Reactive AI is the most basic form of AI, performing the same behaviours in response to certain stimuli with little to no variation in their response. If a reactive AI is faced with a situation it has been faced with before, it will act the same way, meaning it doesn't learn from the past or prepare for the future.

Although reactive AI is the most basic form we currently have, it was actually a huge leap in the field to produce AI that could do this, as it provided the foundation for future developments of AI.

Limited memory AI is distinguished from reactive AI by its ability to learn from the past, use more complex classification systems and make predictions based on past experience.

This type of AI is probably the most commonly used today. However, although it is a great step up from reactive AI, it's limited in its memory (as the name suggests) and much of the information it uses is not stored in the device's long-term memory.

Self-driving cars are perhaps the best example of limited memory AI that we currently have. The car is extremely observant, taking in information from the environment, such as the speed of nearby cars, and uses this information, as well as past experience, to make decisions on the road. How it processes and interprets the data it receives makes it much safer on the roads than a simple reactive AI.

Limited AI is considered narrow AI since it cannot reason and think like humans.

Theory of mind AI is the first step towards truly human-like machines. A theory of mind AI is capable of understanding and remembering human emotions and thus adjusting their behaviour in response to a person's feelings. You could talk to a theory of mind AI like another human.

Currently, theory of mind AI is only theoretical, but we're getting closer to making the theory a reality.  If we do, it will be the first example of general AI in the world.

However, this is difficult because human communication is so fluid. It can be hard for a machine to learn how to respond appropriately to emotions, given they can shift so quickly.

The final and most sophisticated type of AI is self-aware AI. Self-aware AI would have a level of consciousness similar to or identical to humans. This would likely lead to it having emotions, desires and personalities, just as humans do.

To be self-aware, an AI would need to be aware of its own internal state (its thoughts and emotions) and make inferences that would not be possible for lower levels of AI.

Does that mean your device is self-aware if it uses words like "I" or "me"? No - probably, those words were initially written by a human. Even if it did generate this text, it is more likely to be an automatic response by a reactive or a limited-memory AI. But just because the words sound self-aware does not mean that the one generating them understands what they mean.

Currently, we don't have the capability of creating self-aware AI.