1. Types of AI you should know
Artificial Intelligence is one of humanity’s most sophisticated and amazing creations
to date. That ignores the fact that the field is still largely unexplored, implying that
any amazing innovation we encounter today is only the tip of the AI iceberg. Despite
the fact that this point has been reiterated numerous times, a complete
comprehension of AI’s future impact remains elusive.
People are concerned about the inevitability and vicinity of an AI takeover due to
AI’s rapid growth and formidable abilities. Moreover, the effects of AI in numerous
industries have caused business executives and the public at large to assume that we
are nearing the peak of AI research and fulfilling AI’s full potential. Understanding
the perspective and current types of AI will provide a fuller picture of current AI
technologies and the road to their development.
What is AI?
The process of creating intelligent machines from massive amounts of data is known
as artificial intelligence. Systems learn from their past experiences and accomplish
tasks that are similar to those performed by humans. It improves the efficiency,
precision, and efficacy of human efforts. To create computers that can make
judgments on their own, AI employs complicated algorithms and procedures.
Types Of AI
The extent to which an AI system can imitate human ability is used as a measure to
determine the types of AI. As a result, AI can be divided into numerous categories
based on how well a machine corresponds to humans in terms of diversity and
efficiency. In this approach, an AI that can perform more human-like tasks with
2. equivalent standards of accountability will be considered a more sophisticated sort
of AI, whereas an AI with restricted functionality and performance will be regarded
as a simpler and less developed type.
There are two classifications for AI based on this criterion. One approach is to
categorize AI and AI-enabled technologies according to their resemblance to the
human mind and their ability to “think” and possibly “feel” like humans.
Type 1- Based On Capabilities
Based on capabilities, there are three categories of AI:
■ Narrow AI
■ General AI
■ Super AI
1. Narrow AI or Weak AI or Artificial Narrow
Intelligence (ANI)
“Alexa! Set an alarm for 7 A.M.”
Also known as Weak AI, is a level of AI that involves robots that can only do a limited
set of activities. At this phase, the machine has no ability to reason and just conducts
a series of pre-defined operations.
Cortona, Siri, Alexa, self-driving cars, Alpha-Go, Sophia the humanoid, and others
are examples of weak AI.
2. General AI or Strong AI or Artificial General
Intelligence (ANI)
3. AGI, also known as Strong AI, is the step in the development of Artificial Intelligence
when robots will be able to reason and make decisions in the same way that humans
do. It is yet to be demonstrated but is expected to develop intelligence the same as
humans. Many scientists, including Stephen Hawking, believe that strong AI poses a
threat to humanity’s survival.
“The full development of AI could spell the end of mankind. It’d set off on its own,
re-designing itself at a breakneck speed. Humans, whose biological evolution is
slowed, would be unable to compete and would be surpassed.”
3. Super AI or Artificial Super Intelligence (ASI)
Super AI is the stage of Artificial Intelligence at which computers’ capabilities
surpass those of humans. Machines have taken control of the Earth, according to a
hypothetical scenario presented in sci-fi novels and movies.
Given our present rate of development, I believe machines are not far from reaching
this stage.
“You have no notion how fast—it is expanding at a rate that is near to
exponential—unless you have direct exposure to groups like Deepmind. In the next
five years, there is a significant risk of something extremely dangerous happening. At
most ten years.! —According to Elon Musk.
These are the various levels of intelligence that a machine can achieve. Let’s look at
the many types of AI and how they work.
Type 2- Based On Functionalities
1. Reactive Machine
4. They are the most basic and ancient sort of Artificial Intelligence. They imitate a
human’s ability to respond to a variety of stimuli. Because this type of AI has no
memory, it is unable to use previously acquired information/experience to improve
results. As a result, these AI systems lack the ability to learn themselves like the ones
we see today.
Deep Blue, the computer that defeated international grandmaster Garry Kasparov, is
an excellent example of this type of equipment.
The supercomputer was able to detect all of the legal options available to it and its
opponents. It chose the best feasible move based on the options. However, because
these machines have no memory of their own, they are unable to learn from their
previous actions.
2. Limited Theory
This sort of AI, like Reactive Machines, has memory capabilities, allowing it to
leverage prior data and experience to make better decisions in the future. This
category encompasses the majority of the commonly used applications in our daily
lives. These AI applications can be taught using a huge amount of training data
stored in a reference model in their memory.
Many self-driving cars use them to store data such as GPS location, speed of
neighboring automobiles, size/nature of barriers, and a hundred other types of
information in order to drive like a person.
