AIML-20CS51I

WEEK-1

AI CONCEPTS AND FUNDAMENTALS

AI in Industry, Evolution of AI & HMI (Human-Machine Interface)

1. AI in Amazon

• Uses AI for:
o Product recommendations (based on what & why people search)
o Predicting customer demand
o Running Amazon Go – cashier-less stores
• AI personalizes shopping experience
• Helps Amazon understand customer intent, not just keywords

YouTube:

• Amazon Go
• Alexa & AI in Amazon

2. AI in Microsoft

• Integrated AI into:
o Cortana (voice assistant)
o Office 365 (writing help, scheduling, Excel insights)
o Bing, Power BI, MyAnalytics, Pix camera
• Cloud AI through Azure AI, Cognitive Services, and Azure ML Studio
• Key initiatives:
o AI for Earth
o AI for Health
o AI for Accessibility
o AI for Humanitarian Action
o AI for Cultural Heritage

Videos:

• AI for Earth
• Cortana AI
• Azure AI

3. AI in Google

• AI improves:
o Search (Spelling correction, Passage ranking)
o YouTube (Key moments in videos)
o Google Lens (Search using images)
o Google Assistant
o Natural Language Processing and Search Ranking

1
 AIML-20CS51I

Google AI Developments

4. AI in IBM

• AI used for:
o Computer vision
o Speech-to-text
o Natural Language Processing
o Robotics & Automation
o IBM Watson – used in healthcare, finance, customer service

IBM Watson Overview

Latest AI Developments:

• Natural Language Generation (NLG): Turns data into human-like text

• Natural Language Understanding (NLU): Helps machines understand meaning
• Speech Recognition: Converts voice to text
• Machine Learning (ML): Trains machines using data
• Virtual Agents: Chatbots & digital assistants
• Expert Systems: Systems that make decisions like humans
• Robotic Process Automation (RPA): Automates repetitive business tasks

Evolution of AI & HMI:

• AI Evolution:
o From rule-based systems → Machine Learning → Deep Learning
• HMI (Human-Machine Interface):
o Evolved from keyboards & screens → Voice & gesture controls
o Goal: Make tech more intuitive for humans

Watch:

• Evolution of AI
• Evolution of HMI

2
 AIML-20CS51I

1. What is Artificial Intelligence?

• Artificial Intelligence is the ability of machines to mimic human intelligence like

learning, reasoning, and problem-solving.
• AI systems can act without being explicitly programmed and can improve through
experience.

2. How Does AI Work?

• AI works by combining various technologies like Machine Learning, Deep Learning,

and Natural Language Processing.
• It learns patterns from data, draws conclusions, and makes predictions or decisions.

3. Key Subfields of AI

• Machine Learning (ML): Enables machines to learn from past data and improve over
time without manual programming.
• Deep Learning: Uses artificial neural networks to analyse complex data like images,
speech, or text.
• Neural Networks: Mimic the human brain to identify patterns and relationships in data.
• Natural Language Processing (NLP): Helps machines understand, interpret, and
respond in human language.
• Computer Vision: Allows machines to “see” and understand images or videos by
identifying features.
• Cognitive Computing: Tries to simulate human thought processes by analysing
speech, text, and images.

4. Purpose of AI

• The goal of AI is to support or enhance human decision-making, reduce errors, and

automate complex tasks.
• It simplifies processes and allows businesses to use data more efficiently.

5. Types of AI

3
 AIML-20CS51I

Based on Capability

• Artificial Narrow Intelligence (ANI): AI that performs a specific task (e.g., Alexa,
chatbots).
• Artificial General Intelligence (AGI): Hypothetical AI with human-like reasoning
across many domains.
• Artificial Super Intelligence (ASI): Future AI that surpasses human intelligence in all
fields.

Based on Functionality

• Reactive Machines: Respond to inputs with no memory (e.g., chess-playing AI).

• Limited Memory: Learns from past data to make decisions (e.g., self-driving cars).
• Theory of Mind: AI that understands emotions and intentions (not yet developed).
• Self-Aware AI: Hypothetical AI with consciousness and self-awareness.

