An Introduction to Artificial Intelligence
AI for Beginners
AI is a word widely used today. We hear it from Scientists, CEOs, Comedians, Salesmen, YouTubers, High schoolers, Young kids, babies and animals — that is if you listen carefully.
There’s a lot of hype generated around AI, Machine Learning and Data Science. Unfortunately, most of us do not understand what is it, what it does or how it works.
My goal with this story is to give you information about the growing world of Artificial Intelligence.
So, What is Artificial Intelligence?
Is it an emotion-less cyber monster that is going to destroy Earth? or is it a computer program that takes an input and does something to get an output?
If you take a moment to think about the word Artificial Intelligence, you will soon realise that it means Intelligence that is made artificially.
Think of an artificial limb — it looks like a limb, it can do things that a limb can and it was created for a necessity.
Artificial intelligence is an endeavour to replicate human intelligence in machines. It is going to mimic human intelligence, and it was created because there was a necessity.
The need for AI is subjective. You might want to eradicate blindness entirely by having an AI system scan the eyes of people and warn them early on,
or You might want an AI to automate those boring, repetitive tasks at work so you can write a blog instead, or You might want to understand your customers better or perhaps, you just need a friend — OK Google!
Now, you know the meaning of Artificial Intelligence but you still do not know anything about it. You don’t know how to make one, right? and You probably have no idea where it’s being used, right?
AI is strictly used for solving problems. It is everywhere because there are problems everywhere.
So, What types of problems does AI solve?
The problems which AI can solve depends on the type of Intelligence it has attained. AI systems generally fall under three levels of Intelligence —
- Narrow Intelligence: These AI systems are focused on doing specific tasks and doing them really well. Most of the applications used these days fall under this category. Examples of narrow AI are Google search, IBM Watson, Siri, Alexa, etc.
- General Intelligence: This AI system can do what a human can. These AIs exist in dystopian sci-fi stories. They don’t exist in real life yet and probably won't for a long time until there are major technological (and hardware) breakthroughs.
- Super Intelligence: This AI agent will surpass human intelligence and capabilities. They do not exist yet. I think it is possible to create one in the future.
Narrow Artificial intelligence is a system that can learn tasks and solve problems without being explicitly instructed (AKA Programmed) on every single detail. It should be able to do reasoning and abstraction on its own, and easily transfer knowledge from one domain to another. To do this, scientists created technologies that aid the system in reasoning, deducing, inferring and predicting.
Technologies, eh?
Imagine you are a composer for an orchestra. To create a beautiful symphony(i.e AI), you need to understand your tools — guitars, violins, harps, flutes, trumpets, etc.
The deeper your knowledge is about the tools, the better symphony you can compose.
The instruments that are needed to create the symphony that is AI, are:
- Natural Language Processing
- Machine Learning
- Deep Learning
- Reinforcement Learning
- Data Science
So, let’s understand the technologies behind AI
Natural Language Processing
Language is at the core of Human Communication. So, it’s natural that any intelligent agent we build must be able to communicate with us.
Natural Language Processing is an instrument of AI that deals with Human Language.
NLP Algorithms can read, decipher, understand and generate language.
You may have come across applications that use NLP, some of them are:
- Google Translate — It translates from one language to another.
- Grammarly — It can check grammatical errors in a document.
- Smart Assistants like Siri, Cortana, Alexa and OK Google — They communicate with a user by deciphering, understanding and generating language.
Machine Learning
Machine Learning is an instrument of AI that provides machines with the capability to learn and improve from experience without being explicitly programmed.
Machine Learning is divided into three fields —
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
Supervised Learning
It is a subset of Machine Learning that trains a mathematical model, using an algorithm, to predict patterns in data.
Google Photos is an application that uses Supervised Learning to create albums that collect photos automatically.
To understand Supervised Learning, let’s consider the following example —
You want to find out if it’s going to rain or not.
You know that when the humidity is high, it rains. So, you decide to train your model to predict rain based on humidity levels.
The input variable (Humidity) that is used to predict an outcome is known as Predictor Variable.
The target variable or output variable (Rain) that needs to be predicted is known as the Response Variable.
