Class

XI ANSWER KEY
Part-B: Subject Specific Skills Artificial Intelligence

1. Introduction to AI
Exercise
Unsolved Questions
A. 1. Deep Learning 2. Reinforcement 3. Artificial Super Intelligence
4. virtual 5. Machine Learning
B. 1. Machine learning algorithms like Decision Tree is used in fraud detection. This is supervised
learning where classification is performed to categorise fraudulent and regular transactions.
2.
Can do Cannot do
Predictions while commuting - like Traffic Go beyond their tasks
Forecasts on Google Maps.
Product recommendations Learn on small amounts of data

3. • This layer is responsible for processing the input fed into the system
• E
 ach node of the hidden layer has its functionality and hence, the defined machine learning
algorithm gets executed on the data received from the input layer.
• T
 he number of hidden layers in a neural network system depends on the complexity of
the function for which the network has been designed.
4. Data Analyst, Data Scientist, Product Manager (any two)
5.
Data Big Data
The volume of the data is small. The volume of the data is huge.
Can be handled by traditional DBMS. Cannot be handled by traditional DBMS.
Data handling is not expensive. Data handling and management are quite
expensive.

C. 1. Machine learning is the process of computers discovering useful information without being
directed where to seek. Instead, they use algorithms that learn from data in an iterative
process. Machine learning algorithms can process vast volumes of data and extract meaningful

Touchpad Artificial Intelligence-IX (Answer Key) 13

 information by utilising various programming techniques. In this approach, they can improve
on earlier iterations by learning from the data presented to them.
Types of ML

a. Supervised Learning - One of the most fundamental types of machine learning is
supervised learning. The machine learning algorithm is trained on labelled data in this
case.
b. U
 nsupervised Learning - Unsupervised machine learning has the advantage of working
with unlabeled data. This means that no human labour is necessary to make the dataset
machine-readable, allowing the program to work on much larger datasets. Unsupervised
learning lacks labels to deal with, leading in the formation of hidden structures. The
program perceives relationships between data points in an abstract fashion, with no
human input required.
c. R
 einforcement Learning - Reinforcement learning is directly inspired by how humans
learn from data in their daily lives. It has an algorithm that uses trial-and-error to better
itself and learn from new scenarios. Favourable outcomes are promoted or ‘reinforced,
whilst unfavourable outcomes are discouraged or ‘punished’.
Applications of ML – Image recognition, product recommendation, Email filtering and
Spam detection, online fraud detection.
2. N
 o AI and ML are not the same. However, they are definitely connected. AI is a broader
term that refers to the creation of intelligent computers that can replicate human thinking
capability and behaviour, whereas machine learning is an application or subset of AI that
enables machines to learn from data without being explicitly programmed.
 rtificial intelligence systems do not need to be pre-programmed; instead, they use algorithms
A
that function with their own intelligence. It employs machine learning methods such as
Reinforcement learning and deep learning neural networks. AI is being employed in a variety
of applications, including Siri, Google’s AlphaGo, and AI in chess playing.
 achine learning allows a computer system to forecast or make judgments based on
M
historical data without being explicitly programmed. Machine learning makes extensive use
of structured and semi-structured data in order for a machine learning model to produce
reliable results or make predictions based on that data. Machine learning is based on
algorithms that learn on their own utilising past data. It only works in certain domains.
3. a. Snapchat filters – Computer Vision : Computer vision is a domain of AI that enables machines
to derive meaningful information from digital images, videos, and other visual inputs; and
take actions or make recommendations based on that information.
b. Virtual Assistants like Siri and Alexa – Natural Language Processing : NLP employs natural
language for interaction between computers and humans. Natural language processing
(NLP) aims to extract information from the spoken and written word using algorithms, and
natural language refers to the language that is spoken and written by people.

