AI Project Cycle

Data Acquisition
This stage involves collecting accurate and relevant data from authentic and reliable
sources. We collect data needed to train and test the AI model.
Data is raw facts and figures. It can be images, audios, videos, text etc.

Data Sources
● Surveys
● Web Scraping
● Sensors
● Cameras
● Observations
● APIs
Note: An Application Programming Interface (API) allows two applications to interact with
one another and share information.
Eg: Pay with Paypal, Login with Facebook, TripAdvisor, Pokemon Go etc.

Note: One of the most reliable and authentic sources of information are the open-sourced
websites hosted by the government.
Eg: data.gov.in, india.gov.in

Two types of data provided to the model are,

• Training dataset - is the data used to train the model.
• Testing dataset - is the data used to test the model.
Note: Usually, about 80% of the whole data set is used for training and 20% is used for
testing.
Eg: For a salary prediction system, the data fed into the system for training it (Training
Data) is the employee’s previous salaries and the next salary prediction data set is the
Testing Data.

Data Features
Data features refer to the type of data you want to collect about the given problem. After
specifying the Data features, we will understand what sort of data is to be collected and
from where we can get this data.
Eg: In the salary prediction example, data features would be salary amount, increment
percentage, increment period, bonus, etc.

System Map
A system map is a visual representation that shows how different parts of a system are
connected and interact with each other. It helps to understand the relationships and
dependencies between various components within a complex system. ● In a system
map, we represent relationships using arrows.
● Loops within a system map represent a specific chain of causes and effects. ● The
‘+’ and ‘-’ signs indicate the nature of the relationship between elements. ● If an
arrow goes from X to Y with a positive sign, it indicates that as X increases,
Y also increases. Similarly, if an arrow goes from X to Y with a negative sign, it
indicates that as X increases Y decreases.
● A tool that can be used for creating system maps is Loopy.
Data Exploration
This stage analyzes and understands the collected data. The stage involves cleaning,
visualizing, and examining the data to identify trends, relationships and patterns within
the
data, which will help the development of the model. This stage is also known as
Exploratory Data Analysis (EDA).
The technique used for Data Exploration is Data Visualization.
The different types of data representations are, Bar graph, Line graph, Scatter Plot, Pie
chart, Area chart, Pivot Table, Column chart, Area Map etc.

Note: datavizcatalogue.com provides various data visualization techniques.

AI, Machine Learning & Deep Learning

Artificial Intelligence, or AI refers to any technique that enables computers to mimic

human intelligence.

Machine Learning, or ML enables machines to improve at tasks with experience. The

machine learns from its mistakes and takes them into consideration in the next execution. It
improvises itself using its own experiences.
Deep Learning, or DL, enables software to train itself to perform tasks with vast amounts
of data. The machine is trained with huge amounts of data which helps it into training itself
around the data. Such machines are intelligent enough to develop algorithms for
themselves.

Note: Artificial Intelligence is the umbrella term which holds both Deep Learning as well
as Machine Learning.
Deep Learning, is the very specific learning approach which is a subset of Machine
Learning as it comprises multiple Machine Learning algorithms.

Modelling
AI Modelling refers to developing algorithms, also called models which can be trained to
get intelligent outputs. ie, writing codes to make a machine artificially intelligent.

Rule based Approach

It is the AI modeling approach where the rules are defined by the developer. The machine
follows the rules or instructions mentioned by the developer and performs its task
accordingly.
We feed the data along with rules to the machine and the machine after getting trained on
them is now able to predict answers.
A drawback/feature for this approach is that the learning is static. The machine, once
trained, does not take into consideration any changes made in the original training dataset.
That is, if you try testing the machine on a dataset which is different from the rules and data
you fed it at the training stage, the machine will fail and will not learn from its mistake.
Once trained, the model cannot improvise itself on the basis of feedback.

