Narrow AI performs a defined task or limited set of tasks, while general AI would apply intelligence flexibly across unfamiliar tasks at a human-like level. Existing AI systems are narrow; general AI remains proposed rather than established.
This is IB Digital Society syllabus area 3.6A, Types of AI, shared by SL and HL.
| Type | Capabilities | Limitations and status |
|---|---|---|
| Narrow AI, also called weak or domain-specified AI | Processes patterns within a particular domain, such as generating text, recognizing faces or recommending content | Cannot independently transfer its abilities to every unrelated problem. Currently deployed AI systems belong to this broad category. |
| General AI, also called artificial general intelligence | Would transfer knowledge between domains, adapt to unfamiliar situations and perform varied cognitive tasks at approximately human level | No system has demonstrated reliable general intelligence. Its feasibility, measurement and implications remain contested. |
| Strong or full AI | Often used as a synonym for general AI | It may instead imply consciousness or genuine understanding, so its meaning depends on the source. |
The mechanism behind narrow AI is specialization. A system is developed using particular data, objectives and methods, so its outputs reflect that design context. Even when one system produces text, images and code, varied output does not prove general intelligence or flexible, human-level understanding.
A common misconception is that fluent, human-like output proves that a system thinks or understands. It does not. Avoid anthropomorphic claims; explain instead that AI systems process patterns in data according to human-designed objectives.
In an exam, a distinguish question requires clear differences referring to both types throughout. For an explain question, connect specialization and transferability to the distinction. When evaluating future implications, treat general AI as uncertain and contested, not as an existing fact.
