Quick and Complete Intro to Artificial Intelligence

Atul Sharma
Latest posts by Atul Sharma (see all)

Introduction

As a layman, when we hear about Artificial Intelligence, sometimes we think about robots, sometimes computer with super fast speed. Some relate it to Machine learning while for some it is something else.

In this article, we are going to understand the definition of Artificial Intelligence with a brief history and then we will have bird-eye views of the most emerging fields of Artificial Intelligence with use cases. By the end of this article, you will eliminate all your doubts about this most talked technology word.  

Brief History

Since Artificial Intelligence in buzz word today but it is not new. In 1950, Alan Turing published a paper Computing Machinery and Intelligence, this is also known as Turing Test among Computer Scientists. This was a method to determine if a machine can be intelligent.

Then in a conference in 1956, led by John McCarthy, the term Artificial Intelligence was first coined to discuss the scope, goal, and future of Artificial Intelligence. In 1963, John started AI lab at Stanford.

Based on everything – Artificial Intelligence is an interdisciplinary branch of computer science that mimics the learning methodology of human beings.

Now some confuse Artificial Intelligence with Data Science and Machine Learning. An easy definition – Data Science deals with collecting and analyzing the data while Machine Learning is about learning by observing the patterns from the data and predicting the possible outcome. These all can be visualized with this diagram –

Since the scope of this article to have a high-level understanding of Artificial Intelligence and hence without getting in the further detail of the other constituents of the above diagram, we will try to understand in detail the most emerging fields of Artificial Intelligence with use cases and examples and see how it mimics the human learning behavior.  

See also  Angular tutorial for Beginners – Part 5 : Services

Five most emerging fields of Artificial Intelligence

1. Robotics Process Automation (RPA)

Let us begin with the example of a car company. Initially, all parts were built and assembled manually. Their production was low. Then they automated few processes, developed some software and computer programs to perform them faster and more efficiently. Yes, for supervision humans were needed. Further to this, the next level of software application came and they replaced the supervisors and now everything is being monitored by robots. This branch of Artificial Intelligence is known as RPA or Robotics Process Automation. RPA majorly deals with the elimination of repetitive works by computer software.

Use cases for RPA are customer query and order processing, transferring data among servers, collecting the required data from websites, batch processing, large data processing, bulk data entry and processing are just a few items in the list.

2. Deep Learning (DL)

As we see in the above diagram, Deep Learning is a subset of Machine Learning. It means deep learning also does prediction based on data but it imitates the way how the human brain learns and predicts. It uses the algorithm named Artificial neural network.

This field actually mimics the learning of the human brain, which means it takes huge data, slow learning. Other areas of Artificial Intelligence follow some framework and they get the result, in DL there is no framework, it has many layers of networks based on the available huge dataset. The outcome of DL is more credible than any other branch of Artificial Intelligence.

Candidate of use cases of this area is – security and compliance tracking in the banking domain, healthcare diagnosis, and drug discovery, Alexa, Google Assistant, Siri all use Deep Learning.

3. Computer Vision

If we take an example of kids, initially they recognize nothing, not even their parents, but they see them daily and after some time, they can recognize their parents even in pictures and videos. This is how Computer Vision works. It sees a particular object in many images as data, processes them, and start classifying them, even in new images.

See also  What is NFIP and What is changing in it

When Deep Learning is integrated with Computer Vision, it becomes a different area of specialization and known as Convolutional Neural Network (CNN).

Again, this is not new and the traditional example is a bar-code scanner. Facebook’s suggestion to tag your friend is also part of that. Google lens and auto-driving cars also fit in this category. In healthcare, mammography, x-rays, MRI, Scans and many diagnosis techniques are also good candidates. In agriculture, it is used to identify the quality of grains.

4. Reinforcement Learning

This is part of Machine Learning but we have already seen Machine Learning is itself a subset of Artificial Intelligence. To understand Reinforced Learning, we need to have a bird-eye view of 3 main learnings methodologies in Machine Learning –

  1. Supervised Learning – Here we provide input, based on the existing data it predicts the output. Use cases – House price estimation, Weather forecast, identifying the image, identifying Customer behavior,  
  2. Unsupervised Learning – In this branch, the machine will infer the new pattern which is totally different from its past observation.

In the above GIF, the small kid recognizes the food items and which he doesn’t like but when he gets wine, he takes a chance, though he may not have tried it as it is different so wants to give it a try. He may have seen some drinks in the past in the same color.

A similar example could be when we see a totally unseen animal, we try to categorize it in our known categories of animals, maybe mammals, cats, reptiles, or any other and treat in that way.

This is how unsupervised learning works. It draws inference from the dataset, it already has as labelled.

Use cases for Unsupervised learning can be creating a new cluster of customers for new observation, identifying the new animal based on the other animals’ data for fur, skin color, legs, size and categorize them in the best matching group, getting more customer insight, and many similar scenario-based cases.

See also  American History: A Quick Flashback

3. Reinforced Learning – This follows a different path of learning when the agent is allowed to perform the task, if that is correct, it will get awarded else penalized.

This can be best understood by this example, shown in the gif. The kid knows if he doesn’t eat, he will get punished and that is why he eats and stops his love for mobile, though with some initial resistance. This is how machines learn, if some action gets wrong, next time it will not repeat it.

Use cases for Reinforced learning are – Traffic lights controls, robots, personalized recommendations, advertisements, Self-driving cars, healthcare, gaming, news recommendation, etc.

5. Natural Language Processing (NLP)

This field of Artificial Intelligence is concerned with computer and human language. With the help of NLP, computer analyzes and processes the natural language data.  

Natural Language Processing (NLP) has two parts –

  1. Natural Language Understanding (NLU) –

This deals with the understanding of natural language as input and processing. This is the toughest part as one single word may have multiple meanings in any language.

2. Natural Language Generation (NLG) –  

It is the reverse process of NLU means generates the Natural Language from some internal presentation.

Application/Use case of Natural Language Processing –

  1. Sentiment Analysis – It gets the user’s mood about any product.
  2. Speech Recognition – converts audio to text.
  3. Chatbot – for 360x24x7 customer support
  4. Spam Detection – to cleanup our email inbox, before we classify them as spam.
  5. Video Captioning – this feature is used in the zoom, YouTube, Alexa, and Google Assistance.

I hope this article helps you understand the main area of Artificial Intelligence and decide your area of interest.

References –