14 Touchpad Artificial Intelligence-IX (Answer Key)

 4. Artificial General Intelligence, also known as Strong AI, is the stage at which machines will
most likely be able to reason and make decisions in the same way that humans do. Strong
AI has a complicated algorithm that allows it to operate in various scenarios, whereas weak
AIs have all of their actions pre-programmed by humans. Strong AI-powered devices will
have their own minds. They will be able to analyse information and make judgments on their
own, whereas weak AI-based computers can merely replicate human behaviour. Till date,
there are no known cases of Strong AI.
Examples of weak AI – virtual assistants like Alexa and Siri, autonomous vehicles.
5. Data labelling in machine learning is the process of identifying raw data (pictures, text files,
videos, etc.) and adding one or more relevant and useful labels to provide context so that
a machine learning model may learn from it.
Labelled data often starts with a set of unlabelled data and then adds some sort of meaningful
“tag,” “label,” or “class” to each item of that unlabelled data that is either informative or
desirable to know. After collecting a labelled dataset, machine learning models may be
applied to the data so that new unlabelled data can be provided to the model and a likely
label for that piece of unlabelled data can be estimated or predicted.
Example - To construct your training dataset for natural language processing, you must
first manually identify relevant chunks of text or classify the text with particular labels. For
example, you may wish to detect the sentiment or intent of a text blurb, identify parts of
speech, classify proper nouns such as locations and people, and recognise text in photos,
PDFs, or other media. To accomplish this, draw bounding boxes around text and then manually
transcribe the text in your training dataset. Natural language processing models are utilised
for sentiment analysis, entity name identification, and optical character recognition.

In Life
Can have multiple solutions

Deep Thinking
Can have multiple solutions

Ready
1. Can have multiple solutions
2.  L algorithms can supplement field data and automate functions mostly linked to regulation
M
and optimization. Furthermore, machine learning and computer vision have enriched numerous
sectors, including medical diagnostics, statistical data analysis and algorithms, scientific research,
and so on. Such approaches have already been implemented in the fields of smartphone apps,
computer equipment, online websites, cybersecurity, and so on. Examples – Google Maps, Email
Spam Filter.

Touchpad Artificial Intelligence-IX (Answer Key) 15

3. Can have multiple solutions
4.  hidden layer is positioned between the algorithm’s input and output in neural networks, where
A
the function assigns weights to the inputs and guides them through an activation function as
the output. Hidden layers allow a neural network’s function to be broken down into specific data
modifications. Each hidden layer function is tailored to generate a certain result. For example, a
hidden layer function that identifies human eyes and ears may be used in concert with succeeding
layers to detect faces in photos. While the functions to detect eyes alone are insufficient to
distinguish things independently, they can interact together within a neural network.

2. AI Applications and Methodologies

Exercise
Unsolved Questions
A. 1. NLP 2. machine learning 3. Cognitive Computing
4. Sophia 5. Recommendation system
B. 1. Benefits of driverless cars:
• Reduction of traffic congestion (30% fewer vehicles on the road)
• Reduce CO2 emissions by 80% worldwide
2. Google calls the technology “Nowcasting” because it is set up to predict weather zero to
six hours in advance and focus on weather events like thunderstorms that change rapidly
from clear skies to heavy rain to gusty winds and vice versa. In essence, the method turns
weather forecasting into a computer vision problem. Based on progressive images of cloud
formation and movement over a short period, a machine learning algorithm predicts how
the pattern will develop in the next few hours.
3. Semantic segmentation describes the process of assigning each pixel in an image to a class
name (for example, flower, person, road, sky, sea, or car). The idea of segmentation is to
teach computers to process and understand an image at the pixel level. In simple terms,
computers can segment an image, paint objects in the image with different colours, and
predict what is in them. For example, an autonomous car recognizes objects on the road
and labels them as per the classes already fed into it.
4. Some of the advantages of rule-based chatbots are:
• Easy to train (less expensive)
• Very responsible and safe
• May contain some interactive elements and media
• Not limited to text interactions

