Introduction: Artificial Intelligence for Everyone

Definition of Artificial intelligence (AI)

Ability of a machine to learn patterns and make predictions.

• Computer systems designed to perform tasks that typically require human intelligence

Tasks AI can do

• Understand Language

• Recognize Images

• Make Predictions

• Play Games

• Drive Cars

What is not AI

• Traditional Rule-Based Systems

• Simple Automation Tools

• Mechanical Devices

• Basic Sensors

Evolution of AI: 1950

• "Computing Machinery and Intelligence" - Alan Turing’s paper

• Turing test/”Imitation Test"

Evolution of AI: 1956

• Dartmouth Conference - John McCarthy; birthplace of AI as a field

• "Artificial Intelligence" - coined by John McCarthy

Evolution of AI: 1960s - 1970s

• Progress in AI research

  • Expert systems

  • Early neural networks

  • Exploration of symbolic reasoning & problem solving 

Evolution of AI: 1980s - 1990s

• Mixed optimism and skepticism

• Led to "AI winter"

Evolution of AI: 2000s onwards

• Resurgence of interest & progress

• Advancements in

  • computing power

  • data availability

  • algorithmic innovation

Narrow AI

• Single tasks

  • predicting purchases

  • Planning schedules

• consumer application

  • Siri

• Can handle specific tasks effectively; lacks broader understanding.

Broad AI

• Midpoint 

• More versatile; handles a wider range of related tasks

• Businesses 

General AI

• Can perform tasks at the same intellectual level as humans

• Artificial Superintelligence (ASI), self aware machines

  • Aren’t there yet

Data Science

• Data inputs

  • numerical

  • alphabetical

  • alphanumeric

• collection, analysis, and interpretation of large volumes of data; extract insights

• Using

  • statistical methods

  • machine learning algorithms

  • data visualization

Natural Language Processing (NLP)

• Processes text and speech inputs; enable computers to understand, interpret, and generate human language.

• Tasks

  • language translation

  • text summarization

  • speech recognition

Computer Vision

• Deals with visual data inputs; enables computers to interpret and understand visual information.

• Tasks

  • object detection

  • image classification

  • facial recognition

• Applications

  • autonomous vehicles

  • medical imaging

  • augmented reality

Structured Data

• Organized

• rows and columns

• names, dates, addresses, and stock prices

• Straightforward to analyze and manipulate

Unstructured data

• Lacks specific organization

• Examples

  • images

  • text documents

  • customer comments

• Extraction requires specialized tools and techniques.

Semi-structured data

• Metadata to identify certain characteristics and organize data into fields

• Ex: social media video with hashtags

Unstructured data

Semi-Structured data

Structured data

Natural Language Understanding (NLU)

• Understanding the meaning of human language

• Semantic analysis

• Sentiment analysis

• Named Entity recognition

• Example: Spam filters

Natural Language Generation (NLG)

• Generates human language

• Takes structured data as input and turns it into coherent and readable text or speech

• Data-to-text transformation; automatic report generation

Pixels

Grid of tiny colored dots

• Represents a tiny portion of the image

• Contains information about its color and intensity

Resolution

• Total number of pixels along the width and height of the image

• Ex: 1920x1080 px.

  • 1920 pixels horizontally

  • 1080 pixels vertically

AI in cv

• Computers convert them into numbers

  • Each image divided into a series of numbers that represent the color and intensity of each pixel

• Allows AI algorithms to process the image mathematically and extract info.

Cognitive Computing

• mimic the way the human brain works in processing information and making decisions

• Examples

  • IBM Watson

  • Microsoft Cognitive service

• Extends trad. AI’s capabilities using CV, NLP, DS

Machine learning

• Enables computers to learn from data and make predictions or decisions without being explicitly programmed.

Deep learning

• Imitates the human brain in processing data and creating patterns for use in decision making

• Structure of the neurons and their connections

• Permits a machine to train itself to perform a task

• Applications

  • Aerospace and defence: Identifying objects from satellites

  • Medical research: Automatically detect cancer cells

Artificial Neural Networks (ANNs)

Core heart and concept of Machine Learning

Mimic biological neurons

• node layers

  • containing an input layer

  • one or multiple hidden layers

  • an output layer.

Deep Neural Network.

Number of Layers including the Input and Output Layer is more than three

MACHINE LEARNING

• Less accurate

• Trains on CPU

• Divides the tasks into sub-tasks, solves them individually and finally combines the results

• Can train on less data

• Less time to train

DEEP LEARNING

• Requires large datasets

• Highly accurate

• Longer to train

• Requires GPU

• Solves problem end to end

Supervised learning

• Trained on labelled data

  • Input data is accompanied by the correct output

• Learns based on example input-output pairs given

• Goal: Model to make predictions on unseen data

• Example

  • linear regression

  • logistic regression

  • Neural networks

Unsupervised Learning

• Trains on unlabelled data

  • input data is not accompanied by the correct output

• Finds hidden patterns or structure in the input data without explicit guidance

• Goal: Discover inherent structures or relationships

• Example

  • k-means clustering

  • hierarchical clustering

Reinforcement Learning

• Interacting with an environment to maximize cumulative rewards

• trial and error; receives feedback, rewards and penalties

• Goal: learn a policy or strategy; make a sequence of decisions over time

• Example

  • Q-learning

  • deep Q-networks (DQN)

BENEFITS

• Increased efficiency and productivity

• Improved decision-making

• Enhanced innovation and creativity

• Progress in science and healthcare

LIMITATIONS

• Job displacement

• Ethical considerations

• Lack of explainability

• Data privacy and security