Advantages of Artificial Intelligence:

• Reduction in human error

• Available 24×7
• Helps in repetitive work
• Digital assistance
• Faster decisions
• Rational Decision Maker
• Medical applications
• Improves Security
• Efficient Communication

History of Artificial Intelligence (AI):

4
 AIML-20CS51I

• 1943 – McCulloch & Pitts proposed the first artificial neuron model.
• 1949 – Donald Hebb introduced a theory on strengthening neural connections.
• 1950 – Alan Turing proposed the famous Turing Test.
• 1950 – Minsky & Edmonds built SNARC, the first neural network computer.
• 1956 – The term Artificial Intelligence was coined by John McCarthy.
• 1956 – Newell & Simon created the first AI program, Logic Theorist.
• 1959 – Arthur Samuel coined the term Machine Learning at IBM.
• 1963 – John McCarthy started the first AI Lab at Stanford.
• 1966 – Joseph Weizenbaum developed ELIZA, the first chatbot.
• 1972 – Japan built WABOT-1, the first humanoid robot.
• 1974 –1980 – First AI Winter due to lack of funding.
• 1980 – R1, the first commercial expert system, revived interest in AI.
• 1980 – First AAAI Conference on Artificial Intelligence held at Stanford.
• 1987–1993 – Second AI Winter occurred as funding declined again.
• 1997 – IBM’s Deep Blue defeated world chess champion Garry Kasparov.
• 2002 – AI entered homes with the Roomba vacuum cleaner.
• 2005 – US military adopted autonomous robots like Big Dog.
• 2006 – Major tech companies like Google, Facebook, Netflix started using AI.
• 2008 – Google introduced speech recognition in the iPhone app.
• 2011 – IBM’s Watson won Jeopardy by understanding natural language.
• 2012 – Google’s deep learning project recognized cats using YouTube videos.
• 2014 – Google’s self-driving car passed the driving test.
• 2014 – Amazon launched Alexa, a smart voice assistant.
• 2016 – Hanson Robotics unveiled Sophia, the first robot citizen.
• 2020 – OpenAI released GPT-3, a groundbreaking large language model.
• 2021 – DeepMind's AlphaFold 2 achieved a breakthrough in protein folding.
• 2022 – OpenAI launched ChatGPT, gaining widespread public attention for
conversational AI.
• 2023 – OpenAI released GPT-4, a more advanced multimodal language model.
• 2024 – The European Parliament approved the AI Act, the first comprehensive AI legal
framework.
• 2025 – Ongoing integration of AI in productivity tools and increased focus on
responsible AI.

Goals of Artificial Intelligence:

• Build systems that can learn from data and improve over time.
• Enable machines to reason and solve problems logically.
• Allow AI to perceive the environment through sensors and data.
• Create AI that can adapt to new or changing situations.
• Automate tasks to reduce human effort and save time.
• Develop systems that can communicate naturally with humans.
• Enhance productivity and efficiency across industries.
• Assist in decision-making with data-driven insights.
• Simulate human intelligence and behaviour for advanced tasks.
• Improve quality of life through smart applications and services.

5
 AIML-20CS51I

Applications of AI:

a. In Daily Life

• Google Search: Predicts your search based on previous queries.

• Netflix: Recommends shows based on your watch history.
• Facebook: Suggests photo tags using facial recognition.
• YouTube: Auto-plays related videos and recommends content.
• Smart Assistants: Siri, Alexa help with reminders, weather, and more.
• Google Maps: Provides real-time route suggestions using traffic AI.

b. In Business & Industry

• Healthcare: AI diagnoses diseases, analyses scans, and powers robotic surgeries.

• E-commerce: AI personalizes shopping, detects fraud, and powers chatbots.
• HR (Human Resources): Automates resume filtering and tracks employee sentiment.
• Robotics: Powers factory automation, warehouse sorting, and delivery bots.
• Finance: AI helps in credit scoring, fraud detection, and stock trading.
• Education: Offers personalized learning, grading automation, and virtual tutors.

Ethics in AI
AI ethics is about building fair, transparent, and responsible AI systems. It aims to reduce bias,
respect privacy, and ensure that AI benefits all of society.

Ethical Principles (Belmont Report):

• Respect for Persons: Protect individual rights and consent.

• Beneficence: Ensure AI does good and avoids harm.
• Justice: Make AI fair and unbiased in its decisions.

Ethical Concerns in AI:

• Bias and Discrimination: AI may unintentionally reflect societal biases.

