. It differs from current AI, called narrow AI, because the latter excels in specific domains—playing chess, translating texts, generating images—but does not transfer that knowledge to different problems.
Its relevance is both technical and cultural. An AGI would profoundly alter the economy, science, and work, which is why it centers debates on safety, control, and alignment with human values. It is useful to distinguish several related concepts:
- Narrow AI: specialized in specific tasks (what exists today).
- AGI: general competence comparable to human levels.
- Superintelligence (ASI): capability that would vastly surpass human capability.
As of today, AGI does not exist, and disagreement persists regarding whether it is achievable, when, and through which methods. Many recent advances, such as language models, are sometimes presented as steps toward it, although they remain specialized systems without real understanding or intentionality.