How I'd Learn AI in 2024 (If I Could Start Over) | Machine Learning Roadmap

Summary

In this video, Asan Sharma shares his insights on becoming an AI engineer in 2024, highlighting the booming demand for AI professionals. He outlines a step-by-step learning path, starting with mastering math, Python, and data analysis libraries, then moving on to machine learning frameworks and models. Sharma emphasizes the importance of supervised, unsupervised, and reinforcement learning, and suggests practicing on platforms like Kaggle. He also delves into deep learning, neural networks, and generative AI, recommending courses and tutorials to build expertise. The video concludes with advice on leveraging large language models and creating custom AI applications.

Takeaways

• 🚀 AI is predicted to be the biggest trend of 2024 and beyond, with high demand for AI engineers.
• 📚 To become an AI engineer, start with a strong foundation in mathematics, focusing on calculus, linear algebra, and probability.
• 🐍 Learn Python, the most widely used programming language in AI, and understand its basics including data types, control structures, and OOP concepts.
• 📊 Master data analysis with Python using libraries like NumPy, Pandas, and Matplotlib for handling and visualizing data.
• 🛠️ Choose a machine learning framework such as PyTorch, PyTorch, or TensorFlow to create and train models.
• 📝 Understand the three types of machine learning: supervised, unsupervised, and reinforcement learning, and their respective applications.
• 🔍 Practice by solving problems on platforms like Kaggle using various datasets to apply your machine learning knowledge.
• 🧠 Dive into deep learning by learning about neural networks, backpropagation, and hyperparameters.
• 🖼️ Explore Convolutional Neural Networks (CNNs) for image classification and natural language processing (NLP) for text data.
• 🤖 Build generative AI applications by understanding large language models like GPT and learning prompt engineering.
• 📱 Stay updated with the latest AI trends and tools, such as GPT plugins and the GPT store, to create custom AI solutions.

Q & A

What is the prediction for AI in 2024?

AI is predicted to be the biggest trend of 2024, with a booming demand for AI engineers due to the launch of applications like chat GPT and other generative AI tools.

Who is the speaker in the video?

The speaker in the video is Asan Sharma, who started learning about machine learning and AI in 2019.

What is the basic definition of machine learning?

Machine learning is a process through which a system can recognize patterns and predict future outcomes.

What are the key mathematical concepts to understand for AI and machine learning?

The key mathematical concepts include calculus (differentiation and integration), linear algebra, and probability.

Why is Python important for AI engineers?

Python is the most used programming language in the AI field due to its simplicity and the vast number of libraries available for machine learning and data analysis.

What are the three libraries used for data analysis with Python?

The three libraries used for data analysis with Python are NumPy, pandas, and Matplotlib.

Which machine learning frameworks are recommended for beginners?

For beginners, PyTorch and scikit-learn are recommended as they are simpler to use compared to TensorFlow.

What are the three main types of machine learning models?

The three main types of machine learning models are supervised learning, unsupervised learning, and reinforcement learning.

What is the role of Kaggle in learning AI?

Kaggle provides a platform for learners to practice their AI skills by solving real-world problems using various datasets.

What is the significance of neural networks in deep learning?

Neural networks are the foundation of deep learning, consisting of layers of neurons that process inputs and generate outputs, allowing the model to learn complex patterns.

What is a CNN and how is it used in AI?

A Convolutional Neural Network (CNN) is a type of deep neural network used for image recognition and classification by processing pixel data and identifying patterns within images.

Outlines

00:00

🚀 AI and Machine Learning Trends in 2024

The video script introduces AI as the biggest trend of 2024, with a focus on the booming field due to applications like chat GPT. The speaker, Isan Sharma, shares his journey in machine learning and AI since 2019 and encourages viewers to learn AI engineering. He mentions launching a generative AI app called Centur and offers a 20-minute guide to learning AI from scratch, including free courses and steps to build generative AI applications. The script emphasizes the importance of understanding AI and machine learning, defining machine learning as a pattern recognition process to predict future outcomes.

