Classification of AI would refer to trying to distinguish artificial intelligence systems into various groups according to their characteristics, capabilities, and applications which would be helpful for the scientists, developers, policymakers, and the general public understand the nature of different AI systems, their potential applications, and even the potential risks associated with them. Based on this purpose, Artificial Intelligence can be grouped using several frameworks, such as capabilities and functions.
The classification of AI based on capabilities
- Artificial Narrow Intelligence (ANI): Also known as weak AI or Narrow AI systems are created and so used to perform some specific task or function. They do very well within their specific domain but cannot generalize outside of that specified realm.
They are assigned for only single or very specific role. Narrow AI can fail in unpredictable ways if it goes beyond its limits. But it is bit faster and better then a human mind can.
Virtual assistant similar to Siri or Alexa, recommendation system as with Netflix or Amazon and many image recognition systems are very familiar examples of Narrow AI.
2. Artificial General Intelligence(AGI): Also known as General AI, This is a Hypothetical AI system which is capable of understanding, learning, and applying intelligence within a wide range of tasks, similar to the nature of man. Such system does not exist in the world till now.
This is a type of intelligence which could perform any intellectual task with an efficiency of a human.
A truly autonomous robot able to do any intellectual task a human is capable of performing, which remains a goal in AI research is the best example of general AI.
3. Artificial Superintelligence (ASI): If we talk about the ASI , in today world there is No practical applications yet exist, since this is purely a theoretical exercise.
With such system, a machine may surpass human intelligence, and can perform any task better than human with cognitive properties. It is an outcome of general AI. This too is still in the realm of speculation and under ethical debate.
The classification of AI based on Functionality
On the basis of functionality AI is divided into 4 categories
- Reactive Machine AI- The most primitive type of Artificial intelligence is Reactive Machine AI. This is the most basic AI. It reacts to the present and never remembers what happened before; though there are IBM’s chess-playing program Deep Blue, among others. These systems do not know how to learn from past experience. They produce outputs given inputs and do not learn or change their behavior regarding new inputs or previous experience.
It does not possess any learning ability, very simple and straightforward.
2. Limited memory AI – This type of AI learns from earlier experiences or historical data to make decisions for today. They may use data from history to make smarter responses over time.
For example, in the case of a self-driving car, if we consider it, it takes decisions regarding what should be done now in the given present situation on how to drive after it has experienced something in the past.
3. Theory of Mind AI – Such AI would able to understand human emotions, beliefs, and can interact in a socially aware manner. It represents a current class of AI that is purely theoretical: no machine has yet been constructed based upon such a theory.
4. Self-awareness AI – It is a hypothetical type of AI that possesses consciousness and potentially subjective states of experience. It is a theoretical AI form that has consciousness or self-awareness.
These would be more intelligent than human mind. They are far ahead of the AI constructs currently in use. It has not yet been built in real life; it is an abstract concept.