There are three types of machine learning models that can achieve this form of
Limited Memory:
a. Reinforcement learning
5. Through several rounds of trial and error, these models evolve to make better
predictions. Computers are taught to play games like Chess, Go, and DOTA2 using
this technique.
b. Long Short Term Memory (LSTMs)
Researchers reasoned that using past data to predict the next item in a sequence,
particularly in language, would be beneficial, therefore they devised a model based
on the Long Short Term Memory. The LSTM labels more current information as
more significant and those from the past as less essential when predicting the
following parts in a sequence.
c. Evolutionary Generative Adversarial Networks (E-GAN)
Because the E-GAN has memory, it evolves with each evolution. The model
generates a developing entity. Because statistics is a math of chance, not a math of
exactitude, growing entities do not always pursue the same route. The model may
identify a better path, a path of least resistance, as a result of the changes. The
model’s following generation mutates and evolves in the direction of its ancestor’s
incorrect route.
The E-GAN produces a simulation that is analogous to how people have developed
on this planet in several ways. Each child is more poised to have an extraordinary life
than its parent in the event of flawless, successful replication.
3. Limited Memory Types In Practice
While every machine learning model is built with a finite amount of memory, this
isn’t necessarily the case when it’s deployed.
A.I. with limited memory works in two ways:
6. A team is constantly updating a model with new data.
Models are automatically trained and refreshed in the A.I. environment based on
model usage and behavior.
Machine learning must be built-in into the structure of a machine learning
infrastructure in order for it to support a limited memory type.
Active Learning is becoming more widespread in the ML lifecycle. There are six steps
in the ML Active Learning Cycle:
■ Training Data. A machine learning model requires data to train on.
■ Build ML Model. The model has been developed.
■ Model Predictions. The model makes predictions,
■ Feedback. Human or environmental inputs provide feedback on the
model’s predictions.
■ Feedback is converted into data. The data repository receives the feedback
and stores it.
■ Repeat Step 1. Continue to iterate on this cycle.
4. Theory of Mind
It is the next level of artificial intelligence, with little to no impact on our daily
existence. These types of AI are often in the “Work in Progress” stage and are only
available in research labs. Once achieved, this type of AI will have a comprehensive
understanding of human minds, including their needs, likes, emotions, mental
processes, and so on. The AI will be able to change its own response based on its
grasp of human minds and their whims.
The theory of mind AI was implemented at Hanson Robotics’ Sophia. Sophia is able
to see thanks to cameras in her eyes and computer algorithms. She can keep eye
contact with individuals, recognize them, and follow their faces.
7. 5. Self-Aware AI
This is the AI’s final step. Its current presence is simply a rumor, and it can only be
found in science fiction films. These AI systems are capable of comprehending and
eliciting human feelings, as well as possessing their own emotional states. This form
of AI will take decades, if not generations, to develop. Elon Musk and other AI
doubters are wary of this type of AI. This is because once an AI becomes self-aware,
it may enter Self-Preservation mode, viewing mankind as a possible threat and
pursuing efforts to eliminate humanity directly or indirectly.
Branches of AI
By employing the following processes/techniques, Artificial Intelligence can be
utilized to tackle real-world problems:
■ Machine Learning
■ Deep Learning
■ Natural Language Processing
■ Robotics
■ Expert Systems
■ Fuzzy Logic
1. Machine Learning
The science of teaching machines to understand, process, and analyze data in order
to solve real-world issues is known as machine learning.
Machine Learning is divided into three categories:
■ Supervised Learning
■ Unsupervised Learning
■ Reinforcement Learning
2. Deep Learning
8. It is the process of using Neural Networks to obtain insights and build solutions from
high-dimensional data. Deep Learning is a subset of Machine Learning that can be
used for more complex issues.
3. Natural Language Processing
Genuine Language Processing (NLP) is the study of extracting information from a
natural human speech in order to communicate with robots and expand enterprises.
Amazon employs natural language processing (NLP) to better comprehend customer
feedback and improve the user experience.
4. Robotics
It is a branch of AI that focuses on various robot applications and disciplines. AI
Robots are artificial agents that act in a real-world environment to create results by
taking responsible behaviors.
Sophia the humanoid is an outstanding demonstration of artificial intelligence in
robotics.
5. Expert Systems
An expert system is a computer system based on artificial intelligence that learns and
mimics the decision-making abilities of a human expert.
If-then logical notations are used by expert systems to tackle complicated issues. It
does not rely on procedural programming in the traditional sense. Expert systems
are mostly utilized in data administration, medical facilities, loan analysis, and virus
identification, among other applications.
6. Fuzzy Logic
9. Instead of the conventional modern computer logic, which is boolean in nature,
fuzzy logic is a computing approach based on the ideas of “degrees of truth.”
It’s utilized to address difficult challenges that require decision-making in the
medical industry. They’re also employed in automatic transmissions, vehicle climate
control, and other applications.
Conclusion
We may be a long way from constructing self-aware machines that can fix all
problems. However, we should concentrate our efforts on figuring out how a
computer can train and learn on its own and make decisions based on previous
experiences.
I hope this post has clarified the multiple kinds of AI.