There are 4 types of Artificial Intelligence classified by cognitive capability: reactive AI, limited memory AI, theory of mind AI, and self-aware AI. Only the first two exist today — the other two remain academic theory. This classification, developed by University of Arizona researchers, is now used worldwide.
Type 1: Reactive AI
The most basic. It responds to specific inputs but has no memory and doesn’t learn from the past. Every situation is treated as if for the first time.
Classic example: IBM’s Deep Blue, the computer that beat world chess champion Garry Kasparov in 1997. It calculated every possible move, picked the best one, but retained nothing from the match.
Today, basic spam filters and simple recommendation systems still work this way.
Type 2: Limited Memory AI
This is where we are in 2026. These AIs use historical data and recent experiences to make better decisions. This category includes:
- ChatGPT, Gemini, Claude, and other generative AIs
- Autonomous vehicles from Waymo and Tesla
- Facial recognition
- Translation systems
They “remember” during a conversation or trip — but that memory has an expiration date. When you close the app, context is usually lost.
Type 3: Theory of Mind (doesn’t exist yet)
The next leap — and the hardest one. An AI with theory of mind would understand that other beings have their own beliefs, desires, and emotions that influence their behavior.
Simple example: sensing that a person is sad from tone of voice and adjusting the response with genuine empathy — not just mimicking the pattern. Research is underway, but nobody’s there yet.
Type 4: Self-Aware AI (pure theory)
The final level: an AI that is conscious of itself, knows it exists, feels, and develops its own goals. This is the science fiction scenario — and the most debated topic among AI safety researchers at institutions like the Future of Life Institute.
If and when this happens is an open question. Some scientists think it’s impossible; others give it decades.
Comparison table
| Type | Exists today? | Example |
|---|---|---|
| Reactive | Yes | Deep Blue, basic filters |
| Limited Memory | Yes | ChatGPT, autonomous cars |
| Theory of Mind | No | None |
| Self-Aware | No | None |
Why this classification matters
Understanding these types helps you filter hype from reality. When someone says “the AI is thinking” or “feeling,” it’s almost always marketing. What we have today are very sophisticated limited-memory systems — nothing close to consciousness.
To see how these types show up in daily life, check out 10 AI examples you already use or understand better what is AI.