An Origin Story

It was a chilly evening in January of 2019. I just had dinner in the hostel mess and I am returning to my room. On the way, I received the customary daily phone call from my mom. As usual, she sounds worried about my behavior and career and I try to avoid the conversation.

She is not wrong about her worry though. I was rather careless and ignorant about my studies and my career. I was so careless that I missed a class just because I was lazy even though I would have failed that course as a consequence. (As a side note, I eventually didn't fail that course.) I was so ignorant that I kept watching movies and playing video games even when it was time for me to do an assignment. I also didn't have any goals. So, yes, my mom's worry is justified.

Contrary to every day, I had a change of heart this time because I realized this conversation is in my interest too, considering I am in the final year of my bachelor's. I have to let her speak and stop avoiding this conversation. After all, there is no better person to discuss one's future with than a mother because she is the best well-wisher one can ever have.

By the time I reached my room, she told me about an online course. She apparently got to know about the course from a small news article. She tells me that the course teaches deep learning (DL). It sounded like a learning technique using which I can learn lectures deeply. I don't know what that is, so I asked her. I started to lose my temper because she is suggesting me to do a course in which both of us didn't understand what I would be learning. Before I started shouting, the next piece of information calmed me down. The course will be taught by a couple of professors from IIT Madras, one of the finest engineering institutes in the country. On top of that, the course fee is very low. If I don't like the course, I will lose only a small amount of money and a few hours of my time. If I do, I will find an area of interest and I will learn that from one of the best professors. The potential gain is much more than the potential loss from this course. I was convinced that I should give it a try wholeheartedly. I enrolled in the course on the same day.

The day I eagerly waited for has arrived. I am attending the first session of the course from the comfort of my hostel room. The professors explain the definitions of all the common jargon related to DL. I finally know what deep learning means. To put it very simply, it is one of the advanced areas of Artificial Intelligence (AI). Despite this, I found the session to be quite boring. This could be because I don't believe intelligence can be created in a computer. You can give step-by-step instructions to the computer for a task, but how can we encode intelligence into instructions? Towards the end of the session, I kept thinking to myself, "This is only the first session. Moreover, they taught only the definitions of common terms. It is, of course, boring. I should wait for the next session and see."

A couple of days passed and it was time for the second session. I hesitantly opened the video link for the same. Now, they are teaching the basics of Python programming language. I already learned it more than 3 years ago. I expected it to be boring like the first session. I still gathered some energy to pay attention. After a few minutes into the session, I realized I had forgotten most of what I learned about Python. My attention levels soared. Halfway into the session, they started teaching some basic external libraries like numpy and pandas which will be useful for implementing DL algorithms. That is entirely new to me. Now, my interest levels also surged.

For a few more sessions, they covered the fundamentals of DL. That included Python, mathematics, setting up a programming environment for lab sessions, implementing basic building blocks of DL, etc. There were ups and downs i.e., some were interesting and some were boring. I still got through all the basic sessions to find out if the main topic would be interesting or not.

Finally, it is the day of the session where I get to implement my first DL algorithm from scratch. It is a simple feed-forward neural network with only one hidden layer implemented using numpy. The objective is to train the neural network to recognize handwritten digits from images. The dataset to be used to achieve this is called MNIST. I presumed that the training would take a long time since the neural network has to figure out how to recognize the digits by itself. But, it took hardly 2 minutes! I could not believe my eyes when I saw the results. One of the simplest DL algorithms is successful in recognizing 99% of the digits accurately. I didn't have to provide any explicit instructions to do the task. I simply had to provide image and digit pairs to the algorithm and it figured out a way to recognize the digits. My earlier belief that intelligence cannot be created in a computer was shattered at that moment. How naive was I? 🤦🏻‍♂️ After a few minutes, I realized I had fallen in love with DL. 😍

For all the remaining sessions, my interest and attention levels were at an all-time high. I learned many DL algorithms such as convolutional neural networks (CNN), recurrent neural networks (RNN), gated recurrent units (GRU), and Long Short-Term Memory (LSTM) and I even got a sneak peek of recent developments such as transformers which are not part of the course. By the end of the course, I became a deep learning enthusiast even though I didn't know what deep learning meant when the course began.

I found my goal. I wanted to learn even more. In fact, I didn't want to limit myself to deep learning. I wanted to expand the horizon of my knowledge and learn other areas of AI as well. Thus began my search for an appropriate course/program. I found CDAC's post-graduate diploma in AI program to be suitable for me and joined that program. That's how my journey into AI/ML has started.

Thank you, Prof Mitesh Khapra and Prof Pratyush Kumar for initiating the affordable deep learning course and pushing me into AI/ML world.