• Job Displacement: AI may automate jobs, requiring new skill transitions.
• Privacy Issues: AI must handle data responsibly and securely.
• Accountability: Who is responsible for AI’s actions or errors?

6
 AIML-20CS51I

Examples of AI in the Real World:

• Siri, Alexa – Voice assistants that respond to commands.

• Tesla Autopilot – Uses AI to control and drive cars autonomously.
• Chatbots – Used by websites to answer FAQs automatically.
• Netflix Recommendations – Suggests shows based on past viewing habits.
• Google Translate – AI that translates languages instantly.

Significance of Data in AI:

Data is the foundation of Artificial Intelligence. Unlike traditional software that follows
hardcoded instructions, AI systems learn from data to improve their performance.

Why data is critical in AI:

• Learning Patterns: AI uses data to detect trends, relationships, or behaviors (e.g.,

predicting customer behavior from purchase history).
• Training Models: In machine learning (ML), algorithms require large datasets to learn
how to perform tasks (e.g., image recognition).
• Accuracy and Reliability: More and better-quality data means more accurate AI
predictions.
• Feedback Loop: AI systems improve over time as they process new data and receive
feedback.

"Garbage in, garbage out" principle:

If the input data is:

• Incomplete, inaccurate, or biased, then

• The AI model's output will be flawed, misleading, or even dangerous.

Good dataset characteristics:

• Complete: Covers all relevant aspects.

• Consistent: Free of contradictions.
• Valid: Up-to-date and relevant to the problem.
• Accurate: Error-free.
• Unique: No duplicates or noise.

AI Software Development Life Cycle (AI-SDLC):

AI software follows a similar development life cycle as traditional software, but with a
stronger emphasis on data and model performance.

Phases of AI-SDLC:

1. Planning
o Define the AI project’s goals and use-cases.
o Decide whether AI is the right solution.

7
 AIML-20CS51I

2. Requirement Analysis
o Identify data requirements (type, format, volume).
o Evaluate budget, tools, performance metrics.
3. Design
o Choose algorithms (ML, DL, NLP).
o Design the data flow, system architecture, and model pipeline.
4. Development
o Implement models and data preprocessing.
o Use libraries like TensorFlow, PyTorch, Scikit-learn.
5. Testing
o Test model accuracy, bias, fairness.
o Validate using unseen (test) datasets.
6. Deployment
o Integrate model into real-world environment (e.g., a website, mobile app).
o Monitor performance and real-time results.
7. Maintenance
o Regularly update the model as new data arrives.
o Retrain if performance drops.

Comparison: Traditional Software vs AI Software:

Aspect Traditional Software AI Software
Development Rule-based programming with Data-driven with learning from patterns and
Approach fixed logic examples
Behaviour Deterministic – same input gives Probabilistic – output can vary based on
same output learned patterns
Coding Method Explicit rules written by Algorithms learn rules from data (e.g.,
developers machine learning)
Data Dependency Works with minimal data; data Requires large datasets for training and
used for processing accuracy
Flexibility Inflexible to new/unseen scenarios Can generalize and adapt to new/unseen data

Examples Calculator, Database System, Face Recognition, ChatGPT, Fraud

Inventory App Detection
Testing & Debugging Straightforward; bugs are usually Complex; performance measured statistically
traceable (e.g., accuracy)
Performance Tuning Optimized through better Improved through better models and more
algorithms and code data
Output Interpretation Transparent and explainable Often seen as a "black box"; hard to explain
decisions
Learning Capability Does not learn or improve on its Learns and improves from data over time
own
Error Handling Errors handled via code logic Errors minimized via training, validation,
and tuning

Example:
• Traditional software: "If temperature > 38°C, then raise an alert." (uses predefined rules)
• AI software: Learns from 100,000 patient records to decide when to raise an alert — even if
symptoms are mixed and there's no clear temperature threshold.

8
 AIML-20CS51I

Example – Game Rules (Chess) in AI

Chess is a classic AI problem because:

• It involves strategic decision-making.

• There are many possible moves at each step.
• It can be framed as a search problem (e.g., "What’s the best next move?").

How AI plays Chess:

A. Traditional Programming Approach

• Hardcode rules: moves for each piece, winning conditions.

