What is knowledge?
Knowledge includes information & skills acquired through education or experience. It involves theoretical & practical
understanding of a subject.
What is intelligence?
It is the ability to apply knowledge to solve problems.
What are the different types of intelligence?
Howard Gardner of Harvard University proposed the Theory of Multiple Intelligences in 1983.
What is A. I.?
A. I. is a scientific domain that deals with making a system intelligent.
Groundwork for AI:
• 1921: Czech playwright Karel Čapek released a science fiction play "Rossum's Universal Robots" which introduced the idea of
"artificial people" which he named robot.
• 1950: Alan Turing published "Computer Machinery and Intelligence" in which he explored the question "Can machines think?"
He proposed a test of machine intelligence called The Imitation Game. He speculated that by 2000, machines would be able
to pass his imitation game in a way that would convince a significant portion of people. https://youtu.be/3wLqsRLvV-c
• 1950: Claude Shannon demonstrated his mechanical maze-solving mouse, named Theseus.
https://www.youtube.com/watch?v=Q8rVJZ-VDKQ
• 1952: A computer scientist named Arthur Samuel developed a program to play checkers, which was the first to ever learn the
game independently.
Page 1 of 10
Dartmouth Conference, 1956
In 1956, John McCarthy, a Professor at Dartmouth College, put together a group of academicians to clarify and develop ideas about
thinking machines. He picked the name 'Artificial Intelligence' for the new field. The Rockefeller Foundation funded a 2-month
seminar at Dartmouth College. For his contribution to the field, John McCarthy is known as the Father of A. I.
Why is AI needed?
Traditional computers are limited in the sense that they require algorithms. Algorithms use certain rules to process the data.
However, when the rules are changed, the algorithm may not work. Artificial intelligence can work dynamically even when the
rules are change.
Uses of A. I. – A. I. is changing the human world (the new electricity):
1. Banking & Finance - Making quick business decisions. A.I. is used to mine large quantities of data and arrive at quick decisions
in places like the stock market, where time is of the essence.
• Fraud detection mechanisms provide higher quality of services to customers. https://www.americanexpress.com/en-
us/newsroom/articles/products-and-services/backing-american-express-customers-how-machine-learning-
halt.html:~:text=Protecting%20Amex%20Card%20Members&text=At%20Amex%2C%20we%20developed%20a,are%20from%20a%20genuine%20customer
• HDFC Bank has introduced a chatbot Eva (Electronic Virtual Assistant), for their services. https://www.hdfcbank.com/personal/ways-
to-bank/eva
2. Creating New Jobs - About 133 million more engaging, less repetitive new jobs are expected to be created.
3. Bridging Language divides. AI-powered language tools like Duolingo help to learn new languages & translate speech & text in
real-time.
4. Delivering Health Care.
• After doctors at the University of Tokyo had tried for months to diagnose a woman's ailment, IBM Watson analysed her
genetic data and 20 million cancer research papers in 10 minutes to suggest an effective treatment. Watson now aids
doctors and health insurance companies by revealing insights from vast amounts of unstructured data.
https://ictandhealth.com/news/ibm-watson-correctly-diagnosis-rare-form-of-leukemia
• ML models are more accurate than humans in analysing MRI reports & predict whether a patient's tumour is malignant
or benign.
• Fitbits can collect data like sleep patterns of the wearer, calories burnt, heart rate, etc., which can help with early detection
& disease diagnosis. It can even schedule a visit to the doctor.
5. Transport. Automobiles are changing from an engine with a chassis around it to a software-controlled intelligent machine.
Autonomous vehicles use light detection and ranging systems (LIDARs) & A. I. to navigate.
6. A. I. is creative. It can:
• Create art work https://hotpot.ai/art-maker, https://openai.com/blog/dall-e/, https://creator.nightcafe.studio/
• Write articles https://app.inferkit.com/demo
• Compose poems (https://www.poem-generator.org.uk/quick/).