16 Touchpad Artificial Intelligence-IX (Answer Key)

 5. Some of its characteristics are:
•  as the ability to predict and adapt: its ability to understand data patterns is used to
H
make future decisions and predictions.
• Continuously learns: learns from data patterns.
• AI is reactive: it perceives a problem and acts according to the perception.
• AI’s cutting: edge technology will be used in many other areas in the future.
Many applications and tools are using AI, that has a direct impact on our day-to-day life.
So, we need to understand the kind of systems that can be developed using AI.
C. 1. Challenges posed by AI include:
•  ias: AI systems learn from data. If the data is biased, i.e., data favours a particular group/
B
gender/class, the resulting decisions made by the AI model are also biased.
•  nemployment: Due to automation, certain jobs will become obsolete. There will be
U
some highly skilled jobs created too, but some jobs will definitely get affected by AI.
•  isparate control over data: Data is AI fuel—the more data you have, the smarter are
D
the AI machines. Tech giants are investing heavily in data collection and AI projects. This
gives them an unfair advantage over their smaller competitors.
•  rivacy: AI can be used to identify, track and monitor people on multiple devices, whether
P
at work, at home, or in a public place. To make matters worse, AI doesn’t forget anything.
Once the AI knows you, it remembers you forever!
2. i)  griculture: Agriculture is a sector that faces numerous challenges, such as weather, water
A
consumption or soil conditions, etc. Real-time data analysis helps farmers to improve
their crop yields and in turn their profits too.
ii) E
 ducation: Textbooks are digitized with the help of AI, early-stage virtual tutors support
human teachers, and facial analytics measure students’ emotions to determine who is
struggling or bored and to better tailor the experience to your individual needs.
iii) M
 anufacturing: AI-powered robots work with humans to perform a variety of tasks, such
as assembling and stacking, and predictive analytics sensors keep equipment running
smoothly.
2. Various applications of NLP are
i) Text Summarization - Text summary is a technique for shortening long passages of
text into manageable pieces of information. The goal is to develop a logical and fluent
summary that only includes the document’s major ideas.
ii) T
 ext Recognition - In recent years, automatic reading of a vehicle’s license plate,
billboards, or text recognition has acquired a lot of popularity. Optical Character
Recognition (OCR) is a popular text recognition technique that turns scanned images of
text into an editable format.

Touchpad Artificial Intelligence-IX (Answer Key) 17

 iii) S
 peech Processing - Speech processing uses NLP so that smart devices like smartphones
can interact with users through verbal language. One of the best-known examples of
speech recognition technology on a mobile device is Apple’s Siri speech recognition
service.
iv) S
 entiment Analysis - Positive and negative reviews are powerful because they can build
or ruin a brand’s reputation. Did you know that 40% of shoppers form an opinion about
a business after reading the online reviews? This gives you an idea of the importance
of tracking the sentiment of conversations around your brand, whether it’s in reviews,
surveys, social media, emails, and more.
4. These machines can work tirelessly and can even take over the most monotonous job. An
example is SOPHIA. Built by Hanson Robotics, Sophia is a social humanoid robot that can
have a conversation with you, answer questions, and remember you. Hanson designed Sophia
as a suitable companion for the elderly in nursing homes or to help crowds at major events
or in parks.
5. Autonomous systems accomplish tasks by interacting with their surroundings with minimum
or no human involvement. These systems must be aware of the world around them. An
autonomous car, also known as an autonomous vehicle (AV), is so called because it can sense
its surroundings and move safely with little or no human intervention. Self-driving cars work
on a combination of technologies as follows:
•  omputer Vision: Computer vision allows the car to see/perceive its surroundings. It
C
uses ‘High Optic Camera’, ‘Radar’ and ‘Lidar’.
•  eep Learning: This is the car brain that makes driving decisions based on the information
D
gathered through various sources such as computer vision, sensors, etc.

In Life
Can have multiple solutions

Deep Thinking
Can have multiple solutions

Ready
1. a. TIM Brasil – AI powered chatbot
b. Telefonica de Argentina - AI powered chatbot
 uring pandemic, only 50% staff was allowed in organizations. Hence chatbots provided
D
a reduction in burden on call centres and the wait time for users. This also reduced the
need for expensive human resource.