• Use minimax algorithm: simulate all possible future moves and choose the one with
the best guaranteed outcome.
• Add alpha-beta pruning: skip obviously bad moves to save time.

B. AI/ML-based Approach

• Train AI on millions of professional games.

• The system learns patterns, openings, and strategies used by top players.
• Uses deep reinforcement learning (e.g., AlphaZero by DeepMind):
o It starts with no knowledge.
o Plays games against itself.
o Learns from wins/losses.
o Eventually outperforms traditional engines like Stockfish.

Example:

• Deep Blue (1997): Rule-based + brute-force AI beat Kasparov.

• AlphaZero (2017): Learns from scratch, plays creatively, and defeated top engines.

Why Do We Need a Version Control System (VCS)?

Imagine building a project like a mobile app, website, or research paper. Without a version
control system, keeping track of updates becomes confusing and risky. Here's why VCS is
essential:

1. Tracks Changes and History

A VCS logs every change to your files — who made it, what was changed, and when. This
helps you understand how the project evolved and why decisions were made.

2. Enables Collaboration

Multiple people can work on the same project without conflict. Each person works on their
own version, and VCS merges their work efficiently.

9
 AIML-20CS51I

3. Easy Reverts and Recovery

If something goes wrong, you can roll back to a previous version. This is especially helpful
when a new change introduces bugs.

4. Safe Experimentation with Branching

You can create branches to test new features without touching the main project. Once it works,
merge it back — no risk to the original version.

5. Backup and Disaster Recovery

The full history is saved in remote repositories (like GitHub). Even if your laptop crashes, your
code is safe.

6. Accountability and Clarity

Each change is linked to an author and a commit message. This helps track responsibility and
debug issues faster.

Fundamentals of Git:

Git is the most popular Distributed Version Control System (DVCS). Unlike centralized
systems, Git gives each developer a full copy of the project and its history.

Why Git Stands Out

• Offline Work: You can commit, view logs, and branch without an internet connection.
• Speed: Git works fast because most operations happen locally.
• Redundancy: Every clone is a backup — no central failure risk.

Key Git Concepts:

Concept Description
Repository (Repo) A folder tracked by Git that contains your project and its history.
Commit A snapshot of your project. It includes changes, a message, author,
and time.
Branch A line of development. You can make changes without affecting
the main version.
Master/Main The default main branch of most Git projects.
HEAD A pointer to the latest commit on your current branch.
Working Directory The actual files on your system that you're editing.
Staging Area (Index) A temporary area where you prepare changes before committing
them.

Git is a powerful tool that gives you control, safety, and freedom when developing any
project. With practice, it becomes an essential part of modern development workflows.

10
 AIML-20CS51I

Git Installation and Setup:

---Installation by Platform

Windows:

• Visit: https://git-scm.com/download/win
• Download and run the installer.
• Follow prompts (default settings are usually fine).

macOS:

• Recommended: brew install git (requires Homebrew).

• Or: xcode-select --install to get Git with Xcode CLI tools.
• Or: Download from https://git-scm.com/download/mac

Linux:

• Debian/Ubuntu:

sudo apt-get update

sudo apt-get install git

• Fedora:

sudo dnf install git

First-Time Setup (After Installation):

Set your username and email — this info is attached to your commits:

git config --global user.name "Your Name"

git config --global user.email "your.email@example.com"

Optional setup:

• Set default branch name to main:

git config --global init.defaultBranch main

• Set preferred text editor:

git config --global core.editor "code --wait" # for VS Code

git config --global core.editor "nano" # for Nano

11
 AIML-20CS51I

Basic Local Git Operations:

These operations happen locally, without needing an internet connection.

1. Creating a Repository

Initialize Git tracking in a new/existing project folder:

cd path/to/your/project

git init

This creates a hidden.git folder to store all Git data.

2. Cloning a Repository

To download a remote repository (e.g., from GitHub):

git clone https://github.com/username/repo-name.git

This creates a local copy with full history.

3. Making and Recording Changes

Modify files in any editor (VS Code, Atom, etc.).

