AI Types of AI

AI is often grouped into three categories by how broad its capability is — narrow AI, which is what exists today, and general or super AI, which remain hypothetical.

Three Categories, Sorted by Scope

One common way to classify AI has nothing to do with which technique a system uses, and everything to do with how broad its capability is. Under this view there are three categories: narrow AI, general AI, and superintelligence.

CategoryCapabilityExists Today?
Narrow AI (ANI)Performs one task or a limited family of related tasks wellYes — every current AI system falls here
General AI (AGI)Human-level reasoning and learning across arbitrary, unrelated domainsNo — a hypothetical, aspirational research goal
Superintelligence (ASI)Surpasses human intelligence across essentially every domainNo — purely theoretical, discussed mainly in long-term AI safety debates

Narrow AI: What We Actually Have

  • Image classifiers that recognize specific categories of objects
  • Machine translation between particular language pairs
  • Chess and Go engines that play a single game extremely well
  • Chatbots and language models that generate fluent text
  • Recommendation engines that rank content for one specific platform
Note: Even a very capable large language model is still narrow AI by this definition. Its fluency can span many topics, but its competence is bounded by what it was trained on — it does not autonomously set its own goals or transfer skill to unrelated, real-world tasks the way a general intelligence would need to.

General AI: A Research Goal, Not a Product

AGI describes a system that could learn any intellectual task a human can and transfer knowledge across unrelated domains as flexibly as a person does. No such system exists, and there is no expert consensus on when, or whether, one will. It's worth treating AGI as a long-term research direction rather than something already shipping inside a current app, however it is marketed.

Superintelligence: Beyond Human Level

Superintelligence refers to a hypothetical system that would exceed human ability across nearly all domains at once. It appears mainly in discussions about the long-term future and safety of AI research, not as a near-term engineering target.

Note: When a news headline uses the word 'AI', it almost always means narrow AI. Focusing your judgment on the specific task being solved, rather than on the word 'AI' itself, will keep your read on any story more accurate.

Keep Going

Artificial Intelligence Interview QuestionsPractice real questions asked for Artificial IntelligenceAI Engineer SalarySee what AI Engineer roles pay in the marketAuto Apply for JobsApply to open roles related to Artificial Intelligence and tweak your resume