05:00

📚 Foundational Math for AI

The script outlines the foundational math required for AI, including calculus, linear algebra, and probability. It suggests resources like Khan Academy and freeCodeCamp for learning these concepts. The importance of understanding integral calculus, matrix operations, and the concept of probability in AI is highlighted. The speaker reassures that these topics can be learned easily and are crucial for building machine learning models.

10:01

🐍 Learning Python for AI

Python is introduced as the most used programming language in AI, with a focus on its simplicity and widespread use. The script recommends learning Python basics, such as data types, conditional statements, loops, functions, and object-oriented programming. It suggests using resources like Tech with Tim and freeCodeCamp for learning Python. The goal is to understand the basics and apply them to real-world projects, with an emphasis on not getting bogged down in every detail but focusing on the goal of becoming an AI engineer.

15:03

📊 Data Analysis with Python Libraries

The script discusses the importance of data analysis in AI, focusing on three key Python libraries: NumPy, pandas, and Matplotlib. NumPy is used for numerical operations, pandas for handling tabular data, and Matplotlib for data visualization. The speaker explains how these libraries are essential for building and optimizing machine learning models and suggests learning them to analyze and present data effectively.

🤖 Machine Learning Frameworks and Types

The script introduces various machine learning frameworks, recommending PyTorch or scikit-learn for beginners due to their simplicity. It explains the three main types of machine learning: supervised, unsupervised, and reinforcement learning. The speaker suggests taking a machine learning specialization course on Coursera to learn the basics of these types. The script also touches on different types of problems within supervised learning, such as regression and classification, and mentions specific models like logistic regression and K-nearest neighbors.

🧠 Deep Learning and Neural Networks

The script delves into deep learning, starting with the concept of neural networks and backpropagation. It suggests resources like Andre Karpati's YouTube tutorials and the cs50 AI course for in-depth understanding. The speaker explains the importance of learning about hyperparameters and how they affect the model's output. The script also introduces convolutional neural networks (CNNs) for image classification and natural language processing (NLP), recommending courses by Hugging Face and exploring generative AI tools like GPT and stable diffusion.

🌐 Building Generative AI Applications

The script concludes with guidance on building generative AI applications, suggesting tutorials on chat GPT and exploring tools like Hugging Face and Lang chain. It mentions the potential of customizing GPT models and the GPT store, which allows users to build and monetize their own AI tools. The speaker encourages viewers to learn about prompt engineering and building AI tools from scratch, providing a comprehensive roadmap for becoming an AI engineer in 2024.

Keywords

💡AI

Artificial Intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think, learn, and make decisions. In the video, AI is the central theme, with the speaker discussing its growing importance and how to become an AI engineer in 2024.

💡Machine Learning

Machine Learning is a subset of AI that involves the development of algorithms that allow computers to learn from and make predictions or decisions based on data. The video emphasizes machine learning as a fundamental skill for AI engineers, with a focus on pattern recognition and prediction of outcomes.

💡Generative AI

Generative AI refers to AI systems that can create new content, such as text, images, or music. The speaker mentions the launch of chat GPT and other generative AI applications, highlighting the demand for AI engineers skilled in creating such systems.

💡AI Engineers

AI Engineers are professionals who design, develop, and maintain AI systems. The video suggests that there is a high demand for AI engineers due to the booming field of AI and the increasing number of AI applications.

💡Python

Python is a high-level programming language known for its readability and ease of use. In the context of the video, Python is the recommended language for building AI models, and the speaker suggests learning its basics as a foundational step.

💡Data Analysis

Data Analysis involves the process of cleaning, transforming, and analyzing raw data to extract useful information. The video mentions data analysis as a crucial step in AI development, using libraries like NumPy, pandas, and Matplotlib for Python.

💡Deep Learning

Deep Learning is a branch of machine learning that uses neural networks with many layers to model complex patterns in data. The speaker discusses deep learning as an advanced topic for AI engineers, involving concepts like neural networks and backpropagation.

