AI crash course for beginners it is a best course for AI beginner

Contact : 7880113112 & Check Out Free Courses of Data Science & Data Analytics by The iScale
Free Course Link in Description box
Complete Crash Course
On Artificial Intelligence (AI)

Group - 1 Group - 2

Group - 1
Group - 2

Group - 1 Group - 2
AI
AI
AI

• AI Infrastructure & Model Creators
• Company that uses AI Infrastructure
• Types of AI Tools
• What is Artificial Intelligence ?
• Evolution of AI
• Discriminative Model (Classifier & Predictor)
• Generative Model (Content & Data Creation)
• Agentic Model (AI with Decision-Making Abilities)
• Hybrid Models (Combination of Multiple Approaches)
• Structure of AI
• Machine Learning
• Supervised ML
• Unsupervised ML
• Reinforcement ML
• Deep learning
• Neurons and neural network
• Face detection
• Computer vision
• Evolution of CV
• Natural language Processing
• Evolution of NLP
• Companies using NLP
• Case works
• Architechture
• LLM
• Building of LLM
• Agentic AI
• Features of AI agents
Topics to be covered today

AI Infrastructure & Model Creators
Open AI Google
DeepMind
Meta Anthropic Microsoft NVIDIA Amazon Tesla
Company that uses AI Infrastructure
Tech Companies Healthcare Finance & Stock Market E-commerce & Marketing

3 types of AI Tools
Standalone AI Tools Integrated AI Tools Customized AI Tools

Artificial Intelligence (AI) is a branch of computer science that focuses on creating machines that can
perform tasks that typically require human intelligence. These tasks include:
Learning – AI learns from data and improves its performance over time.
Reasoning – AI can analyze information and make logical decisions.
Problem-Solving – AI can find solutions to complex problems.
Understanding Language – AI can process and generate human language (like ChatGPT!).
Perception – AI can recognize images, sounds, and patterns.
What is Artificial Intelligence ?

Evolution of AI
1950s
Sabse pehle, AI ka concept Alan Turing ne introduce kiya tha. 1950 mein unhone ek paper likha tha,
"Computing Machinery and Intelligence," jisme unhone Turing Test introduce kiya. Is test ka goal tha
yeh dekhna ki kya ek machine soch sakti hai jaise insaan karta hai.

Evolution of AI
1956 (Dartmouth Conference):
Yeh moment AI ki duniya ka turning point tha, jab John McCarthy aur unke colleagues ne "Artificial
Intelligence" shabd ko define kiya aur officially is field ka shuruat ki. Yahaan se AI ka journey start hota
hai.

Evolution of AI
1960s to 1970s
1960s se lekar 1970s tak, AI mein bohot basic systems aur programs banaye gaye.
•ELIZA (1966): Yeh ek chatbot tha, jo Joseph Weizenbaum ne banaya tha. ELIZA ek simple script ke through logon se baat
kar sakti thi. Yeh ek early AI conversational system tha.
•Shakey the Robot (1969): Yeh ek robot tha, jise Stanford Research Institute ne develop kiya tha. Is robot mein decision-
making aur problem-solving capabilities thi.

Evolution of AI
The Winter of AI - 1970s to 1980s
AI ka initial excitement zyada din tak nahi chal paya. 1970s aur 1980s mein AI mein bohot funding aur
research kam ho gayi thi, isliye is period ko AI Winter kaha jata hai. Is time par logon ne socha tha ki AI
utna promising nahi hai, jitna initially laga tha. Us waqt hardware aur resources kaafi limited the.

Evolution of AI
1990s to Early 2000s
1990s ke end tak, AI ka second boom start hota hai, jab machine learning aur neural networks ka development start
hota hai.
Deep Blue vs Garry Kasparov (1997) : Yeh moment AI ki history mein bahut important tha, jab IBM ka Deep Blue chess
world champion Garry Kasparov ko harata hai. Is se AI ke potential ko duniya ne seriously lena shuru kiya.
Speech Recognition: 1990s mein, speech recognition systems bhi kaafi improve hue, jisme Dragon NaturallySpeaking
jese software aaye.