18 Touchpad Artificial Intelligence-IX (Answer Key)

 2. Can have multiple solutions
3. Both AI and the human. AI can pinpoint erratic behaviour and human can verify or follow it
up.
4. Can have multiple solutions

3. Maths For AI
Exercise
Unsolved Questions
A. 1. True 2. True 3. False 4. True 5. True
B. 1. 1
–3i + j
2
1 7
a21 i=2j=1 × –3(2) + 1
2 2
1 11
a32 i=3j=2 –3(3) + 2
2 2

2. 1×2+2×1+3×5 1 × 2 + 2 × 0 + 3 × –1
A×B= –1 × 2 + 4 × 1 + –2 × 5 –1 × 2 + 4 × 0 + –2 × –1
2×2+0×1+5×5 2 × 2 + 0 × 0 + 5 × –1

19 –1
= –8 0
29 –1

3. a = (3,4,2) and b = (4,3,2)

i j k
3 4 2 4 2 3 2 3 4
a×b= i –j +k
4 3 2 3 2 4 2 4 3
(8–6)i – (6–8) j + (9–16)k 2i + 2j –7k
4. Given that u = (4, 9, 2) and v = (–2, 5, 0)
u.v = (4×–2) + (9×5) + (2×0) = 37
v.u = (–2×4) + (5×9) + (0×2) = 37

5. The cartesian product of two relations is the concatenation of tuples belonging to the two
relations. The cartesian product is a binary operation and is denoted by (x). The cartesian
product of two relations will produce the relation which has a degree = sum of columns of
parent relations and a cardinality =product of rows of parent relations.

Touchpad Artificial Intelligence-IX (Answer Key) 19

C. 1. The union operation (Relational Algebra)
     Student      Student 1
Rollno Name Marks
Rollno Name Marks
100 Prakash 256
101 Vivek 256
102 Sudha 210
103 Punit 269
103 Punit 269
105 Jimmy 210 104 Kumar 287
106 John 276 106 John 276

Result of the union operation
Rollno Name Marks
100 Prakash 256
101 Vivek 256
102 Sudha 210
103 Punit 269
104 Kumar 287
105 Jimmy 210
106 John 276

The union operation (Set Theory)

 he union of two sets A and B is the set of elements that are in A, in B, or in both A and B.
T
It is denoted by A ∪ B. For example: If A = {20, 21, 22, 23} and B = {23, 24, 25}, then: A ∪ B
= {20, 21, 22, 23, 24, 25}. (The common elements occur only once).
Intersection Operation (Relational Algebra)
(Assuming the tables Student and Student1)
Intersection Student ∩ Student1
Result of the Intersection operation

Rollno Name Marks

103 Punit 269
106 John 276

Intersection Operation (Set Theory)

 he intersection of sets A and B is the set of elements that are common in A and B. It is
T
denoted by A ∩ B. For example: If A = {21, 22, 23} and B = {23, 24, 25}, then A ∩ B = {23}.
2. Mean = (18 + 22 + 19 + 10 + 12)/5 = 16.2 = 16 (approx)
18 22 19 10 12
I. –16 –16 –16 –16 –16
2 6 3 6 –4

20 Touchpad Artificial Intelligence-IX (Answer Key)

 II. Squaring all differences.
22     62     32     (-6)2     (-4)2   Variance = 97
4     36     9      36     16     std. Dev = 97 = 9.84
III. 4 + 36 + 9 + 36 +16 101
IV. 101 – 4 = 97
3. 4 3
P=
8 7
4 3
P = 8 (4×7) – (8×3) 4
7