💡Neural Networks

Neural Networks are computational models inspired by the human brain, consisting of layers of interconnected neurons. The video describes neural networks as the building blocks of deep learning, where the model learns to optimize outputs through training.

💡CNN (Convolutional Neural Networks)

CNNs are a type of neural network architecture specifically designed for image recognition and classification. The video mentions CNNs as a key concept in AI for processing and understanding image data.

💡NLP (Natural Language Processing)

NLP is a field of AI that deals with the interaction between computers and human language. The video suggests that understanding NLP is important for AI engineers working with language-based data and applications.

💡Generative Models

Generative Models are AI systems that can create new data samples that resemble a given dataset. The speaker encourages viewers to explore generative models like chat GPT and other tools to build their own AI applications.

Highlights

AI is predicted to be the biggest trend of 2024.

The demand for AI engineers is booming with the launch of chat GPT and other generative AI applications.

The basic definition of machine learning is a process for systems to recognize patterns and predict future outcomes.

Mathematics, specifically calculus, linear algebra, and probability, is the foundation for AI learning.

Python is the most used programming language in the AI field and is essential for AI engineers.

Data analysis with Python involves using libraries like NumPy, pandas, and Matplotlib.

Frameworks like PyTorch, scikit-learn, and TensorFlow simplify the creation of machine learning models.

There are three main types of machine learning: supervised, unsupervised, and reinforcement learning.

Supervised learning involves labeled data, unsupervised learning deals with unlabeled data, and reinforcement learning optimizes for a reward.

Deep learning involves neural networks, which are layers of neurons for input and output processing.

Convolutional Neural Networks (CNN) are used for image classification by understanding pixel patterns.

Natural Language Processing (NLP) and Recurrent Neural Networks (RNN) are advanced AI topics for understanding language and sequences.

Generative AI and Large Language Models (LLMs) like chat GPT are the next frontier for AI engineers.

Learning about prompt engineering and customizing LLMs is crucial for building AI tools.

Chat GPT plugins and the GPT store offer opportunities to build and monetize custom AI models.

The video provides a comprehensive guide to becoming an AI engineer from zero to building generative AI applications.

The importance of understanding the basics and continuously learning is emphasized for AI engineers.

Practical applications and projects, such as Kaggle competitions, are recommended for hands-on learning.

The video concludes with a call to action for viewers to engage with the content and ask questions.