7. Social Media - Social Media is not just a platform for networking & self-expression. It shapes our choices, ideologies &
temperament by using tools which work silently in the background, showing us posts that we "might" like and advertising
products we "might" find useful based on our browsing history.
Page 2 of 10
8. Entertainment - Netflix & Amazon Prime, rely on info collected from users to make recommendations. New content is added
every minute. It is difficult to classify them and make them easy to search. AI tools analyse the content of videos frame by
frame and identify objects to feature appropriate tags. Netflix found from a survey that customers tend to switch channels if
they do not find something suitable to watch in 60 to 90 seconds. By building suitable personalized recommendation systems,
they estimate that they gain $1billion annually.
9. Surveillance - Following the 2013 Boston Marathon bombing, 10 TB of data (call logs, text messages, photos, videos and social
media feed) was collected and analysed using A. I. Within 24 hours, 2 suspects were identified and they were apprehended
within 4 days.
What are the risks from A. I.?
• In 2017, Elon Musk claimed that technology poses "a fundamental risk to the existence of human civilization."
• Prof. Stephen Hawking worried that people aren't concerned enough about the pace of change.
1. Invasion of privacy. Facial recognition can identify any individual in a crowd. The data gathering ability of A. I. also mean that
a timeline of our daily activities can be created, accessing CCTV footage and social networking sites.
2. Japan's Hotel Hennna, has culled more than half its robot contingent owing to a variety of complaints like a bot in a room kept
waking a man up, as it could not decipher his snoring.
In 2018, Facebook admitted that 5.62 lakh people in India had been "potentially affected" by a global data leak.
3. Autonomous weapons. Drones can search targets to pre-programmed instructions. (Bayraktar drones from Turkey in Russia -
Ukraine war). If they were developed by governments that do not care about human life and deployed anonymously. In 2015
a robot crushed a man to death at a Volkswagen production plant in Germany. The technician was installing a stationary robot
programmed to perform a series of assembly line tasks. The robot grabbed and crushed the technician against a metal plate.
4. Loss of jobs. According to McKinsey Global Institute, 800 million jobs could be lost worldwide due to automation by 2030.
Robots do not need to be paid, so the owners of A. I. – driven companies would make all the profits, increasing the gap
between rich and poor even further.
5. Deep Fakes. Realistic looking fake images & video eat at the reliability of evidence https://youtu.be/gLoI9hAX9dw
Page 3 of 10
Turing Test: https://youtu.be/3wLqsRLvV-c https://youtu.be/uSMm6p8H_LA
Dartmouth Conference: https://www.youtube.com/watch?v=5Ur-Nf85ARw&t=77s
Eliza: https://web.njit.edu/~ronkowit/eliza.html
Deep Blue: https://youtu.be/KF6sLCeBj0s
Roomba: https://www.youtube.com/watch?v=sbr1hfgwJ0A&pp=ygUGcm9vbWJh
Jeopardy: https://www.youtube.com/watch?v=P18EdAKuC1U&pp=ygUTSmVvcGFyZHkgaWJtIHdhdHNvbg%3D%3D
Eugene Goostman: https://youtu.be/oHL1JpPTle0
Alphago vs Lee Sedol: https://youtu.be/mzZWPcgcRD0
Fields closely related to A. I.
• Machine Learning: Statistical algorithms that enable A. I. implementation through data.
• Deep Learning: Subset of ML, which uses neural networks.
Types of AI
Weak / Narrow AI: It is the most common type of AI found today. It can perform only a single task. Examples would be Siri virtual
assistant, IBM Watson, AlphaGo.
Strong / General AI: At present, such machines don’t exist. Once developed, they will have the ability to perform a wide range of
human tasks, like independent thought. E.g., what we see in movies like Terminator.
Super AI: This is also beyond the realm of today’s technology. Once developed, it would be smarter than humans and would be
able to create technologies beyond human ability. This could lead to a singularity, where AI become self aware or reach an ability
for continuous improvement so powerful that it will evolve beyond our control.
