Diverse types of Artificial Intelligence: A Must-know for AI Enthusiasts

A precursory article that explains various categorizations of artificial intelligence, some real-life examples and concepts.

Table of contents

• Introduction
• Types of Artificial Intelligence
• Capabilities-based AI categorization
• Functionalities-based AI categorization
• Conclusion

Introduction

Artificial Intelligence has been the hot tech buzz worldwide among computer scientists in the last few years. It is nothing but a branch of computer science involved with the design of computers or other programmed mechanical devices having the capacity to imitate human intelligence and thought.

Despite some of the already successful innovations and advanced research, the technocrats claim that the depth of AI that we have seen until now is just like a drop in an ocean, the time is yet to deliver the advanced sophisticated phase of this technology.

In this article, I am trying to explain the basic categorization followed in the innovation happening in the AI tech industry which will be a piece of very useful yet foundational knowledge for those students and professionals who are eager about this concept.

Types of Artificial Intelligence

All of the AI based solutions that have been already developed and those which are in pipeline for the future ideation and implementation are categorized broadly according to two congregations-

• Capabilities-based AI categorization 
• Functionalities-based AI categorization

Under the Capabilities bases AI categorization, we have the following types of solutions-

• Narrow Artificial Intelligence
• General Artificial Intelligence
• Super Artificial Intelligence
  

Under Functionalities-based AI categorization, we have the following types of solutions-

• Reactive Machines
• Limited Memory
• Theory of Mind
• Self Awareness

Capabilities bases AI categorization

Cataloguing AI solutions based on their "power or ability to do something" comes under this type of categorization. Let us go through each one of them in detail.

Narrow Artificial Intelligence

Artificial Narrow Intelligence (ANI), also called Weak AI or Narrow AI is the only type of artificial intelligence we have successfully perceived until today. Narrow AI is AI that is programmed to perform a single task and it cannot perform the tasks beyond its limitation.

These systems can solve problems in real time but their entire knowledge about this universe is restricted to the specific datasets that the researchers used while training the models. This makes them unable to solve any problem outside of the patterns recognized from the training datasets. It has no consciousness, sentiment or human emotions.

As I mentioned earlier, all of the advanced AI applications designed and delivered in recent times belong to Narrow AI such as-

Google Translate - A multilingual neural machine translation service developed by Google to translate text, documents and websites from one language into another

English Wikipedia's homepage translated into Portuguese  (Source: Wikipedia)

IBM Watson - A question-answering computer system capable of answering questions posed in natural language

Watson demo at an IBM booth at a trade show (Source: Wikipedia)

Apple’s Siri - A private intelligent virtual assistant that helps to make calls, send texts, use apps, and get things done with just your voice. Etc.

Siri Remote for the Apple TV (Source: Wikipedia)

You can obviously doubt why we call these solutions “Weak AI” although the magnitude of the problem they solve is huge and they are derived out of complex algorithmic computations. The reason is that they are still very distant from actual human intelligence. They are deprived of the capacity to think of themselves to solve an actual problem instead, most of the problems and corresponding solutions are coded to their artificial intelligence in a patterned way using various algorithms.

For example, We consider Siri as the most advanced private virtual intelligent assistant to date. It can help you in assisting the routine work and let you remember the daily schedules, official or education-related doubts, health-related queries etc. but it cannot answer the questions if you need some advice or guidance related to personal life that your close friend or parents can provide by analyzing rationally, logically and emotionally by understanding your circumstances.

Although the lack of understanding of real-life surroundings is considered a limitation of Narrow AI, we must accept its positive side such as the speed that it can provide while solving a problem that a human cannot even think of. They can automate various mundane tasks, give insights about various data and sometimes acts as a helper in various decision-making activities.

General Artificial Intelligence

General AI, also referred to as “Strong AI” or “Deep AI” is a concept of a machine with general intelligence that mimics human intelligence, with the ability to think, understand, learn and apply its intelligence to solve any problem as humans do in any given situation.

These types of AI solutions come along with systems that are conscious, sentient, and driven by emotion and self-awareness.

Still, human intelligence is considered superior to Strong AI because humans can strategize plans and creatively think about new ideas, and our thoughts and decision-making ability exponentially grow with the experience while interacting with this universe on a daily basis. These aspects of human intelligence are not yet replicated by Strong AI machines.

Although General Artificial Intelligence is lacking consciousness and emotions, it can solve problems and make appropriate judgements under uncertain conditions. Also, it uses its prior knowledge for somewhat imagining various scenarios and making suitable decisions.

K Computer (Source: Wikipedia)

Fujitsu, a Japanese Information and communications technology company built one of the fastest supercomputers called the “K computer” which is considered a notable attempt at achieving Strong AI. It took 40 minutes to simulate a single second of neural activity. Hence, this scenario made computer scientists clear that Strong AI is difficult to achieve in the foreseeable future.

Riken Advanced Institute for Computational Science (AICS) in Kobe, which housed the K computer (Source: Wikipedia)

Super Artificial Intelligence

Super Artificial Intelligence is defined as a form of AI capable of surpassing human intelligence by manifesting cognitive skills and developing thinking skills of its own. Its existence is still hypothetical.