Transcripts

here is my prediction AI is going to be

the biggest trend of 2024 and going

forward in this decade hi everyone I'm

asan Sharma I started learning about

machine learning and AI back in 2019 and

2020 and today this field is booming

like never before with the launch of

chat GPT and other generative AI

applications there is so much demand for

AI engineers and because of that if you

want to become one this video is the one

you need to watch we even launched

centur which was a generative AI app

last year I'll share with you everything

you need to learn to become an AI

engineer from zero to building

generative AI applications in the next

20 minutes we'll break it into steps

talking about the exact free courses

that you can take and become an AI

engineer for completely free and if that

sounds exciting hit the like button and

subscribe and let's start with the topic

of today's video how I would learn Ai

and machine learning in 2024 if I had to

start all over again the most important

thing to understand is what is AI or

machine learning the basic definition is

machine learning is a process through

which a system can recognize patterns

and predict future outcomes that is the

basic idea of it so imagine you have a

blackbox you first give it an input and

you tell what the outut come is and

hence you are training this blackbox

what happens next is it tries to

identify patterns between the input and

the output and once you've given it

enough training data you then ask it a

question and then you ask it to give you

an outcome and a probability for that

outcome and that is what machine

learning is all about again you don't

need to stress about anything we'll

break everything down into steps in the

next few minutes the first first

important thing to learn is going to be

maths maths will be the foundation upon

which you'll build all of your learnings

so the most important thing to learn in

maths will be calculus differentiation

and

integration understanding about linear

algebra and lastly probability you need

to have an intuitive idea of what these

things really mean first of all you

should understand integral calculus and

differentials you should understand how

matrices really work and what is the dot

product of two matrices and at the end

you should understand understand about

probability probability is going to be

very important as we move on because any

result that a model gives is not a yes

or no it is a probability it will say

that I think this photo looks like a dog

the probability of that is

.95 I think this photo does not look

like a dog the probability for that

is98 all of this is very simple you can

you can learn it from three blue one

Browns videos on calculus as well as

matrices and probability or you can go

to KH Academy or free code Camp has some

amazing videos on mathematics so go and

have a look at those it's very simple

and once you understand the basics of

maths now we can move on to the next

step right now all of the things that we

are learning in maths we'll be applying

when we'll be building these machine

learning models when we will be training

these models when we will get an outcome

from this model so it's all going to be

relevant stick around till the end the

second step is to learn about python now

python is the most used programming

language I think in this age after

JavaScript maybe and it's super easy to

learn it's similar to English I learned

it myself in 2019 I even taught python

to thousands of students back in 2020

and 21 on an academy so it's a very

simple language to learn right I I

taught it to school kids if you're in

college you can very simply learn it now

Tech with Tim has a great playlist which

will give you the basic idea of python

or you can also check out free code cams

tutorials on python but you need to

understand the basics of this

programming language because this is

what you will use to build all the

models the first thing to learn is data

types learn about F's conditional

statements learn about for Loops how can

you create functions and then learn

about oop Concepts and when you have

have all of these things figured out try

building some simple projects try just

rebuilding some games and see how well

can you do it and you can even take a

step further and learn about the

different libraries that python really

has and that will give you a basic idea

of how do you use your knowledge of

python Concepts and apply it into

building something in real world right

and that is super important so once you

have the basic idea of python again you

don't need to know every single thing

because I think learning in 2024 is a

lot less about completing the entire

textbook knowing all the rules of the

game it's more about understanding the

basics going about with it if you face

any difficulty error or something that

you don't know Google it stack exchange

you can also just ask chat GPT for some

answers but most important is to learn

the basics and then keep moving forward

because what will happen is you'll find

it too hard to learn everything about

python python has just too many

libraries too many packages too many

applications don't focus on all of that

just understand the basics and keep your

eye on the goal which is to become an AI

engineer the next step after learning

python is to learn about data analysis

and data analysis with python is all

about using three libraries the first

one is numai second one is pandas and

the third one is is matte plot Li numi

will basically help you with numbers you

will be storing all the data in arrays

and then you'll have multi-dimensional

arrays which will become matrices and

then you'll be doing a lot of dot

product of matrices to come up with

answers again this will be relevant when

you'll be building the models when

you'll be coming up with parameters when

you'll be optimizing an equation I'll

talk all about that in later but you

will have to use numpy arrays to do all

of that the second thing is pandas

pandas is a library which will basic

basically help you with data which is in

tabular form right so any data that you

have stored in columns and rows pandas

will help you get it sorted and this is

super important because any data set

that you use or you make it it's all

going to be in a file called CSV comma

separated values and you should know how

to use it and how to extract data from

that table from that tabular form and

you can do that very easily with pandas

again it all has functions that you can

use so it'll be like PD do this or the

same is the case with numpy arrays so

you create an array and then you say np.