Where do we find Artificial Intelligence?
• Virtual Assistants - AI can understand and respond to what you say, like virtual assistants such as Siri or Alexa.
• Email Spam Filters - Email providers have spam filters that sort mail into Primary, Social, Promotion, etc., saving our time.
• Make Predictions - AI analyses data to predict the weather or suggest what movie you might like to watch next.
• Social Media - Object recognition allows faces to be identified and tagged.
• Drive Cars - AI can help cars drive themselves by sensing the road and making decisions to stay safe.
• Play Games - AI can play games and learn to get better at them, like playing chess (Deep Blue) or Go (AlphaGo).
Page 4 of 10
Smart gadgets vs AI gadgets:
Famous AI machines
• Sophia is a social humanoid robot developed by Hanson Robotics (Hong Kong) in 2016. Sophia is a Saudi Arabian citizen. She
was named the UNDP's first Innovation Champion. https://www.youtube.com/watch?v=Bg_tJvCA8zw#t=2m17s
• ASIMO (Advanced Step in Innovative MObility) is a humanoid created by Honda in 2000. The name was chosen in honour of
Isaac Asimov. ASIMO can distinguish between voices and recognize 10 faces & address them by name.
https://youtu.be/QdQL11uWWcI
• BigDog, Spot, Atlas, and Handle. Created by Boston Dynamics, an American robotics company founded in 1992. It has now
owned by the Hyundai Motors. https://youtu.be/M8YjvHYbZ9w
• Roomba is a series of vacuum cleaners sold by iRobot. Introduced in September 2002, they have a set of sensors that enable
them to navigate the floor area of a home. These sensors can detect the presence of obstacles, dirty spots on the floor & steep
drops (to avoid falling down stairs). https://youtu.be/sbr1hfgwJ0A
• IBM Watson is a question-answering computer system capable of answering questions posed in natural language. It was
developed to answer questions on the quiz show Jeopardy! In 2011, Watson competed on Jeopardy! winning $1 million. In
2013, IBM announced that Watson software would be used to manage decisions in lung cancer treatment.
https://youtu.be/P18EdAKuC1U
• Waymo is an autonomous driving technology development company and part of Google. Waymo operates a commercial self-
driving taxi service. It operates without safety backup drivers in the vehicle.
Domains of A. I.
1. Data Science: It deals with alphanumeric data inputs. It involves the collection, analysis & interpretation of data to extract
insights & patterns using statistics, ML & data visualization.
Types of Data
• Structured Data - It comes in tabular form, with rows and columns that make it easy to work with. It can be in the form
of an Excel file, an SQL table or a CSV file.
• Unstructured Data - It lacks specific organization, making it more challenging to analyse compared to structured data.
Examples include images, videos, songs & song lyrics. Extracting meaningful insights from it requires specialized tools and
techniques.
• Semi-structured Data - It falls between structured and unstructured data. Examples of semi-structured data timestamps
of photographs and other files.
2. Natural Language Processing (NLP): It focuses on processing text & speech to enable computers to understand, interpret &
generate a response. It includes language translation, sentiment analysis, text summarization & speech recognition. Its sub-
fields are:
Page 5 of 10
• Natural Language Understanding: It analyses text & speech, extracting information, intent & sentiment.
• Natural Language Generation: It analyses input and turns it into text or speech.
Semantris: NLP-based game by Google https://experiments.withgoogle.com/semantris
3. Computer Vision: It deals with enabling computers to interpret and understand images & video, perform tasks such as object
detection, image classification, facial recognition & scene understanding, enabling applications such as autonomous vehicles,
medical imaging, and augmented reality. CV gives computers the ability to see and understand the world like humans use their
eyes to perceive their surroundings."
A digital image is a grid of tiny, coloured dots called pixels. Each pixel represents a tiny portion of the image and contains
information about its colour and intensity.
