Start reading, only if you have the commitment to stay with me till the end & will not come to any conclusions in the middle.

Assuming you are quite familiar with these buzz words, Artificial Intelligence, Machine Learning, Data Science & Deep Learning. If not then just google those words, watch some videos on youtube & come back.

AI in a nutshell

Think of it as a mathematical equation, that takes numbers as input and after performing some computations, spills out numbers(again) as the output.

Any format of data eg., text, images, videos & audios will first get (Encoded) converted into a representation of ‘numbers’ allowing us (& the computers ) to perform almost any kinda mathematical operations from additions to differentiation on them & get the resultant, later those results will again get converted(Decoded) into our desired format.

here are a few examples of how data is getting converted into a set of numbers:

An image is a huge set of tiny units called pixels that contains red, green & blue color content in ’em (RGB values).

Whereas a video is just a collection of 24 images per second arranged in sequential order.

Text is a collection of letters and words which can be represented in several ways. here, each word in the sentence was assigned to a vector of size of the number of unique words.

Similarly, even the audio format of the data can be converted into a set of numbers ranging from fixed bounds.

Mostly we convert all these representations of a set of numbers into something called “Tensors”. These tensors, flow throughout our model, hence the name “TensorFlow” was given to the popular deep learning framework by Google.

Every AI application you were using so far, first takes the input, converts that into tensors, feeds those into a mathematical equation (say BlackBox)that consists of some thousands to billions of variables, perform computations & then the resultant tensor will again get converted to your desired format.

Our job is to model the Black Box in such a way, that it would give fruitful results based on the “hidden pattern” captured by it while being trained.

The technology or sub-field of AI we will be used depends on the task we want our Black Box to accomplish.

If it is required to work on image/video data then we use something called Computer Vision.

Eg., Face recognition, Object detection, Image enhancement, etc.

Check out how Tesla uses Computer vision for self-driving.

When a task involves textual or audio data format we use something called Natural Language Processing (NLP).

Eg,. Chatbots, Virtual assistants like Siri & Alexa.

Deep Reinforcement Learning is another sub-feild of AI being used to empower intelligence in robots & bots.

Eg., AlphaGo & OpenAI Five, etc.

Where to start?

As it would be good to have some certification under your belt, I would highly recommend you to take IBM professional data science followed by Deep learning by Andrew ng specializations on Coursera.

Do check out guided projects by Coursera, where unlike traditional online courses you would have hands-on experience in real-time..!!

There are a hell lot of great *FREE * resources out there including Youtube channels, Books & Blog posts you will come across.

No matter where learning from, be it a blog, book, youtube video, or even a StackOverflow query. Just make sure you had learned the following,

0. Python programming language

Trust me, you don’t need to master the language. Basic syntax, loops, functions, inheritance & import statement is enough..!!

No need for learning Data structures & Algorithms..!!

Most students learn Python & get stopped there itself. Just because they aren’t aware of all these. I hope you will share this article with your friends who already or learning python or any other programming language & join me in promoting STEM Education.

NOTE: You don’t need to mug up all the functions, you will be googling a lot while writing code anyway. So you just need to know the functionality & the various features of the packages.

0.5. Surfing

You got to be good at surfing internet, get familiar with websites like Medium, Quora, StackOverFolow, Github & Google Colab.

1. Basic Python libraries

>OS(to read files),

>NumPy(for math),

>Pandas(for .csv files),

>Matplotlib(for graphs),

>OpenCV(for image processing),

>SciPy(for text processing),

>Librosa (for audio processing).

2.Machine Learning

>Sklearn

>XGBoost

3.Deep learning

Don’t get worried about the framework, just go for Keras (since it is very beginner-friendly). later learning PyTorch won’t take much time anyway.

The following are a few great resources to get started with Deep learning.

>deeplizard

>Jeff Heaton

3.5 NLP with the “Hugging Face” library

This library allows you to download & use huge pre-trained language models like GPT2 & BERT for Natural Language Understanding (NLU) tasks, such as analyzing the sentiment of a text, and Natural Language Generation (NLG), such as completing a prompt with new text or translating in another language.

The best thing I love about this library is, it offers something called “Pipelines” that allows you to use STOA huge language models for infrence in a SINGLE LINE OF CODE!. Making it very useful for rapid pprototyping.

For all the libraries mentioned above their ‘’Official Documentation” not only does a great job in explaining almost everything but also saves hell lot of time!

Also, the official documentation allows you to peek into the source code in GitHub & get an in-depth understanding of “What makes something do something”.

4.Kaggle contests & Hackathons

You gotta respect all domains..!!

Start doing projects, once you had learned everything it is important to start building stuff on your own.

Since you are using the libraries built by the experts if you ran into an error, it is very likely due to incorrect datatype or dimensions rather than a logic error.

While participating in Hackathons, you could able to share your work through a web/android/iOS app where the model is deployed on the cloud.

For that, I would highly recommend learning web frameworks and cloud platforms for yourself instead of relying on your teammates.

Do check out Streamlit, a minimal web framework & Heroku, a cloud platform that lets you turn your AI project into a web app almost instantly..!!

Kaggle competitions were not that easy. It requires a lot of practice to ace ’em. Learn from the solutions made by the winners of past contests. By doing so, you will be learning a lot, till now you’ve learned how to build AI models but this allows you to learn how to transform the data provided in such a way we get most of out of it.

Going beyond the “known solvable problems”

By now, you would have some projects & a lot of experience to put in your portfolio that lands you in good internships(maybe even a job). Start applying for related roles through LinkedIn. Also, join the Slack channels related to AI where you can grow your network.

Now, it’s time to start thinking like a researcher. Since till now, you had worked on the AI solutions which were already being implemented by Tech giants Eg., Face recognition, Chatbots & Autonomous vehicles.

Try to read & understand the research papers (I know it’s really hard) and mimic the work done by the tech giants like OpenAI, Deepmind & Nvidia.

This is the final stage of your journey, Do a lot of experimentation, no matter how much you learn there will always be new things to learn. Take the world from another point of view. Try to explore other interesting fields like Deep Reinforcement Learning, Quantum Machine Learning & Artificial General Intelligence.

Remember, never stop learning..!!

Ending note

I hope you will come up with innovative ideas, one day start writing your research paper, start contributing to this world and make it a better place💖.

While going through your journey, you could write books, blogs, college article, start giving seminars to your peers & others, make youtube videos, record podcasts & host ’em on your college website to educate people, build a community to share your knowledge to do large scale projects, learn & grow together.

Check out my book, YouTube channel & blog.

Feel free to WhatsApp me +91 8886940907.

Stay motivated

Follow Daniel Bourke, Lex Fridman & Two Minutes Papers to stay updated.

Remember the fact that AI has the potential to solve numerous problems including health, poverty, education & the environment. And never lose hope..!!

THANK YOU