AGI (Artificial General Intelligence)
AGI (artificial general intelligence) describes a theoretical level where a machine can transfer knowledge across domains the way humans do, solving new problems without retraining for every scenario.
Current models are still narrow compared to this definition: they perform well in language, code, and multimodal tasks but often lack a robust world model and long-term autonomous behavior. See the difference vs ASI and the core concept artificial intelligence.
Key characteristics
- Describes hypothetical general intelligence, not a product you can deploy today.
- Relies on transferring knowledge across domains without retraining for each task.
- Is often used in strategic discussions about future capability, safety, and AI governance.
Example
A model that is strong in law but weak in physics is still "narrow AI."
By the most common definition, AGI would switch between law, physics, economics, and programming at comparable quality without retraining for each domain.