Eg:- Decision Tree (Real world example - Customer Support Chatbot) The basic
structure of a Decision Tree starts from the root. From there, the tree diverges into
multiple directions with the help of arrows called branches. In the end, the final
decision is where the tree ends. These decisions are termed as the leaves of the tree.
Learning Based Approach
It refers to AI modeling where the machine learns by itself.
The AI model gets trained on the data fed to it and then is able to design a model
which is adaptive to the change in data. ie, the model would modify itself
according to the changes which occur in the data.

Types of Machine Learning

● Supervised learning
Eg: Classification model (Eg: Spam or Not Spam in Email)
Regression model (Eg: House Price Prediction)
● Unsupervised learning
Eg: Clustering (Eg: Medical imaging, Genetic research)
Association (Eg: Recommendation System, Market Basket Analysis, Medical
Diagnosis)
Dimensionality Reduction (Eg: Image compression, Speech recognition) ●
Reinforcement learning
Evaluation
• Evaluation is the process of calculating the efficiency and performance of the AI
model, based on outputs, by feeding the test dataset into the model and comparing
the output (prediction) with actual answers(reality).
• The efficiency of the model is calculated on the basis of the following
parameters:- Accuracy, Precision, Recall and F1 Score

Model Evaluation Terminologies (Confusion Matrix)

A confusion matrix is a table used to evaluate the performance of a classification
model.
Note: It’s not recommended to use training data which we used to build the model to
evaluate it. This is because our model will simply remember the whole training set, and will
therefore always predict the correct label for any point in the training set. This is known as
overfitting.

6. Deployment
Deployment is the final stage in the AI project cycle where the AI model or solution is
implemented in a real-world scenario where it can be used by the users. This includes
integrating the model with existing systems, monitoring its performance, and making
necessary adjustments.
Eg: successful AI projects- self driving cars, medical diagnosis systems, and chatbots.

AI ETHICS & AI ETHICS & MORALITY

Ethics
Ethics can be defined as a system of moral principles that govern an individual's actions or
behavior.

AI Ethics
“AI Ethics” is a term which is used to deal with all the ethical concerns and issues related to
AI systems.
AI Ethics is the principle of designing artificially intelligent systems. It ensures that an
automated system is able to respond to situations in an ethical way.

Morals
Morals usually refer to personal beliefs influenced by factors such as society, culture, and
individual experiences.
Morals Ethics

The beliefs dictated by our society. The guiding principles to decide what is
good or bad.
They are not fixed and can be different These are values that a person
for different communities/societies. themselves chooses for their life.

Eg: Eg:
● Always speak the truth ● Is it good to speak the truth in all situations?
● Always be loyal ● Is it good to be loyal under all
● Always be generous circumstances? ● Is it necessary to always be
generous?
Ethics & Personal Data
Ethics & Personal Data refers to the principles and practices that ensure the respectful and
responsible handling of individuals' personal information. It involves safeguarding privacy,
obtaining informed consent, and ensuring transparency about data collection and usage.

AI Ethics Principles
The following principles in AI Ethics affect the quality of AI solutions:
▪ Human Rights
▪ Bias
▪ Privacy
▪ Inclusion

1. Human Rights:
When building AI solutions, we need to ensure that they follow human rights. The system
must not take away freedom, discriminate among people, deprive people of jobs etc. Eg:
An AI algorithm to consider job applications should be designed to avoid discrimination
based on race or gender and must ensure equality.

2. Bias:
AI bias is the tendency of AI systems to make unfair, prejudiced decisions due to biased
data, algorithms, or design choices.
Bias is partiality or preference for one over the other. The bias in training data or algorithms
affects the output and performance of the model.
● Eg: Amazon recruitment system (rejecting applications from female candidates) ●
Google Photos (search results showing Females images for Teacher/Chef, Males for
Doctor)
● Voice Assistants (mostly Female voice)
● Problem of Facts and Interpretation of Facts (Microsoft’s chatbot TAY(Thinking
About You) on twitter, was designed to experiment with conversational
understanding. But it started mimicking the offensive behavior of other Twitter
users).