7 –3
Adjoint of r =
–8 4
1 7 –3 1 7 –3
P–1 = =
P –8 4 4 –8 4

= 7/4 –3/4
–2 1
4 3
P=
8 7
4 3
P = 8 (4×7) – (8×3) 4
7

Adjoint of r = 7 –3
–8 4
1 7 –3 1 7 –3
P–1 = =
P –8 4 4 –8 4

= 7/4 –3/4
–2 1

1 0 0
S= 1 1 0 Expanding along Row 1
1 1 1

1 0 1 0 1 1
S = 1 – 0 + 0
1 1 1 1 1
= 1 so inverse exists
Adjoint of 5 First row
1 0 1 0 1 1
1 1 1 1 1 1
1 1 0

Touchpad Artificial Intelligence-IX (Answer Key) 21

 Second row
0 0 1 0 1 0
1 1 1 1 1 1
0 1 1
Third row
0 0 1 0 1 0
1 0 1 0 1 1
0 1 1
Adjoint Matrix with sign

1 –1 0 1 –1 0 1 –1 0
1
0 1 –1 A–1 = 0 1 –1 = 0 1 –1
1
0 0 0 0 0 0 0 0 0

4. Mean = (7+16+12+35+51+10+80+71+16+9+11+16)/12 = 27.83
Median = 12/2 = 6th item = 16

7 9 10 11 12 16 16 16 35 51 71 80

Mode = 16
5.
Height f cf
120-130 4 256
130-140 6 256
140-150 12 210
150-160 20 269
160-170 8 287
50

Median th

Me = size of N item
2
th

Me = size of 50 item
2
Median class 150 – 160
N/2 – c.f
Me = l1 + ×i
f
= 150 + 25 – 22 × 10
20
3
= 150 + = 150 + 1.5 = 151.5 cm
2

22 Touchpad Artificial Intelligence-IX (Answer Key)

 Mode
Modal Class 150 – 160
f1 – f0
Z = l1 + ×i
2f1 – f0 – f2
20 – 12
= 150 + × 10
(2 × 20) – 12 – 8
4
8 8
= 150 + × 10 = 150 + × 10 = 154 cm
40 – 20 20

In Life
Can have multiple solutions

Deep Thinking
Can have multiple solutions

Ready
1. A tensor is a container that can hold N dimensions of data. Tensors are generalisations of
matrices to N-dimensional space that are frequently and incorrectly used interchangeably
with the matrix (which is precisely a 2-dimensional tensor).
2. A sparse matrix is one that contains largely zero values. Sparse matrices differ from dense
matrices, which have a large number of non-zero values.
3. Artificial Neural Network (ANN) - Each layer of an ANN is made up of many perceptrons
or neurons. Because inputs are exclusively processed in the forward direction, ANN is also
known as a Feed-Forward Neural Network. This sort of neural network is one of the most
basic neural network versions. They send data in one way, through multiple input nodes,
until it reaches the output node. The network may or may not contain hidden node layers,
which makes its operation more understandable.
Convolutional neural networks (CNN) are among the most widely used models today. This
neural network computational model employs a variant of multilayer perceptrons and
includes one or more convolutional layers that can be completely linked or pooled. These
convolutional layers provide feature maps that capture a portion of an image, which is then
divided into rectangles and sent out for nonlinear processing.
Recurrent neural networks (RNN) are more complex. They record the results of processing
nodes and input them back into the model (they do not pass the information in one direction
only). The model is said to learn to anticipate the outcome of a layer in this manner. Each
node in the RNN model functions as a memory cell, continuing calculation and operation
implementation. If the network’s forecast is erroneous, the system self-learns and continues
to backpropagate towards the correct prediction.