Evolution of AI
AI Revolution - 2010s
2010 ke baad, AI ka real revolution dekhne ko milta hai, jab
deep learning aur neural networks ka use improve hota hai.
In technologies ke through, machines ko image recognition,
speech recognition, natural language processing jaise
complex tasks perform karne ke liye train kiya gaya.
Deep Learning and Neural Networks:
Google’s DeepMind ne AlphaGo ko train kiya, jo 2016 mein
Go game ke world champion ko harata hai.
Chatbots and Personal Assistants:
2010s mein, AI-powered chatbots aur personal assistants
jese Siri, Alexa, Google Assistant market mein aaye, jo daily
tasks ko automate karte hain.

Evolution of AI
AI ka Future (2020s and Beyond)
Aaj ke time mein, AI bahut rapidly evolve ho raha hai, aur
generative AI (jese ChatGPT aur DALL·E), self-driving cars,
agentic AI, aur AI ethics jaise concepts ki taraf hum move kar
rahe hain.
•Generative AI: Jaise ki ChatGPT, DALL·E, MidJourney, jo new
content generate karte hain (text, images, etc.).
•Agentic AI: Jaise AutoGPT, jo independent tasks perform karne
mein capable hote hain.

Discriminative Model (Classifier & Predictor)
Discriminative models are used for classification and prediction.
Examples in AI:
•Spam detection (Spam or Not Spam)
•Face recognition (Is this face John’s or not?)
•Fraud detection in banking

Generative Model (Content & Data Creation)
Generative models create new data based on training data.
Examples in AI:
•ChatGPT, GPT-4, BERT (Text generation)
•Stable Diffusion, DALL·E (Image generation)
•WaveNet (Speech synthesis)

Website – www.theiscale.com Contact : 7880-113-112 theiscale theiscale.founders
The acceleration of AI research,
envisioning businesses utilizing AI
agents for customer interactions
agentic AI as a "new labor
model, new productivity model,
and a new economic model
The age of agentic AI is here,"
emphasizing the emergence of AI
agents capable of performing
complex tasks autonomously
Nadella introduced AI agent tools designed to
act autonomously on behalf of users, capable
of tasks like reviewing customer returns and
checking supply-chain invoices

Agentic Model (AI with Decision-Making Abilities)
Agentic models are AI systems that can take actions and make decisions autonomously. These models go beyond
classification and generation—they interact with the environment and take actions accordingly.
Examples in AI:
•Self-driving cars (Deciding when to stop, turn, accelerate)
•AI-powered robots (Automating warehouse operations)
•Game-playing AI (AlphaGo, OpenAI Five for Dota 2)
•Personal AI assistants (AutoGPT, BabyAGI)

Hybrid Models (Combination of Multiple Approaches)
Some AI systems use a combination of discriminative, generative, and agentic
models for better performance.
Examples:
Self-driving cars (Use CNNs for image recognition + RL for decision-making)
Chatbots with memory (Use transformers for text generation + RL for adaptive
learning)
AI art generators (Use GANs for image generation + CNNs for style transfer)

Structure of AI
Artificial Intelligence
Machine Learning

Structure of AI
Machine Learning
Deep Learning

Structure of AI
Machine Learning
Deep Learning
Discriminative

Structure of AI
Machine Learning
Deep Learning
Discriminative
Generative AI
LLM
Agentic AI

Machine Learning (ML) is a type of technology that allows computers to learn from data and make decisions without
being directly programmed.
Machine Learning
Imagine teaching a child to recognize
fruits. If you show them many apples
and tell them, "This is an apple," they
will eventually learn to identify apples
on their own. Similarly, in ML, we
provide a computer with a lot of data,
and it learns patterns to make
predictions or decisions.
Give
Data
Learn Patterns Make Predictions
Improvise with
time

Trained Model
Machine Learning
Training Data
New Data
Model
Predictions

Types of Machine Learning
Supervised Learning Unsupervised Learning Reinforcement Learning
Examples:
•Spam detection (Email is spam or not)
•House price prediction
•Image classification (Cats vs. Dogs)
Examples:
•Customer segmentation in marketing
•Anomaly detection (fraud detection banking)
•Topic modeling in NLP
Examples:
•Self-driving cars
•Game-playing AI (e.g., AlphaGo)
•Robotics

Supervised Machine Learning
Supervised Learning ek aisa Machine Learning type hai jisme hum model ko labeled data (yaani ki input-output pairs)
ke saath train karte hain. Matlab, model ko pehle se pata hota hai ki kaunsa input kis output se match karta hai. Phir
model ye pattern samajhne ki koshish karta hai taaki naye data ke liye bhi sahi prediction kar sake.