Resolution is the total number of pixels along the width and height of the image. An image with a resolution of 1920x1080
pixels has 1920 pixels horizontally and 1080 pixels vertically. Higher resolution images have more pixels, providing more detail.
The computer converts each pixel into numbers. These numbers allow AI algorithms to process the image mathematically and
extract meaningful information from it. For instance, AI algorithms might learn to recognize patterns in these numbers that
correspond to specific objects, like cars or faces. By analysing large amounts of labelled image data, AI systems can "learn" to
identify objects accurately.
Emoji Scavenger Hunt: https://emojiscavengerhunt.withgoogle.com/
Quickdraw: https://quickdraw.withgoogle.com/
Autodraw: https://www.autodraw.com/
Terms connected with A. I.
• Machine Learning – It is a subset of artificial intelligence (AI) that focuses on developing algorithms and models that enable
computers to learn from data and make predictions or decisions without being explicitly programmed.
• Deep Learning – It is a subset of ML, which uses Artificial Neural Networks, to simulate the working of neurons in the human
brain. They comprise of an input layer, one or more hidden layers, which do the processing and an output layer. If the output
of a node is above a specified threshold, that node is activated, sending data to the next layer of the network. Else, no data is
passed along to the next layer of the network. If there are multiple hidden layers, it is called a Deep Neural Network.
Diagram of a neural network
Page 6 of 10
Example:
The MNIST (Modified National Institute of Standards & Technology) dataset uses a dataset of 70,000 grayscale images of
handwritten digits. Each image is 28 × 28 pixels, labelled 0–9. It is used to process cheques, numberplates and mail rapidly.
To improve the performance of the machine, you expose it to more data to ensure that the machine is trained on all variations of
each digit. The more data you provide for the algorithm, the better the model gets.
Types of Machine Learning
Supervised Learning
• the goal is to learn a mapping function from inputs to outputs so that the model can make predictions on unseen data.
• the algorithm maps input data to output labels based on example input-output pairs provided during the training phase.
• algorithms include linear regression, logistic regression, decision trees, support vector machines & neural networks.
Page 7 of 10
Videos from Andrew Ng
• Regression: https://www.youtube.com/watch?v=SwuP6cz6cHc&list=PLWD7QtH5pagQevEwjEOCQi1Cgqe3zKf2s&index=3
• (Classification): https://youtu.be/p-jKwSzwOYU?list=PLWD7QtH5pagQevEwjEOCQi1Cgqe3zKf2s
Unsupervised Learning
• the model learns from unlabelled data (the input data is not accompanied by the correct output).
• the goal is to explore and discover relationships within the data, such as clusters, associations, or anomalies.
• algorithms include k-means clustering, principal component analysis & autoencoders.
Reinforcement Learning: https://youtu.be/yP-QM4hTjz0
• an agent learns to make decisions by interacting with an environment to maximize rewards.
• the agent learns through trial and error by taking actions and receiving feedback from the environment in the form of
rewards or penalties.
• the goal is to learn a policy or strategy that guides the agent to take actions that lead to the highest cumulative reward
over time.
• algorithms include Q-learning.
Types of ML: Supervised and Unsupervised (ML Tidbits):
https://youtu.be/wy-m6sd1BOA
Applications of A. I.
• Earn a credential on A. I. Fundamentals on IBM Skills Build:
https://students.yourlearning.ibm.com/activity/PLAN-CC702B39D429
• Semantris is an NLP game by Google based on Word association powered by semantic search.
https://experiments.withgoogle.com/semantris
• Quickdraw is a game built with machine learning. We draw, and a neural network tries to guess what you're drawing.
https://quickdraw.withgoogle.com/
• The experiment based on the computer vision domain of AI. It identifies that you draw and suggests the related images.
