The Four Types of Artificial Intelligence
A.I. has its subsets like machine learning and deep learning, but did you know it also has different types?
In a previous blog, I talked about what exactly artificial intelligence is and how A.I. differs from machine learning which differs from deep learning. I’m going to dive a little deeper into this rabbit hole and talk about the four different types of A.I.
Artificial intelligence can be categorized into four different types: reactive machines, limited memory A.I., A.I. with a theory of mind, and finally, self-aware A.I.
As the name states, reactive machines function by taking in an input and reacting to that input. They are given a specific function and cannot work on its previous experience to improve itself. It’s the most basic form of A.I. that we have. Some examples of this are IBM’s Deep Blue, which was used to beat a chess grandmaster, and Google’s AlphaGo, which beat a top human Go player. In chess, the A.I. is able to recognize chess pieces and placements, which are the input, and move the pieces to specific places, which is the output. That’s their only function and they aren’t able to do anything else with it.
A bit more advanced would be A.I. with limited memory. This refers to how the A.I. is able to store previous information and act upon that information. This may include pre-programmed knowledge, or knowledge obtained throughout the course of use such as through its actions or observations. In order to create an A.I. with limited memory, there has to be a learning model. There are three types of learning models: reinforcement learning, long short term memory (LSTM) , and evolutionary genetic adversarial networks (LGANs). These will be gone over in significantly more detail in a future blog post. An example of A.I. with limited memory would be autonomous cars with their ability to recognize objects on the street, pedestrians, bikers, and so on. They store the information and use it to perform their job. With more input information, they’re able to do their job even better.
Moving on to even more advanced is A.I with a theory of mind. These A.I.’s are capable of perceiving human thoughts and emotions and then react to them. They can interpret the world and the environment around them, as well as understanding how they affect human behavior. Our society has only begun to scratch the surface of creating A.I. with a theory of mind. One example is Kismet, a robot head created in the 90’s, which is capable of mimicking human emotions as well as recognizing them, but is unable to do much else beyond that. Perfecting theory of mind A.I would be a huge leap in technology.
In the distant future, we will have self-aware A.I, or an A.I. that has a consciousness and is capable of having a personality and make self-informed decisions. In my opinion, it’s not a matter of will self-aware A.I. be possible, it’s a matter of when.
Having talked about the four types of A.I., you might be wondering if any of the voice assistants like Google, Siri, or Alexa are considered A.I. They’re just really good speech recognition tools and they don’t have any actual intelligence. In order for learning to occur, which then turns into intelligence, there has to be a machine learning active learning cycle. However, that’s another rabbit hole to dive into!
