Introduction:
AI has got to be one of the most compelling and astounding discoveries made by man. And what may incite your interest even more is the fact that we have only touched the tip of the iceberg. A vast majority of the field is yet to be explored. It is still hard to perceive how much impact AI may actually have on mankind in the future. One way to map out or at least give us a clearer picture of the existing AI technology is to define the major categories/types of AI that are in use today.
AI Classification and its Criteria
The basic criterion used to classify machines using AI is how well they can replicate human capabilities in terms of both versatility and performance. So those which show more human-like functions along with proficiency in their work will be considered to be on a higher level, a bit more evolved level you could say. Whereas those which show limited functionality and performance will be considered a less evolved type.
According to this criterion, there are two ways of classification of AI. One depends on how alike AI-enabled machines are to the human mind.
First System of Classification:
Using this system we can classify AI into four types:
- Reactive Machines
These are the oldest and most basic form of Artificial Intelligence. They have very limited capabilities. They do not store any ‘memories’ therefore they do not use experiences gained from past situations in order to show response to present situations, in other words, they do not have the ability to learn. They follow the basic ability of the human mind to respond to different stimuli according to simple perception. They can be used for automatic responses to a limited set of inputs. IBM’s Deep Blue (defeated chess grandmaster Garry Kasparov 1997) is the perfect example of a reactive machine
- Limited Memory
This type of AI works on the principle of limited memory which is dependent on both pre-programmed knowledge and observations carried out over time. They are capable of learning from data gained through past situations. Most of the existing AI falls under this category. Present-day AI systems store training data in their memory to form a reference model which will help it to solve problems. They can however only retain data for a short or a ‘limited’ period of time. Self-driving cars or ‘Autonomous vehicles’ are the perfect examples of limited memory AI, they make observations about vehicles moving around them, their speed, distance and other information which will help them to navigate roads. Other examples include chatbots and virtual assistants (VAs).
- Theory of Mind
This AI is still only a ‘concept’, or a work in progress you could say. This type of AI has the ability to interpret the world around it and the people in it. It requires an understanding that the people and things in an environment can alter feelings and behaviors. These machines will have the ability to understand needs, emotions, beliefs and thoughts. They must understand human needs and perceive humans as individuals whose thoughts can be shared, therefore ‘understanding’ humans
- Self-aware
This is the most advanced type of artificial intelligence which involves machines having consciousness and self-awareness. Scientists haven’t been able to build something like this i.e. it only exists ‘hypothetically’ but they believe them to be the ultimate goals of AI development. This type of AI will not only be able to understand emotions but will have emotions, needs, beliefs and potential desires. (Yes, that does tend to raise a few red flags, especially if you’ve been watching The Terminator all day) Self-aware robots may be problematic. Where they’ll have their uses they will have their consequences. We can only hope we build a fail-safe way to keep AI-enabled machines under check for the sake of humanity.
Second System of Classification:
The second system of classification which is more widely used divides AI into three types:
- Artificial Narrow Intelligence (ANI)
This type of AI represents all existing AI to date. ANI-enabled machines can only perform specific tasks autonomously using human-like capabilities. They can only do what they are programmed to do and thus are limited. This type of AI corresponds to the prior mentioned Reactive AI and Limited Memory AI.
- Artificial General Intelligence (AGI)
This type of AI has the ability to learn, perceive, understand and function as a human being. They could successfully perform any intellectual task that a human being can. A system comprising of this type of Ai would be able to build multiple competencies and form connections across domains independently without the need for training making AI systems just as capable as humans.
- Artificial Super intelligence (ASI)
This type of AI would form the peak of AI research. It refers to a time when the capability of computers will surpass humans, making them the most capable form of intelligence on Earth. They will have greater memory, faster analysis and data processing, and decision-making capabilities. Where they may sound glorious, they may threaten humanity’s very existence. It could solve all world problems or be the last invention ever made.
For now, one thing is clear. This type of technology is far out of our reach. It may take years of hard work, maybe even centuries before we come across this kind of tech. Like we said before, we have only touched the tip of the iceberg, we have yet to dive into the ocean. We must be optimistic for what is to come as the uncertain future lies ahead.