array and then you perform functions on

those arrays right that is how it works

the last library that I talked about in

data analysis is going to be matte plot

lib matte plot lib is really important

because once you run the model you have

some outcomes you have some answers you

have a probability and you have to

Showcase it you need to show what the

end result is and that is what you do

with the help of matplot lib it's a

amazing Library through which you can

actually plot bar charts you can plot

pie charts you can plot various ways to

represent data right and that is where

Matt plot lip comes very handy you use

numpy you use pandas and you use Matt

plot lip to analyze any data set that

you would be getting so once we've

learned about python once we learned

about data analysis the next step is to

pick pick a framework now there are

various Frameworks which make your life

easy you can use these Frameworks and

you can very easily create different

types of machine learning models you can

either choose py torch pyit learn or

tensorflow or there are many others as

well but these are the three most

popular ones I would suggest you to go

with pytorch or psyit LA because these

are simpler Frameworks to use as a

beginner tensor flow just hides a lot of

the beginner stuff away from you so it

might look really clean you can just use

three four lines of code to execute

something but you won't understand the

essence of what you are building hence

tensorflow you can pick once you have

the basic idea of building models and

tensorflow makes it really simple and

you can quickly spin up models with the

help of tensorflow and kasas now the

most important thing here is to pick one

framework and master it you can find

numerous tutorials on free code Camp

about p TCH along with psychic learn so

you focus on learning each of these what

will basically happen is you learn about

the various types of machine learning

models essentially there are three types

of machine learning models the first one

is a supervised learning the second one

is unsupervised learning and the last

one is reinforcement learning in

supervised learning the data that we

give to the model is labeled data so we

are actually telling it that this photo

is of a cat and this photo is of a dog

and when you have un learning in that

case data is unlabeled right there's no

label on the data the model can just

tell that this is category a and this is

category B that is what unsupervised

learning is all about and the last thing

is reinforcement learning and

reinforcement learning is all about

creating an incentive for a character in

a game or any environment and then it

learns how to optimize for that number

for example if you have a character in a

game you create an incentive that when

it walks towards the price it gains more

points when it walks away from it it

will lose points and hence you are

basically training the model telling it

that you should move closer to the price

to increase your points and the goal of

this game is to increase the points as

much as possible so that is your

reinforcement learning and these are the

three main types of machine learning

models that you can create now there's

an amazing course on course era that you

can take by Deep learning AI called as a

machine learning specialization you can

individually enroll and audit each of

the three courses that are in that

specialization for completely free and

you can learn about all of these Basics

right so when you go through that it is

I think the most popular course in

machine learning so take that course and

you learn the basics of supervised

learning un supervised learning along

with reinforcement learning now in

supervised learning you have various

types

you can either have regression problems

or you can have a classification problem

right regression is the one in which you

have some data and you have to predict

future data and in the case of

classification as the name suggests you

have to classify between two or three or

more categories so you can create

multiple types of models regression

analysis can have logistic regression

then you have polinomial regression then

you have classification models along

with that you also have something called

as K nearest Neighbors and all of these

are going to help you predict something

it is all going to help you train that

model these are all approaches to solve

a particular problem you need to choose

which one is the best and when you go

through different practices when you try

things out a lot you will understand

which model to use for which type of

problem that is essentially how it works

in UNS supervised learning you have

clustering which is called as K means

clustering and then in reinforce po

learning there's just one type which

I've already explained to you in which

you are optimizing for the number of

points of that character now for

reinforcement learning particularly Tech

with Tim on YouTube has a tutorial on

Flappy Bird AI so go check that out and

you will learn exactly how to build it

from scratch and you will understand how

can you incentivize a character which is

the Flappy Bird to actually play the

game on its own with the help of

reinforcement learning so go check that

out but basically once you learn the

different types then you can start going

deeper and building projects The Next

Step would be to practice as much as

possible with the knowledge that you now

have accomplished so you go to kaggle

and you start solving problems they will

be data sets you take them and then you

try to solve the problem that they have

stated there are very famous data sets

like the Titanic data set the most

cliched one you have cfr1 you have Mist

you have fash mnist and you can start

applying all of your learnings on these

data sets Titanic data set would have

information about the passengers on

board Titanic and you have to predict

something the second one cfr1 has random

images of 10 different objects and

animals now mnist is a collection of

images which will be having numbers

written in it and you have to basically

build a model that recognizes those