To play the game, visit the following link on any computing device with speakers.
https://www.autodraw.com/
Page 8 of 10
Exercises
A. Multiple-choice questions (MCQs):
1. Who is referred to as the "Father of AI"?
a. Alan Turing b. John McCarthy c. Marvin Minsky d. Herbert Simon
2. In which year was the term "Artificial Intelligence" coined by John McCarthy?
a. 1930 b. 1956 c. 1970 d. 2000
3. What does the term "Data is the new oil" imply?
a. Data is as valuable as oil. b. Data is used as fuel for machines.
c. Data is a non-renewable resource. d. Data and oil are unrelated.
4. Which category of ML category occurs in the presence of a teacher?
a. Unsupervised Learning b. Reinforcement Learning c. Supervised Learning d. Deep Learning
5. What is the role of reinforcement learning in machine learning?
a. Creating rules automatically b. Recognizing patterns in untagged data
c. Rewarding desired behaviours and/or penalizing undesirable ones
d. Mimicking human conversation through voice or text
6. Which AI application automatically separates emails into "Spam" and "Not Spam" categories?
a. Gmail b. YouTube c. Flipkart d. Watson
B. Fill in the Blanks:
1. To determine if a machine is AI-based, consider its ability to perform tasks that typically require ______________
intelligence.
2. Artificial intelligence enables a machine to carry out cognitive tasks typically performed by ______________.
3. The categories of machine learning are ______________, ______________ and ______________.
4. Machine learning can be used for online fraud detection to make cyberspace a ______________ place.
C. True or False:
1. Chatbots like Alexa and Siri are examples of virtual assistants.
2. Supervised learning involves training a computer system without labelled input data.
3. Unstructured data can be easily analysed using traditional relational database techniques.
4. ML is not used in everyday applications like virtual personal assistants & fraud detection.
D. Short Answer Questions:
1. How is machine learning related to AI?
2. Define data. List the types of data.
3. Define machine learning.
4. What is Reinforcement Learning? Name any 2 applications of Reinforcement Learning at School.
5. How do you understand whether a machine is AI based or not? Explain with the help of an example.
Page 9 of 10
E. Competency Based Questions
1. Rahul is an architect. He has designed and built a beautiful home for his client in Pune. He has installed solar water heater,
security cameras, rainwater harvesting system, cleaning robots, smart lighting, automatic door, Siri, automatic washing
machine. Which of these are AI systems. How did you choose these as AI systems?
AI systems are: smart TV, cleaning robots, smart lighting, Siri.
The parameters to choose are: ability to make decisions, problem-solving, recommendations, adapt to new situations & learn
from past experiences.
2. You were designing a robot to sort recyclable items like glass, plastic, and paper. Which type of learning would be used to help
the robot?
Supervised learning
3. You have a bunch of fruits, and you want the AI system to organize them into groups based on similarities? Which type of
learning would be used?
Unsupervised learning
4. Shankar runs a company that deals with services to customers related to financial investments. He uses AI in his company due
to which his employees are facing less job responsibility, customers are feeling insecure about their data. What is this scenario
known as?
Potential impact of AI on society and lack of data privacy.
5. Jatin is a student who has just enrolled in a course in AI. He attended a few introductory classes and learned that systems can
learn from the data using algorithms to perform a task without explicitly programming it. In some situations, the system mimics
the human brain's learning process. Identify the concepts in this scenario.
Machine learning and Deep learning
Videos to watch
• Understand about AI: https://youtu.be/ad79nYk2keg?feature=shared
• History of AI: https://www.youtube.com/watch?v=L-9ZqkeNNJE
• Introduction to Data Science: https://youtu.be/X3paOmcrTjQ?feature=shared
• Natural Language Processing: https://www.youtube.com/watch?v=CMrHM8a3hqw
• Introduction to Computer Vision: https://www.youtube.com/watch?v=puB4LuRNys
• Cognitive Computing: https://www.youtube.com/watch?v=Zsl7ttA9Kcg
Page 10 of 10
