What Are the Core Subjects in Artificial Intelligence (AI)?

What Are the Core Subjects in Artificial Intelligence (AI)?

6 mins read14.5K Views Comment
Updated on Dec 3, 2021 17:40 IST

Artificial Intelligence (AI) is a buzzword nowadays. Companies like Google, Facebook, Tesla, Amazon are doing cutting-edge work in AI. So, what is AI? Is it the same as computer science? What are the different topics covered under AI? These are the questions that might plague one given the recent interests and mentions it has generated. Even if we don’t know about the intricacies of AI, AI has started pervading our lives.

2021_12_iStock-957654454-1.jpg

 

AI as Part of Our Daily Experience

AI applications now help run our day-to-day lives whether we realize it or not. For e.g., there is AI in action when we travel and use Google Maps extensively. The Facebook algorithm automatically tags the photos when we upload pictures after meeting friends or some get-together at home. Siri and Alexa now interact with us, recognizing our voices and intelligently responding. There is an automatic reply (depending on the content of the trailing matter) suggested by the software when we get to reply to our emails. There are drones now which do topographical surveys and transport stuff in hard-to-reach terrains. There are many games on our smartphones that are powered by Artificial Intelligence. So, AI is not a futuristic technology anymore. It helps run our lives now in various forms.

AI in Scientific Research and Business Applications

Beyond the above day-to-day applications, AI is being used in scientific research across disciplines. Neural networks and other techniques of machine learning were used in developing the Covid vaccine. Needless to mention that AI is being used by businesses in various capacities across functions be it in the supply chain (e.g. tracking wastage, vehicle running status using sensors) or in finance (e.g. algorithmic trading) or in HR (in recruitment, training, and development) and in sales and marketing (marketing analytics, customer churn analysis).

Also Read – Top Real-World Artificial Intelligence Applications

Do I Need to Know About AI?

One might think that given the technical nature of AI, the study of AI is suitable only for people with computer science backgrounds; or, it is meant only for engineers. One might doubt- if I am a non-techie, can I attempt to learn about AI? The answer is – ‘Yes. AI is for everybody. For e.g., when the IT (Information Technology) revolution started in India in the late 90s, most of the people employed were non-computer science graduates. Later on, the IT industry needed people who understood the business models and processes of other core industries such as banking, manufacturing companies, oil and gas refineries, mining companies, hospitals, etc.

Thus, people having experience in various industries and sectors were employed with these companies. Note that these people had industry understanding and not necessarily the knowledge about coding or other IT skills. These skills could be provided as needed within the IT company by in-house training and development. Similarly, various functional skills were also in demand. So people with functional expertise in HR, supply chain, finance and accounts, engineering, sales got employment with these companies. In the same vein, AI is for everyone and not only for techies. You might be a fresher, a salaried individual, an entrepreneur, or a freelancer – understanding about AI and its transformative use cases would empower and reward you as a professional and would help your organization too in getting ahead of the race.

You may like – Future of Artificial Intelligence: Top Industries It Will Affect

What Does AI Comprise?

So, what should one learn about AI? – 1) As a generalist and non-techie and 2) As a learner interested in knowing about technical aspects of AI.

  1. We shall see that the subject matter of AI contains an understanding of Mathematics and basic Computer Science. However, even without knowing these technicalities, the following items are required to be understood by any AI generalist.
  • The meaning behind common AI terminology such as Neural Network, Label, Feature, Parameters, Hyperparameters, Training Data, Validation Data, Test Data, Bias, Variance, Regression, Classification, Activation Function, Regularization, etc.
  • Understanding about AI use cases and how AI technology could be applied to solve your unique business problem
  • What does it take to build an AI project – what kind of skill sets do you need? What kind of tools and processing capacity do you require?
  • Whether AI is required for your business problem? Could it be solved by conventional, non- AI tools? Because AI would require commensurate cost and might not be the most efficient way, given the constraints of your organization
  • Ethical issues in AI (garbage in, garbage out. So, the right kind of data for the problem at hand is important)

 

2021_12_iStock-998499074.jpg

AI is truly an interdisciplinary subject

 