To make an accurate prediction, you will have to train a mathematical model with large amounts of data. This data is called the Training data.
Once you have trained the model, you can evaluate its performance by testing it against Testing data.
Unsupervised Learning
It is a subset of Machine Learning, in which unlabeled input is fed to the mathematical model and the model predicts an outcome.
The Model infers the relationship between various unlabeled data items (Think of Sherlock Holmes trying to deduce the occupation of John Watson from his clothes, appearance, wrist, etc) to predict an outcome. To do this requires a large training data set. Thanks to you all (and me) — Data is enormous.
Although it is supposed to be unsupervised, yet human intervention (judgement) is inevitable for now.
Deep Learning
Deep learning, a form of machine learning, is able to learn without human supervision, drawing from data that is both unstructured and unlabeled.
Here, we feed input to a neural network that predicts the outcome. Although artificial neural networks are analogous to our brain’s neural networks, they are not related in any way.
Neural networks are layers through which data is transformed. It is analogous to writing — You write the first draft, it's not very good so you have to fine-tune (or edit) it. Then it's better but still bad, so you delete certain parts or expand certain parts. You keep fine-tuning it until you are satisfied with the end result.
Likewise, neural networks adjust the input parameters on their own until they are happy with the outcome. Deep Learning algorithms can be applied to unsupervised tasks.
Tesla uses deep neural networks to detect roads, cars, objects, and people in video feeds from cameras installed around the vehicle. It combines Machine Learning with Computer Vision.
But, deep learning requires a lot of training data. This means a Tesla in the U.S.A is more powerful than the same Tesla in India because it has been trained on data of American roads.
Reinforcement Learning
Reinforcement Learning is a subfield of Machine Learning that learns by exploring. It learns by itself and there is no human intervention.
AlphaGo is a computer program that plays the board game Go. It was able to do this using reinforcement learning.
AlphaGo’s artificial neural network was extensively trained, both from human and computer play. It was trained to identify the best moves and percentages of winning using those moves.
The neural network combined with other algorithms became so powerful that it has defeated all the human Go players and It became the best Go player in the world.
Soon, it was succeeded by AlphaGo Zero, which learnt the game without any data — it learnt by playing the game against itself.
A paper released by DeepMind (The team behind AlphaGo) claimed that they generalized AlphaGo Zero’s approach into a single AlphaZero algorithm, which achieved within 24 hours a superhuman level of play in the games of chess, shogi, and Go by defeating all world-champion programs.
Data Science
Due to the Information explosion, new fields were created like Data Mining, Big Data, Data Analysis which would deal with extracting data from raw information and analyzing it.
Data Science is a new field that is a combination of existing fields like Statistics, Machine Learning, Big Data, Computer Science, etc.
The Ultimate objective of Data Science is to analyze a large amount of data and extract insights that lead to actionable decisions.
We are all used to the following phrases:
- Customers who viewed this item also viewed so-so ( on Amazon )
- Viewers of X also watch Y ( on Youtube )
Companies employ Data Scientists to analyze data pertaining to a customer and come up with suggestions or recommendations ( decisions ) that the customer is more likely to take. If the Data Scientist does a good job, then the company will massively profit from it.
Recommendation/Suggestion is one use-case of Data Science and of course, there are many!
The following are the various Components in Data Science —
- Acquiring data from RDBMS ( RDBMS is an abstraction of a server )
- Cleaning data
- Summarizing
- Exploratory Analysis ( Finding patterns )
- Making Predictions
- Deciding a Plan of Action
From music recommendations, fraud prevention, online banking to autonomous vehicles, AI is everywhere.
I believe it is important that all of us at least understand the basics of artificial intelligence and the technologies behind it.
This brings us to the end of this story. Stay tuned for more on AI!
Author’s Note: I am new to Artificial Intelligence. So, if there are any mistakes, please do let me know. All feedback appreciated.
© Shreya Sinha 2021
Written by Shreya Sinha, Thank you for reading.
Follow the Author on Instagram: Click here
This post includes affiliate links.
Read My Sci-Fi Short Story :