3. Privacy:
Data privacy refers to the protection of personal information from unauthorized access or
misuse. It involves ensuring that people's data is collected, stored, and used in a way that
respects their rights and confidentiality.
Eg: Online shopping recommendation tracking your browsing and purchase history, Fitness
tracking app collecting personal health data.
4. Inclusion: (AI Access)
This principle ensures that AI systems are accessible and beneficial to all individuals,
including marginalized or underserved groups.
Eg: A language translation AI should support a wide range of languages to accommodate
people from different linguistic backgrounds.

Other Ethical concerns related to adoption of AI systems:

1. Job Loss (Unemployment)
2. Increasing inequalities
3. Negative Adoptions
4. Black Box Problem (Neither AI systems nor the developer explain the process of a
particular decision making. This is black box problem.)

Q1. Identify which of the following are examples of classification/regression/clustering.

a. Making a diagnosis for a patient on the basis of their symptoms - Classification b. Price
prediction for a house coming up on sale - Regression
c. HR shortlisting applications for interview based on information provided in candidates'
resume - Classification
d. Credit Card Fraud prevention - Classification
e. SPAM filters - Classification

Q2. What do you understand by the terms True Positive and False Positive?
True Positive = Correctly identifying a positive case.
False Positive = Incorrectly identifying something as positive when it’s actually negative.
Q3. The process of breaking down the big problem into a series of simple steps is
known as:
a. Efficiency
b. Modularity
c. Both a) and b)
d. None of the above
Modularity - refers to breaking down a complex problem into smaller, manageable, and
independent parts or modules. Each module can be developed and tested separately, making
the overall problem easier to solve.
Q4. The guiding principles to decide what is good or bad is known as morality. Q5.

When building AI solutions, we need to ensure that they follow ethical guidelines.

Q6. Praneet has taken extra packets of mouth freshener after dinner from a restaurant. Is it
considered theft?” Is it a Moral or Ethical concern? - Moral concern
Q7. Rakshit and Aman are talking about purchasing a new mobile. They discuss various
features which they want in their mobile. Aman finds that he started getting notification of
various models of Mobiles that meet his requirement? Write which ethical concern the
above example depicts. - Privacy
Q8. “Preference for one over the other” is known as bias.
Q9. Artificial Intelligence and machine learning systems can display unfair behaviour if not
trained properly. (True/False) - TRUE
Q10. Searching results for images of personal secretary on Google, displaying
predominantly the images of Women is an example of bias.
Q11. An Ethical AI framework makes sure that transparency, fairness and
accountability is developed into the systems to provide unbiased results.
(True/False) - TRUE

Q12. Craft a description of how considerations for inclusivity are addressed during the
development of AI models.
When creating AI models, we ensure inclusivity by using diverse data that represents
different people and groups. We regularly check the model to make sure it doesn't favor or
hurt any group unfairly. We involve a variety of people in the design process to get different
viewpoints and spot potential problems. We also keep things transparent by explaining how
the model works and what data it uses. Finally, we listen to feedback from different users to
make continuous improvements and address any inclusivity issues.

Q13. As Artificially Intelligent machines become more and more powerful, their ability
to accomplish tedious tasks is becoming better. Hence, it is now that AI machines have
started replacing humans in factories. While people see it in a negative way and say AI
has the power to bring mass unemployment and one day, machines would enslave
humans, on the other hand, other people say that machines are meant to ease our lives.
If machines over take monotonous and tedious tasks, humans should upgrade their
skills to remain their masters always. What according to you is a better approach
towards this ethical concern? Justify your answer.
The best approach is to use AI to handle boring and repetitive tasks while helping people
learn new skills for different jobs. This way, AI can make work easier and more efficient,
and people can find new opportunities instead of losing their jobs. By focusing on training
and education, we can ensure that AI benefits everyone and creates a better balance between
technology and human work.