  1. As an engineer or computer science/IT student/professional, you would like to go more in-depth and understand the technical aspects of AI. AI is truly an interdisciplinary field. Its canvas is vast. It’s an emerging field and vast opportunities lie ahead. Broadly (not exhaustively), the following topics across disciplines are covered understudy of AI

Mathematics

  • Calculus (Including partial derivatives) – useful in understanding how parameters (weights) get updated in backpropagation in neural network apart from other uses
  • Linear algebra – Matrix operations, Rank, Basis, Eigenvalue, Eigen Vectors, Dimensionality reduction (useful in Principal Component Analysis)
  • Probability – Basic probability, Expectations, Bayes theorem for Bayesian network and learning

Statistics

  • Basic statistics – Mean/Median/Mode, Standard deviation, Probability distributions such as normal distribution and its applications
  • Statistics – Skewness, Mean squared error (MSE), Hypothesis testing, ANOVA
  • Statistics – Correlations, Type I and Type II error, Precision, and Recall

Must Read – Statistical Methods Every Data Scientist Should Know

Computer Science

  • Sorting algorithms (quicksort, merge sort, insertion sort, etc.)
  • Shortest path algorithms (Dijkstra’s, A*)
  • Tree algorithms (pre-, in-, post-order traversal)
  • Memory requirement and computational cost
  • Data structures – Trees: binary search tree, heap, Queues, stacks, priority queues, Linked lists, Hash map, and Hash table
  • Various search algorithms – Breadth-first search, Depth-first search, etc. Uniform search, Iterative deepening search
  • Constraint satisfaction
  • Propositional logic, 1st Order logic, Backward and Forward chaining, Resolution Method
  • Markov decision processes (MDP)
  • Programming

(Being given separately as not every programmer is a Computer Science graduate)

  • Any object-oriented programming – C++/Java etc.
  • Any modern programming language such as Python
  • Basic data science operations – such as cleaning the data (removing duplicates), identifying missing data, parsing the data, data visualization (various charts, etc.)

Machine Learning – Basics

  • Types of learning – supervised, unsupervised, reinforced learning
  • Regression (Linear, Polynomial and Logistic regression), Classification
  • Various activation functions and loss functions
  • Gradient descent
  • Bias, variance tradeoff
  • Imp of training, test, and validation data

Deep Learning

Deep learning – A neural network with many hidden layers. Requires greater computing power.

  • Pre-training, Transfer learning
  • Autoencoders, Ensemble methods, Dropout, etc.

Computer Vision (Convolutional Neural Network – CNN)

CNN is used in image recognition

  • Convolution operation, 2D, 3D Filters, Max pooling
  • ConvNet, ResNet, GoogLeNet

Recurrent Neural Network (RNN)

RNN is used in sequential learning problems such as text/audio/video prediction

  • Word embeddings, LSTM algorithm
  • Backpropagation through time (BPTT)

Reinforcement Learning (RL)

RL is used in autonomous car driving, speech translation, Gaming (Famous AlphaGo program), Robotics, Algorithmic trading, etc.

  • Exploration – Exploitation tradeoff, Bandit Algorithm
  • Policy Gradient, Value Function
  • Temporal difference learning, Q learning
  • Dynamic Programming
  • Function Approximation

Deep Reinforcement Learning (Deep RL)

Internet of Things (IoT)

  • Involves learning about some hardware too)
  • Different types of sensors, Actuators, Wireless protocols
  • Cloud computing
  • Machine to Machine (M2M) and V2V (Vehicle to Vehicle) communication
  • Smart homes, Smart Grid, Smart city

Conclusion

Thus, it is clear that AI is a vast topic and spans multiple disciplines right from advanced mathematics, statistics, neuroscience, psychology, operations research, linguistics, robotics, etc. Its canvas is vast. It’s an emerging field and vast opportunities lie ahead. So, how to learn so much? The trick is to get started. As you become more familiar with the jargon and basic concepts, you would develop more comfort and interest in AI topics. Keep on building your learning. Like compound interest, sooner than later, you would be positively surprised with your learning mastery of the subject.

Read More:

About the Author

This is a collection of insightful articles from domain experts in the fields of Cloud Computing, DevOps, AWS, Data Science, Machine Learning, AI, and Natural Language Processing. The range of topics caters to upski... Read Full Bio