Learning to use AI tools like ChatGPT can
make you more productive at your job. But, learning to build AI tools like ChatGPT will
make sure you have a job to be productive at. Will AI take over your job? Well, I don’t know
about that. But one thing I do know is that building AI tools would be one of the last jobs AI
can replace. So, if you want to be future proof, you might want to invest some time into
learning to build AI. And if that’s not a good enough reason for you, you will be
shocked to know that OpenAI, the company who built ChatGPT pays almost 1 million dollars
to its AI engineers. In this video, I will give you a step by step guide on everything you need to
learn to be able to create AI tools like ChatGPT. This video is especially important for someone
who already knows a little bit of programming or Math and wants to transition into
an AI related job.
Let’s do this. Human Intelligence comes from the transmission
of information through Neurons. Neurons are nothing but a bunch of interconnected
nodes in our brain. In the same way, Artificial Intelligence also comes from
a network of interconnected nodes called Artificial Neurons. And these networks are
also called Artificial Neural Networks or simply Neural Networks. To be able to build AI
tools like ChatGPT, we need to learn how to build these Neural Networks. But to get there,
we will have to take many smaller steps. You see, neural networks are part of this
field called Deep Learning. In Deep Learning, we take a neural network, train it by showing
a lot of data and use the trained network to predict something.
What to predict can vary
depending on the task you assigned to the network when you trained it. In the case of
ChatGPT, the network is actually predicting the next word of a sentence. I know it doesn’t
make sense right now, but stay with me until the end and I’ll explain this in much more detail.
Moving on, Deep Learning is a subset of this field called Machine Learning in which machines
acquire the ability to learn. In other words, Deep Learning is not the only way a machine
can learn to predict. There are many other ways to do that and all of them combined
are part of Machine Learning. To be able to learn Neural Networks and Deep Learning,
we need to master Machine Learning first.
But how do we do that? Well, Machine
Learning has 3 pillars. Mathematics, Statistics and Programming.
Let’s tackle them one by one. Let’s start with Mathematics because you will need
good knowledge of Math to learn Statistics. Linear Algebra, Calculus and Probability theory sit
at the core of Machine Learning. And most of us have been burnt by these either in high school
or college.
But the good news is that you need these concepts only to learn how different
Machine Learning algorithms work. Once you have learnt that, you will just write programs
that implement these algorithms for you. So, I don’t believe that you have to be a Math
genius to be able to do Machine Learning. To learn Math for Machine Learning, I recommend
this specialization called Mathematics for Machine Learning and Data Science on Coursera.
This course is created by DeepLearning.ai which was founded by Dr. Andrew Ng who is arguably the
most well known professor of Machine Learning. Later in the video, we will go back to Prof. Ng
when we want to study Machine Learning. Anyway, this specialization consists of 3 courses. One for
Linear Algebra, another for Calculus and the last one for Probability. If you think all of this is
too overwhelming for you, you can also check out this Data Science Math Skills by Duke University.
This course is not as comprehensive as my other recommendation.
If you are someone who doesn’t
want to get too involved with the Math behind Machine Learning or already knows the Math and
just wants to brush up, this course is for you. Now that you know Math, let’s move on to
Statistics. Now, Statistics is a very vast field and it requires many many years to
fully understand it. But luckily for us, we don’t need to know everything for
Machine Learning. For Statistics, we will use a Breadth First Approach for learning.
We will learn some basic core concepts and then build upon them as we encounter new Machine
Learning algorithms later. To learn all the key concepts that you actually need, I recommend
this course called Introduction to Statistics by Stanford University. This course covers all the
important ideas like Probability Distributions, Central Limit Theorem, Confidence Intervals
and Regression etc.
By the end of this course, you would know all the Statistics you
need to get started with Machine Learning. Before we can finally do some Machine Learning,
we need to learn some programming. To be more specific, we need to learn programming in Python
because it’s the most popular choice when it comes to Machine Learning. Now, there is some talk
in the town about this new programming language called Mojo which is compatible with Python and
is 35,000 times faster. But it’s still too early to predict the future of Mojo. So, we are going
to stick with the time tested Python for now.
For the purpose of Machine Learning, we don’t need
advanced level programming skills. If you know the basics like if statements, loops, functions
and classes etc., you should be able to pick Machine Learning easily. So, we are not going to
build any crazy projects in Python at this step and would focus on the basics. But we will build
some Machine Learning projects using Python later in the video. To learn these basics, simply go to
learnpython.org and do some hands-on exercises. At last, we have reached a place where we can
start doing some Machine Learning. We are just one step away from building tools like ChatGPT
now. For Machine learning, we need to go back to Prof. Andrew Ng. Check out his Machine Learning
Specialization on Coursera. This specialization is divided into 3 courses. One for supervised
learning algorithms like Linear and Logistic regression.
Another one for unsupervised
learning algorithms like Clustering. And the last one for advanced algorithms that
also introduces you to Neural Networks. If you really want to have a deep understanding
of ML, this is the best course out there. The only caveat I would like to mention here is that
the code samples and the Jupyter notebooks that let you actually play with the code are
not available for free with this course. Once you are done with the course, head over to
Kaggle and do some hands-on practice. On Kaggle you can see the projects that other people have
built.
You follow along in the beginning and build some confidence. When you are comfortable, you
can participate in one of their competitions. This will do 2 things. One, It will give you confidence
that you can complete Data Science projects independently. Two, you will build a portfolio
of projects that you can write in your resume. Now that you feel confident about your Machine
Learning skills, let’s go back to our original goal which was to build AI tools like
ChatGPT using Neural Networks and Deep Learning. Machine Learning specialization by Dr.
Ng already gives you an introduction to Neural networks. But it’s not comprehensive enough
for you to be able to understand advanced AI systems like ChatGPT. For that, you will have
to develop your skills in Deep Learning. But there’s no need to worry because Dr. Ng also
offers a specialization in Deep Learning. First 3 courses in this specialization cover
basics of Deep Learning which is basically how to train Neural Networks.
In the fourth
course, you will learn about Convolutional Neural Networks which is basically Computer
Vision. In Computer Vision, you will learn how to train machines to recognize patterns
in images which has applications in Autonomous Driving and Face Recognition etc. But if your
main goal is to understand how ChatGPT works, that’s part of Natural Language Processing which
is the last course in this specialization. This course covers Transformer architecture
which is what Chat GPT uses. By the end of this specialization, you will know everything
you need to have a successful career in AI. I know that this path can seem very long to many
people. But, that’s the cost you pay to work on the next generation technology.
Another path
that is closely related to Machine Learning is that of Data Science. In Data Science, you use
data to develop insights but you don’t need to be a Machine Learning expert. If you want to
know the fastest way to learn Data Science, watch this video. My name is Sahil
and I will see you in the next one..