Check what’s changed:

git status

It shows:

• Modified files
• Untracked files
• Staged changes

4. Staging and Committing Changes

Staging:

git add <file_name> # Stage one file

git add. # Stage everything

Committing:

git commit -m "Short and clear commit message"

Example: git commit -m "Add contact form to footer"

12
 AIML-20CS51I

5. Viewing History

• Show full commit history:

git log

• One-line summary per commit:

git log --oneline

• Graph view of branches:

git log --graph --oneline --all

• View specific commit changes:

git show <commit_hash>

6. Undoing Changes

• Discard unstaged changes:

git restore <file_name> # WARNING: Deletes unsaved work

• Unstage a file (remove from staging area):

git restore --staged <file_name>

• Revert a commit safely (keeps history):

git revert <commit_hash>

>Reset to a previous commit (rewrites history):

• Keep changes staged:

git reset --soft <commit_hash>

• Keep changes unstaged:

git reset --mixed <commit_hash>

• Discard all changes permanently:

git reset --hard <commit_hash>

13
 AIML-20CS51I

Git Branching and Merging (Basic):

Branching in Git lets you work on new features or fixes without affecting the main code.
Merging helps you bring those changes back when ready.

1. Creating and Switching to New Branches

To create a new branch: git branch new-feature

To create and immediately switch to it: git checkout -b new-feature

Example: git checkout -b login-page

This creates a branch called login-page and switches to it.

2. Switching Between Branches

To switch to another branch: git checkout branch-name

Example: git checkout main

This moves you back to the main branch.

3. Merging Local Branches Together

Once your work on a feature branch is done, merge it into the main branch:

1. Switch to the branch you want to merge into (usually main): git checkout main
2. Merge your feature branch: git merge new-feature

Example: git merge login-page

This brings all commits from login-page into main.

Fast-Forward vs. Merge Commit:

• If there's no new commit in main, Git performs a fast-forward merge (simply moves
the pointer).
• To force a merge commit, use: git merge --no-ff new-feature

Summary of Commands

Action Command
Create new branch git branch feature-x
Create + switch git checkout -b feature-x
Switch branches git checkout main
Merge feature into current git merge feature-x
Force merge commit git merge --no-ff feature-x

14
 AIML-20CS51I

GitHub – Concepts and Commands:

1. Basics of Distributed Git

• Git is a distributed version control system (DVCS).

• This means every developer has a full copy of the repository, including all versions
and history.
• You can work offline and sync with others using push (upload) and pull (download).
• No central server dependency – everyone’s clone is a backup.

GitHub is a cloud-based hosting platform for Git repositories — it adds tools for
collaboration, sharing, and management.

2. GitHub Account Creation and Configuration

To use GitHub:

1. Go to https://github.com/join
2. Enter your username, email, and password
3. Verify your account via email
4. (Optional) Set up SSH keys for secure authentication:

ssh-keygen -t ed25519 -C "your.email@example.com"

3. Create and Push to Repositories

A. On GitHub (create repo):

1. Click the "+" icon → "New repository"

2. Enter repo name and description
3. Choose public/private
4. (Optionally) initialize with README
5. Click "Create repository"

B. In your local system (push your project):

git init

git remote add origin https://github.com/your-username/repo-name.git

git add .

git commit -m "Initial commit"

git push -u origin main

This uploads your local code to GitHub.

15
 AIML-20CS51I

4. Versioning

• GitHub keeps the entire version history of your project.

• Each commit = one version of the project with details like author, date, message.
• You can:
o Use git log to view versions
o Revert to any old version
o Compare changes between versions

Versioning helps with tracking progress, debugging, and safe rollbacks.

5. Collaboration on GitHub

GitHub makes teamwork easy through:

Feature Purpose
Branches Work on features without affecting main code
Pull Requests Propose and review changes before merging
Issues Report bugs or suggest features
Forks Copy someone's repo to your own account
Code Reviews Review and approve others' code

Team members can comment on code, suggest improvements, and merge once approved.

6. Migration

Migration = moving repos between users or platforms.

Examples:

• Move from one GitHub account to another

• Migrate from GitHub to GitLab or vice versa
• Transfer ownership to an organization

To transfer a repo on GitHub:

1. Go to the repo → Settings

2. Scroll to "Danger Zone"
3. Click "Transfer"
4. Enter the new owner’s name and confirm

All collaborators, issues, and history are preserved.

16
 AIML-20CS51I

Popular AI Cloud Services by Leading Vendors:

1. Amazon Web Services (AWS):

Overview:
AWS is a pioneer in cloud computing and offers a broad range of AI services for both
consumers and businesses.

