Types of Artificial Intelligence

 Artificial Intelligence can be divided into various types. There are mainly two categories based on capabilities and functionalities. 

Types of Artificial Intelligence based on capabilities:

There are three types of artificial intelligence based on capabilities. 

• Weak AI or Narrow AI
• Strong AI
• General AI

Now we will discuss them one by one:

Narrow AI:

• Narrow AI, also known as weak AI, focuses on a particular task and can not perform beyond its limitations.  It targets a subset of cognitive abilities. These machines can perform a single task, they are programmed for and cannot add anything else in output using their intelligence.
• All existing work that is performed is mostly included in this type, which means all belong to narrow AI including the most capable and complicated work done in this field.
• Example:

1. Apple Siri is an example of narrow AI that performs a limited pre-defined set of functions. It may cause a problem if it faces issues beyond its range.
2. Google translator
3. Image recognition system
4. Speech recognition
5. E-commerce
6. Self-driving cars
7. Games using AI

General AI:

• General AI which is also known as Strong AI, is the type of artificial intelligence that can perform any task with human-like efficiency. In this type AI agents can learn, perceive and understand any problem similarly to humans.
• This allows the machine to apply knowledge and skills in different contexts. AI researchers have never been able to achieve General AI. You need to find a way to be machine aware and program a complete set of cognitive abilities. General AI has received a $ 1 billion investment from Microsoft through OpenAI.
• Fujitsu manufactured Kyo, one of the fastest supercomputers in the world. This is one of the key attempts to achieve powerful AI. It took nearly 40 minutes to simulate one second of neural activity. Therefore, it is difficult to determine whether General AI will be realized in the near future.
• Tianhe-2 is a supercomputer developed by the  National University of Defense Technology in China. It holds a record of cps (calculations per second) at 33.86 petaflops (hundreds of billions of cps). It sounds exciting, but the human brain is estimated to be one exaflop or one billion cps.

Super AI:

• Super AI goes beyond human intelligence and can perform better tasks than humans. The concept of artificial intelligence does more than just understand AI as it has evolved to be very similar to human emotions and experiences. It also evokes one's own feelings, needs, beliefs, and desires. Their existence is still a hypothesis. Key features of Super AI include thinking, puzzle solving, judgment, and independent decision-making.

Types of Artificial Intelligence based on functionalities:

There are four types of artificial intelligence based on functionalities.

• Reactive machine
• Limited memory
• Theory of mind
• Self-awareness

Reactive machine:

• Reactive machines are the most basic type of artificial intelligence. They are the oldest forms of AI systems with very limited functionality. They mimic the ability of the human mind to respond to different types of stimuli.
• They do not store memory or use past experience to determine future behavior. Works only with current data. They perceive the world and react to it. Reactive machines are equipped with specific tasks and have no function other than those tasks.
• An example of a reactive machine is Google's AlphaGO.
• Another example of a reactive AI machine is IBM's Deep Blue, which defeated chess grandmaster Garry Kasparov in 1997.

Limited memory:

• Machines with limited memory are machines that can learn from historical data and make decisions, in addition to the capabilities of purely reactive machines. But they can store data for a very limited time.
• For example, image recognition AI is trained with thousands of images and their labels to teach them to name the objects they can. When an image is scanned by such an AI, it uses the training image as a reference to understand the content of the presented image and label the new image with high accuracy based on its "learning experience".
• Limited Memory AI monitors how other vehicles are moving now and over time. This continuously collected data is added to the static data of AI engines such as lane markings and traffic lights. These are involved when the vehicle decides when to change lanes, avoid disconnecting other drivers, or collide with nearby vehicles.

Theory of mind:

• Theory of Mind represents a class of advanced technology and exists only as a concept. These types of AI need to fully understand that people and things in the environment can change emotions and behaviors. You need to understand people's feelings, feelings, and thoughts. There are many improvements in this area, but this type of AI is not yet complete.
• AI mind-level theory helps us better understand the entities with which AI interacts by recognizing AI needs, emotions, beliefs, and thinking processes.
• Theory of Mind A real example of  AI is Kismet. Kismet is a robot head created by researchers at the  Massachusetts Institute of Technology in the late 1990s. Kismet can imitate and recognize human emotions. Both abilities are important advances in the theory of mind AI, but Kismet cannot track the line of sight or pay attention to humans.

• Hanson Robotics Sophia is another example of the implementation of the theory of mind AI. By combining Sophia's eye camera and computer algorithms, Sophia can see. She can maintain eye contact, recognize people and follow her face.

Self-awareness:

• This is the final stage of AI development and currently only exists virtually. Self-aware AI is a self-evident AI that has evolved to be sufficiently similar to the human brain and has developed self-awareness. Creating this type of AI has been decades, if not centuries, since its realization and remains the ultimate goal of all AI research.
• This type of AI  not only understands and arouses the emotions of interacting AI, but also has emotions, needs, beliefs, and in some cases its own desires. And that's the kind of AI that tech denials are wary of. The development of self-awareness can spike our progress as a civilization, but it can also potentially lead to catastrophe. This is because when AI is self-aware, it can have ideas such as self-preservation that can directly or indirectly spell the end of humanity. Because such an entity can easily defeat human intelligence and make elaborate plans for humanity.

Conclusion:

We may be far from creating a self-aware machine that can solve all problems. However, we must focus on understanding how machines can be trained and learned on their own and have the ability to make decisions based on past experience.