Supervised Machine Learning
Some Algorithms
• Linear Regression
• Logistic Regression
• Decision Trees
• Random Forest
• Support Vector Machines (SVM)
• K-Nearest Neighbors (KNN)
• Naive Bayes
• Gradient Boosting (e.g., XGBoost, LightGBM)
• AdaBoost

Unsupervised Learning ek aisa Machine Learning type hai jisme model ko bina kisi labeled data ke train
kiya jata hai. Matlab, model ko input diya jata hai, lekin output ka pata nahi hota. Model khud patterns
aur relationships find karta hai data ke andar.
Unsupervised Learning

Unsupervised Learning
Some Algorithms
• K-Means Clustering
• Hierarchical Clustering
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• Principal Component Analysis (PCA)
• Independent Component Analysis (ICA)
• t-Distributed Stochastic Neighbor Embedding (t-SNE)
• Gaussian Mixture Models (GMM)
• Autoencoders

Reinforcement Learning
Reinforcement Learning (RL) ek Machine Learning ka type hai jisme ek agent apne environment se
interact karta hai aur reward system ke through seekhta hai.

Reinforcement Learning
Reinforcement Learning (RL) ek Machine Learning ka type hai jisme ek agent apne environment se
interact karta hai aur reward system ke through seekhta hai.
Q-Learning
Deep Q-Networks (DQN)
SARSA (State-Action-Reward-State-Action)
Policy Gradient Methods

Deep Learning is a type of Artificial Intelligence (AI) that helps computers learn and make decisions
just like humans.
Deep Learning
brain-inspired system

Dendrites: Ye part
information ko receive
karta hai dusre neurons
se.
Cell Body (Soma): Ye
part received
information ko
process karta hai.
Axon: Jab neuron process
kar leta hai, toh output
signal ko phir dusre neurons
tak bhejne ka kaam axon
karta hai.
Hidden Layers: Ye layers multiple neurons se banayi jaati hain jo input data ko process
karte hain. Har neuron apne weights aur activation function ke through input ko
transform karta hai.

Nose
Ears
Eyes

Nose
Ears
Eyes
This is Panda
Yes

Nose
Ears
Eyes
This is not a Panda
Yes

Nose
Ears
Eyes
This is a Panda Yes

model
Recognize
panda face

Difference Between Deep Learning & Machine Learning
Image Recognition
•Machine Learning:
• Uses algorithms like SVM, Random
Forest.
• Needs manual feature extraction (edges,
colors, textures).
•Deep Learning:
• Uses CNN (Convolutional Neural
Networks).
• Learns patterns automatically and
performs better on complex images.

Important topics of Deep learning
1. Basics of Deep Learning
•What is Deep Learning?
•Difference between Machine Learning and Deep Learning
•Neural Networks and their working
•Activation Functions (ReLU, Sigmoid, Tanh, Softmax)
2. Neural Networks
•Perceptron and Multi-Layer Perceptron (MLP)
•Forward and Backpropagation
•Gradient Descent and Optimizers (SGD, Adam, RMSprop)
•Loss Functions (Cross-Entropy, MSE, Huber Loss)
3. Deep Learning Frameworks
•TensorFlow
•PyTorch
•Keras
4. Convolutional Neural Networks (CNNs) – For Image
Processing
•Filters and Kernels
•Feature Maps and Pooling Layers
•CNN Architectures (AlexNet, VGG, ResNet,
EfficientNet)
•Transfer Learning

Important topics of Deep learning
5. Recurrent Neural Networks (RNNs) – For Sequential Data
•RNN Basics and Challenges (Vanishing Gradient Problem)
•Long Short-Term Memory (LSTM)
•Gated Recurrent Units (GRU)
•Applications in NLP and Time-Series Forecasting
6. Natural Language Processing (NLP) with Deep Learning
•Word Embeddings (Word2Vec, GloVe, FastText)
•Transformers (BERT, GPT)
•Attention Mechanism and Seq2Seq Models
•Sentiment Analysis, Chatbots, and Speech Recognition
7. Generative Models
•Autoencoders
•Generative Adversarial Networks (GANs)
8. Advanced Topics
•Self-Supervised and Semi-Supervised Learning
•Deep Reinforcement Learning
•Meta-Learning
•Explainable AI (XAI)
9.AI Deployment & Optimization
•Model Deployment (Flask, FastAPI, TensorFlow Serving)
•Model Optimization (Pruning, Quantization)
•AI on Edge Devices (TensorFlow Lite, ONNX)

Natural Language Processing (NLP)
Natural Language Processing (NLP) is a branch of artificial intelligence (AI) that enables computers to
understand, interpret, and generate human language. It allows machines to process and analyze vast
amounts of natural language data, such as text and speech.

