"AI" gets used for everything from spam filters to imagined gods. A few distinctions cut through the confusion.
Narrow → General → Super
| Term | Means | Status |
|---|---|---|
| ANI (Narrow) | Superhuman at one domain | Everything today — chess, LLMs, image models |
| AGI (General) | Human-level across most tasks | The goal; debated how close |
| ASI (Super) | Beyond human in every domain | Hypothetical; the singularity target |
Strong vs Weak AI (a different axis)
This is about mind, not capability — and people constantly conflate the two:
- Weak AI — the system acts intelligent (a tool, however capable). All current AI.
- Strong AI — the system genuinely has a mind / understanding. The Chinese Room debate.
A system can be ASI and still "Weak" — superhumanly capable yet nobody home. Capability and consciousness are separate questions, just as the orthogonality thesis separates capability and goals.
When someone says "AGI is near" or "AI is conscious," ask which of these they mean — the argument usually evaporates into a definitions dispute.
Related: The Technological Singularity · The Chinese Room · Emergence and Emergent Capabilities