Touchpad Artificial Intelligence-IX (Answer Key) 23

4. AI Values
Exercise
Unsolved Questions
A. 1. False 2. True 3. True 4. True 5. False
B. 1. Societal AI bias occurs when an AI acts in ways that reflect social intolerance or institutional
discrimination. Initially, the algorithms and data themselves seem unbiased, but their results
reinforce societal biases.
2. •  mazon discovered in 2015 that their algorithm for hiring staff was skewed against
A
women. The explanation for this was that the algorithm was trained to prefer men over
women based on the amount of resumes submitted over the previous 10 years, and
because the majority of the candidates were men.
•  he COMPAS (Correctional Offender Management Profiling for Alternative Sanctions)
T
algorithm, which is used in US court systems to forecast the possibility of a defendant
becoming a recidivist, is arguably the most well-known example of AI bias. The model
predicted twice as many false positives for recidivism for black criminals (45 percent) as
white offenders due to the data collected, the model chosen, and the whole process of
constructing the algorithm (23 percent ).
3. Structured data collection that allows for a variety of viewpoints. For a single data point,
there are frequently numerous correct opinions or labels. Your model will be more adaptable
if you collect their viewpoints and allow for valid, often subjective, conflicts.
Recognize your training data. There are classifications and labels in both academic and

commercial datasets that can introduce bias into your algorithms. You’re less likely to be
surprised by unwanted labels if you understand and own your data.
Recruit a diversified ML team to ask a variety of questions. We all bring unique

perspectives and ideas to the office. People with various backgrounds — colour, gender,
age, experience, culture, and so on – will naturally ask different questions and interact with
your model in different ways.
4. Following are some of the issues related to driverless cars:
1. 
The engineers need to strike a balance between safety and speed of the car. It can’t be
too slow or too fast.
2. Engineers must balance mobility and environmental impact.
3. 
It is difficult for the cars to recognize hand-gestures, signals given by cyclists or
pedestrians.
4. 
It is difficult for self-driving cars to decide between the lives of its passengers and the
lives of pedestrians in case of accidents.

24 Touchpad Artificial Intelligence-IX (Answer Key)

 4. AI bias is a discrepancy in the output of machine learning algorithms caused by flawed
assumptions made during the algorithm development process or biases in the training data.
C. 1. i) In 2019, Facebook was found to be in violation of the United States constitution for
enabling advertisers to target ads based on gender, ethnicity, and religion, all of which
are protected categories under the country’s legal system. Job advertisements for
nursing and secretarial positions were targeted predominantly towards women, but job
advertisements for janitors and taxi drivers were targeted mostly at men, particularly
those from minority backgrounds.
ii) W
 hen translating from languages with gender-neutral pronouns, Google Translate is
prejudiced against women. The app’s AI engine is more likely to produce translations
like “he invests” and “she looks after the children” than vice versa.
2. i) i)Unemployment. What happens when a job comes to an end?
ii) Inequality. What is the best way to disperse the money generated by machines?
iii) Humanity. What impact do machines have on human interactions and behaviour?
iv) Artificial insanity What can we do to avoid making mistakes?
v) Robots that are racist. How can we get rid of AI bias?
vi) Security. How can we keep AI safe from outside threats?
vii) Singularity. How can we keep a sophisticated intelligent system under our control?
viii) Robots have certain rights. What does it mean to treat AI with humanity?
3. Can have multiple solutions. Teacher can discuss points from the following website.
Refer: https://techcrunch.com/2018/01/06/the-light-and-dark-of-ai-powered-smartphones/
4. The first step is to make a list of all the algorithms in use. Stakeholders should be consulted
by developers to learn how and when algorithms are employed.
The second stage identifies the key to tackling algorithmic bias. Make certain that you
understand what the algorithm’s goal should be. It’s a lot easier to make sure the algorithm
is hitting the current target if you have that understanding. Organizations are encouraged to
do a table screening for label choice bias. Organizations will fill up the table with information
on the algorithm, the ideal target, the actual target, and the risk of bias.
Step three is either updating or removing any inefficient algorithms that are no longer serving
the organisation well.
Finally, researchers should keep an eye on the algorithms and conduct audits to make sure
they’re on track to meet their objectives.
5. 1. Invest in end-user education to increase AI confidence.
Inform your leaders and teams about what you’re doing and how artificial intelligence
works. Your customers must be able to see how AI will benefit them.
2. Set the stage for AI predictions by providing context and transparency.

Touchpad Artificial Intelligence-IX (Answer Key) 25