Theory
Three levels of capability
People lump all AI together, but it helps to classify AI by capability into three types: narrow, general, and super. This classification is also where hype and reality most need separating, because two of the three do not exist yet.
This lesson defines all three honestly. The single most important fact to carry away: everything called AI today is narrow AI. General and super AI are goals or speculation, not current technology. Knowing this keeps you grounded when you hear grand claims about 'AI'.
At a glance
| Type | What it would be | Exists today? |
|---|---|---|
| Narrow AI (ANI) | Skilled at ONE specific task or narrow range | Yes: all current AI is narrow |
| General AI (AGI) | Human-level intelligence across ANY task | No: a goal, not yet achieved |
| Super AI (ASI) | Surpasses humans in essentially every field | No: purely hypothetical |
Theory
Narrow, general, super
Narrow AI (ANI), also called weak AI, is designed for one specific task (or a narrow range). It can be superb within that task, a system that spots crop disease in photos, a voice assistant, a recommendation engine, but it cannot do anything outside it. All AI that exists today is narrow AI.
General AI (AGI), or strong AI, would have human-level intelligence, able to understand, learn, and apply knowledge across any task the way a person can. It does not exist; it is a long-term goal and research aspiration.
Super AI (ASI) would surpass human intelligence in essentially every field. It is purely hypothetical, a subject of speculation, not something that exists. So the honest picture: narrow is real, general is not yet, super is imagined.
Quiz
Which type of AI actually exists today?
- Super AI, which surpasses humans in every field
- Narrow AI: all current AI is task-specific (like crop-disease detection or voice assistants), skilled at one thing but not generally intelligent
- General AI, with full human-level intelligence across all tasks
- All three types exist equally
Show the answer
Narrow AI: all current AI is task-specific (like crop-disease detection or voice assistants), skilled at one thing but not generally intelligent
Only narrow AI (ANI) exists today: every current AI system is designed for a specific task, image recognition, voice assistants, recommendations, crop-disease detection, and cannot operate outside it. Option A, super AI, is purely hypothetical, it does not exist. Option C, general AI (human-level across any task), does not exist either; it remains a research goal. Option D is false: general and super AI are not real, so the three do not exist equally. The honest, non-hype fact to remember: today's AI is all narrow, however impressive a specific system is, it is still specialised, not a general or superhuman mind.
Think first
Why is it important to know that general and super AI do not exist yet?
Why stress that only narrow AI is real, when general and super AI are talked about so much? Then tap.
Show the answer
Because confusing today's real, narrow AI with the hypothetical general or super AI leads to badly mistaken expectations, decisions, and fears, so distinguishing them keeps your thinking accurate. Much public conversation and marketing blurs these categories, describing narrow systems as if they were general minds, which causes several problems. First, EXPECTATIONS: if you think an AI has general, human-like understanding, you will expect it to reason, use common sense, and handle novel situations it was never built for, and you will be misled when it fails, because a narrow AI genuinely cannot do those things; it only does its one task. Second, TRUST and SAFETY: overestimating an AI leads people to hand it judgements it cannot actually make (assuming it 'understands' the way a person does), which can cause real harm when the system confidently gets things wrong outside its narrow competence. Third, FEAR: much anxiety about AI imagines super-intelligent machines taking over, a scenario about ASI, which does NOT exist and is speculative, whereas the real, present concerns about AI are different and more mundane (bias in narrow systems, privacy, job impacts, errors), and conflating the two distracts from the issues that actually matter now. Fourth, EVALUATION: knowing the categories helps you cut through hype, when a product is called 'AI', you can ask what it really is (almost always a narrow system for a specific task), rather than imagining a general mind. So the distinction is not academic pedantry; it is essential for realistic expectations, safe use, sensible discussion of risks, and clear-eyed evaluation of claims. Today's AI is narrow and powerful within limits; general and super AI are, respectively, an unachieved goal and speculation, and keeping that straight is the foundation of thinking about AI honestly. Know what is real versus imagined, and you reason about AI soundly.
Summary
Key takeaways
- AI is classified by capability into narrow, general, and super.
- Narrow AI (ANI, weak AI) is skilled at one specific task and cannot operate outside it; all AI today is narrow.
- General AI (AGI, strong AI) would have human-level intelligence across any task; it does not exist yet (a research goal).
- Super AI (ASI) would surpass humans in essentially every field; it is purely hypothetical.
- The single key fact: everything called AI today is narrow AI, however impressive.
- Confusing narrow AI with general or super AI causes mistaken expectations, misplaced trust, and misdirected fears.
- Memory hook: narrow is real (one task), general is not yet (human-level), super is imagined (beyond human).