Key AI/ML Offerings:

• Amazon Lex: Builds conversational interfaces (e.g., chatbots) using the same
technology as Alexa.
• Amazon Polly: Converts text to natural-sounding speech (TTS).
• Amazon Recognition: Performs real-time image and video analysis (facial recognition,
object detection).
• Amazon SageMaker: A complete ML platform to build, train, and deploy models at
scale.
• Amazon Machine Learning (AML): Simplifies the process of building ML models
with visualization tools.
• Alexa Voice Services (AVS): Integrates voice AI into consumer products.

Specialty:
Focus on voice, vision, and user-friendly ML model creation without deep ML knowledge.

2. Google Cloud:

Overview:
Google leads in AI research and tools, offering powerful infrastructure and APIs for developers
at all levels.

Key AI/ML Offerings:

• TensorFlow: Industry-standard open-source deep learning library.

17
 AIML-20CS51I

• Vertex AI (formerly AI Platform): Unified ML development environment for training

and deploying models.
• AutoML: Enables non-experts to train high-quality models with minimal effort.
• Cloud Vision API: Detects objects, faces, text, and landmarks in images.
• Cloud Natural Language API: Analyzes and extracts meaning from text.
• TPUs (Tensor Processing Units): Google’s custom chips for high-performance ML.

Specialty:
Advanced AI infrastructure, research-backed tools, and integration with Google’s own
technologies (Search, Translate, etc.).

3. IBM Cloud (Watson AI):

Overview:
IBM is known for enterprise-grade AI solutions, especially under the Watson brand.

Key AI/ML Offerings:

• IBM Watson Studio: Tools for building, training, and deploying AI models with
automated data preparation.
• Watson Assistant: Builds intelligent chatbots and virtual agents.
• Watson Discovery: Extracts insights from complex documents.
• Watson Machine Learning: Trains and deploys models on cloud or on-premise.
• Watson Services for Core ML: Enables app integration with Apple’s Core ML
framework.

Specialty:
Strong in enterprise AI, hybrid deployment (cloud/on-prem), and data-rich AI model building.

4. Microsoft Azure:

Overview:
Azure provides an extensive set of modular AI capabilities across services, tools, and
infrastructure.

18
 AIML-20CS51I

Key AI/ML Offerings:

• Azure Cognitive Services: Pre-trained models for vision, speech, language, and
decision-making.
• Azure Machine Learning (AML): Comprehensive ML platform with model training,
deployment, and MLOps.
• Azure Notebooks & Visual Studio Tools for AI: Developer tools for building and
training models.
• AI Infrastructure: Includes Azure Data Lake, Kubernetes Service, and high-
performance VMs.

Specialty:
Flexible ML model development for both beginner and enterprise users, deeply integrated with
Microsoft ecosystem (Office 365, Teams, etc.).

5. Salesforce (Einstein AI):

Overview:
Salesforce offers embedded AI through its platform "Einstein," primarily focused on CRM and
business intelligence.

Key AI/ML Offerings:

• Einstein Prediction Builder: Predict outcomes using custom ML models.

• Einstein Bots: Create smart customer service chatbots.
• Einstein Language: Classify intent and sentiment from text.
• Einstein Vision: Image recognition integrated into Salesforce apps.
• Einstein Recommendation Builder: Suggest personalized products or content.

Specialty:
AI built natively into CRM — ideal for sales, marketing, and customer relationship
enhancement.

19
 AIML-20CS51I

Questions:

1. Define Artificial Intelligence. Explain the types of AI with examples.

2. How does AI works?
3. What are the goals and applications of AI?
4. Describe how AI influences companies like Amazon, Google, and IBM.
5. Discuss how AI impacts daily life, jobs, products, and services.
6. Why is data considered the backbone of AI?
7. Explain the AI Software Development Life Cycle.
8. Compare traditional software development vs AI software development.
9. List and explain any 3 cloud services for AI/ML/DL and their providers.
10. What are the major ethical concerns in AI?
11. Give 2 real-world examples of AI applications with brief explanation.
12. Why do we need a Version Control System?
13. List and explain 5 basic Git operations.
14. What is GitHub? How does it help in team collaboration?
15. Explain Git branching and merging with simple steps.

20