digits and lastly we have fash mnist

which is going to be clothes so it

should be able to identify is this a

jacket is this a lower is this a cap is

this a shoe things like that I hope

you're understanding what I'm saying now

when you're are solving for amnest or

fash amnest you need to learn about

something called as neural networks and

that is where we start our journey into

deep learning right neural networks is a

very interesting and fascinating concept

just think of it like a layer of neurons

in which you'll be having inputs and

outputs you are having multiple inputs

you're putting that through a black box

which comprised of a layer of neurons

and then you have an output you are

training this model to actually optimize

and get the best result possible and

here is where you should learn about the

concept of back

propagation right the next step is for

you to have multiple layers of neural

networks and that is deep neural

networks right so you would be having

multiple layers you would be having an

input layer an output layer and multiple

layers in the middle depending on how

complex the problem is so the more

random the situation is the more layers

you will need to effectively solve a

problem and that is essentially how it

works you will also then have to learn

about something called as hyper

parameters you learn about keywords like

learning rate weights and biases and how

do you manipulate these three numbers to

get the optimal output to get the best

probability for the right output so that

is how it works again to learn about

neural networks you can check out this

amazing tutorial by Andre kapati on

YouTube it's completely free you can

also check out this amazing cs50 AI

course wherein they take you in the

depths talking about machine learning

and the neural networks so check out

that part where they talk about neural

networks that will open up your eyes a

lot so once you know about neural

networks The Next Step then is to move

to the next thing which is called a CNN

convolutional neural networks which will

again be discussed in the cs50

artificial intelligence course so you

will learn how do you use images and you

classify between images see an image is

built up of pixels and each pixel has a

RGB value red green and blue so this

will be having three values on one pixel

and there can be let's just say 108 T

into 1920 now think of these pictures in

a data set which is labeled so it says

that this is a dog this is a cat and

you're feeding it into a neural network

and hence you would basically create a

neural network which looks sort of like

this and this is the idea of CNN

convolutional neural networks again what

happens in between is the blackbox and

your model will try to understand the

patterns and will optimize those weights

and biases and learning rate and EPO and

everything to come up with the accurate

outcome and the more you train it the

better the outcome will become so that

is essentially how that works you will

learn about CNN the next step is up to

you you can learn about NLP natural

language processing you can also learn

about rnns and there's an amazing course

by hugging face on NLP that you can

check out and get started with that so

now that you've learned the basics of

deep learning deep neural networks

convolutional Neal networks the next

step is for you to start tinkering

around with ch GPT and generative AI you

can now perfectly start building AI

tools because you have the basic

underlying knowledge of how these neural

networks really work the next step is to

check out deep learning AIS tutorials on

chat GPT the first one is about how chat

GPT works you'll understand how large

language models operate and you'll also

learn about the art of prompt

engineering ing that will get you

started they have other courses as well

which talk in depth about using the open

AI apis to build your own models from

scratch it also talks about how do you

build a large language model from

scratch how do you customize a large

language model how do you use tools like

Lang chain to actually use models in any

particular environment so there's lots

of things that you can explore now check

out

deeplearning.ai they have some amazing

short courses that you can take to learn

the basics of large language models you

can learn about how do you deploy these

models how do you take advantage of the

existing tools like hugging face and

chat GPT and stable diffusion and mid

journey and others to build your own

tool on top of it you can also explore

the chat GPT documentation to learn more

about how does it really work and what

you can do with it you can do the same

with other tools like hugging face like

Lang chain and others and that will give

you a perspective on how do you build

with these tools from scratch the next

step is for you to learn about Chad GPT

plugins and the GPT store now this is a

amazing opportunity it's like the App

Store from the 2008 in which you would

basically able to build and train your

own custom GPT you can launch it and

people would pay you to get access to

those gpts again you will have to

customize these gpts you will have to

learn the art of prompt engineering and

that is how you can build on top of chat

GPT on the GPT store so that is what you

need to learn if you want to become an

artificial intelligence engineer in

2024 there's a lot more to learn you are

just getting started description will

have a ton of resources that you can

take and get started in your journey of

becoming an AI engineer if you have any

questions let me know Below in the

comment section but this is how you do

it in 2024 learn about python then

python libraries then learn about

machine learning then learn about neural

networks and then finally learn about

NLP and then go into llms and start

building your own generative AI

applications thank you so much for

watching any questions let me know Below

in the comments if you're still watching

in the comment section I watch till the

very end you can also click a screenshot

and post this on social media and tag me

at isan Sharma 7390 I will catch you in

the next video

byebye

[Music]