It can surpass human intelligence in all aspects ranging from creativity to general wisdom, to problem-solving. Researchers often define a scenario called “Technological singularity” which is nothing but a hypothetical point in time at which technological growth becomes uncontrollable and irreversible, resulting in unforeseeable changes to human civilization. Machines having Artificial Super Intelligence might be able to do certain things that are not humanly possible. If these machines start to emerge then that becomes a singularity point.

For example,

What if a Super intelligent machine starts to replicate itself by creating similar machines and starts to rule the world by considering humans as their slaves?

Photo by Brian McGowan on Unsplash

It will have the capacity and knowledge to create weapons, form the government and control us by framing certain rules.

Actually, when we reach the singularity point, we won't be knowing how those machines will react to us. Some researchers believe that there are 2 chances in the midst of uncertainty-

• An artificial super-intelligent machine with good character
• An artificial super-intelligent machine with a bad character


Image by Lukas from Pixabay 

If the machine is having good character then it will almost become like a God. It will be immortal, It can save humankind from various natural disasters, treat and heal physical diseases, solve political and economical crises, etc.

But if the scenario is the opposite then we can't even imagine the degree of disaster that we are going to face.

Functionalities-based AI categorization

Cataloguing the AI solutions based on their “quality of being suited to serve a purpose well” comes under this type of categorization. Let us go through each one of them in detail-

Reactive machines

Reactive machines are those AI solutions that cannot form memories or use past experiences to influence present-made decisions. They can only react to currently existing situations, hence the name “reactive”.

An AI program created by IBM in the 1980s named “Deep blue” is considered an example of a Reactive machine. The objective of this AI solution was to compete against humans for playing and winning the chess game.

One of the two cabinets of Deep Blue in its exhibit at the Computer History Museum, California (Source: Wikipedia)

Although it doesn't learn and improve its performance from past experience since it has no memory, it can optimally make a decision for moving a piece on the chess board according to the rule of games to win against the opponent.

Garry Kimovich Kasparov, a Russian chess grandmaster and former World Chess Champion was defeated by this AI program to become the first computerized program to defeat a human opponent. This experiment showed the potential of the “reactive” ability of AI to make quicker, more accurate and optimal decisions based on the environment.

Garry Kasparov playing a simultaneous exhibition in 1985 (Source: Wikipedia)

In challenging aspects, Reactive machines don't possess the capabilities to perform tasks beyond their simple programmed knowledge. The lack of memory and no understanding of the past, present and future makes them unfit for solving further complex problems.

Limited Memory

Limited memory overcomes the challenges faced by Reactive machines in many aspects of problem-solving. It can derive knowledge from previously-learned information, stored data, or events. Although they can store the data, their memory is usually short-lived. They use both programmed logic and patterns recognized and learnt from a lot of trained datasets to solve complex problems.

Self-driving cars are considered an example of Limited memory. These cars use their memory while driving for understanding the environment and memorizing the scenario such as-

• Is there any obstacle coming in front of the car?
• Is the car keeping enough distance from the pedestrians ?
• Are there any traffic lights active now ?
• How other vehicles near this car is placed on the road? etc.


A prototype of Waymo's self-driving car, navigating public streets in Mountain View, California in 2017 (Source: Wikipedia)

These pieces of information are memorised by the AI and it combines this information with already present pre-programmed logic to make appropriate decisions.

Theory of Mind

Theory of Mind is a kind of AI that aims to understand human needs by detecting emotions, and sentiments, discerning beliefs and mapping thought processes. This technology is in the early stage of research and we haven't achieved its completion yet.

Two notable examples are the robots Kismet and Sophia, created in 2000 and 2016 respectively.

Kismet, developed by Professor Cynthia Breazeal, was capable of recognizing human emotions based on facial signals and could replicate said emotions with its face, which was structured with human facial features.

Kismet now resides at the MIT Museum in Cambridge, Massachusetts, United States. (Source: Wikipedia)

Sophia is a humanoid bot created by Hanson Robotics. She has the ability to see and respond to interactions with appropriate facial expressions.

Sophia in 2018 (Source: Wikipedia)

These two humanoid robots are considered successful initial steps for achieving “Theory of Mind AI” in the near future.

Self Awareness

Self Awareness AI can understand their states, internal traits, and conditions and perceive human emotions. Self Awareness is not currently in existence but would be considered the most advanced form of artificial intelligence that can be imagined by a human brain. They will be able to think for themselves, have desires, and understand their feelings.

Image by Gerd Altmann from Pixabay

The major difference between the “Theory of Mind” and “Self Awareness” is that the “Theory of Mind” only focuses on the aspects of comprehension and replication of human practices. However, “Self-awareness” takes it a step further by implying that it can and will have self-guided thoughts and reactions.

Conclusion

In this article, we discussed one of the most important but foundational categorizations of artificial intelligence. Also, we went through some of the real-life experiments that the industry has already extrapolated in the past couple of years. I hope by going through the reference sections of this article, AI enthusiasts will be able to surf through more in-depth contextual and elaborative information. 

References

• Poole, David; Mackworth, Alan (2017). Artificial Intelligence: Foundations of Computational Agents (2nd ed.). Cambridge University Press. ISBN 978-1-107-19539-4.
• Russell, Stuart J.; Norvig, Peter (2009). Artificial Intelligence: A Modern Approach (3rd ed.). Upper Saddle River, New Jersey: Prentice Hall. ISBN 978-0-13-604259-4.