Three terms describe points along a range of AI capability.
Narrow AI
Systems built for specific tasks: spam filters, recommendation engines, image classifiers, game-playing programs. They can be superhuman in their domain but can't transfer skills elsewhere.
General AI
Systems able to perform well across a broad range of tasks, adapting to new ones without being rebuilt. Modern large language and multimodal models are much more general than earlier AI, which is why debates about AGI have intensified.
Superintelligence
A hypothetical system far exceeding the best human abilities in virtually every domain, including science, strategy and social skills. Some researchers think it could follow AGI relatively quickly if AI systems help improve AI; others are sceptical.
Blurry Boundaries
Real systems don't fall neatly into categories. A model may be superhuman at some tasks, human-level at others and weak at a few. Capability profiles are uneven.
Why the Distinctions Matter
Policy, safety research and business planning depend on which capabilities exist and which are emerging. Precise language helps: talk about specific capabilities rather than labels where possible.