NLP has evolved over the years from simple rule-based methods to
advanced deep learning techniques:
1. Rule-Based Systems (1950s-1980s): Early NLP relied on hand-
crafted rules and dictionaries.
2. Statistical Methods (1990s-2010s): Machine learning models like
Hidden Markov Models (HMMs) and Support Vector Machines
(SVMs) improved NLP accuracy.
3. Deep Learning & Transformers (2015-Present): Models like
Word2Vec, LSTMs, and Transformers (BERT, GPT) revolutionized NLP
by achieving human-like language understanding.
Evolution of NLP

Computer Vision
Computer Vision (CV) is a field of Artificial Intelligence (AI) that enables machines to interpret and
understand visual data from images or videos, just like humans. It allows computers to recognize patterns,
detect objects, and analyze scenes.

Computer Vision
Evolution of Computer Vision
The field has evolved significantly over time:
1.1960s-1980s: Early computer vision systems relied on basic image processing and edge
detection.
2.1990s-2010s: Introduction of Machine Learning (ML) techniques like Support Vector Machines
(SVMs) and Convolutional Neural Networks (CNNs).
3. 2012-Present: Deep learning revolutionized CV, especially with AlexNet in 2012, followed by
ResNet, YOLO, and Transformers (ViTs).

LLM (Large Language Model)
LLM (Large Language Model) ek AI model hai jo bahut bade text datasets se train kiya jata hai aur
human-like text generate kar sakta hai
Example:
•ChatGPT, Google Gemini, Claude, Llama
•AI Chatbots (Siri, Alexa, Bard)
•Code generation (GitHub Copilot)

1) Data (Bahut Sara High-Quality Data)
•LLM ko train karne ke liye bahut saare text datasets chahiye.
•Example: Common Crawl, OpenWebText, Wikipedia, Code datasets
2) Neural Network Architecture (Transformer Model)
•LLMs Transformer-based models ka use karte hain, jaise:
•GPT (Generative Pre-trained Transformer)
•BERT (Bidirectional Encoder Representations from Transformers)
•T5 (Text-to-Text Transfer Transformer)
What is needed to build an LLM ?

3) Powerful Hardware (GPUs & TPUs)
•LLM ko train karne ke liye bahut powerful GPUs ya TPUs ki zaroorat hoti hai.
•Example: NVIDIA A100, H100 GPUs ya Google TPUs
4) Training Algorithms
•LLMs ko unsupervised learning aur self-supervised learning se train kiya jata hai.
•Techniques:
• Masked Language Modeling (MLM) – BERT ke liye
• Causal Language Modeling (CLM) – GPT ke liye
• Reinforcement Learning from Human Feedback (RLHF) – LLM ko better responses dene ke liye fine-
tune karta hai.

5. Optimization Techniques
•Gradient Descent aur Backpropagation se models optimize kiye jate hain.
•Fine-tuning & Transfer Learning se LLMs ko specific tasks ke liye improve kiya jata hai.
6. Deployment & Scaling
•LLMs ko APIs aur cloud platforms par deploy kiya jata hai.
•Example: OpenAI API, Hugging Face, Google Cloud AI, AWS, Azure AI

Connect with us
Website – www.theiscale.com Contact : 7880-113-112 theiscale theiscale.founders
theiscale
theiscale.founders
Instagram Handle
Siblings - Nishant Dhote & Swati Dhote
The iScale Organization Handle

AI crash course for beginners it is a best course for AI beginner

More Related Content

Similar to AI crash course for beginners it is a best course for AI beginner (20)

More from RajKumar64326 (6)

Recently uploaded (20)

AI crash course for beginners it is a best course for AI beginner