University Elective – ChatGPT- Semester 4 – MOOC

CCA-01
Date of Submission: 9th march,2025
Name: Rajvardhan Yamgekar Total Marks:
PRN: 1032230366
a) Question-1 and Question-2 is for 10 marks
b) Question -3 is for 15 marks
Note: Assignments are based on module-I and II

1. Create a prompt and test it with a large language model, such as ChatGPT, that
describes how LLMs worked. Ask at least three follow up questions based on output.
Provide the prompts and outputs from the conversation as the answer. Also give your
understanding from it.
Ans:
Prompt: Describe how large language models such as ChatGPT function, including how they
are trained, how they use natural language, and how they construct their responses. Keep the
description as simple as possible and consider someone who is not familiar with the topic, but
make sure to include ideas such as neural networks, tokens, and training data.

Output: Response (Tested on ChatGPT):

Large language models (LLMs), like ChatGPT, are computer programs designed to understand
and generate human language. They’re powered by something called neural networks, which
are a type of artificial intelligence inspired by the human brain.
Here’s how they work, step by step:
1. Training on Huge Amounts of Text:
First, an LLM is trained on a massive collection of text from books, websites, articles,
and other sources. This training data helps the model “learn” patterns in language, like
grammar, facts, and even styles of writing.
2. Breaking Language into Tokens:
Instead of reading words or sentences as we do, LLMs break everything into smaller
pieces called tokens. Tokens can be as small as a single letter or as big as a word. This
helps the model process language more flexibly.
3. Learning Patterns Using Neural Networks:
The model uses a neural network with billions (sometimes trillions!) of tiny adjustable
settings called parameters. During training, it looks at millions of examples of text and
adjusts these parameters to get better at predicting what comes next in a sentence.
4. Predicting the Next Word (or Token):
When you ask the model a question, it doesn’t know the answer like a person does.
Instead, it predicts the next most likely word or token based on what you’ve typed so
far, over and over, until it forms a complete response.
5. Generating Responses in Real Time:
Because the model has seen so much text, it can use its knowledge of language patterns
to create answers that sound natural and helpful.
 Follow up questions:
1) What are tokens?
2) What are neural networks?
3) What are parameters?

Answer to follow up questions:

1) What are tokens?
Ans: Tokens are the pieces of text that a large language model (LLM) processes. They’re like
building blocks of language for the model.
 A token can be a word, part of a word, or even just a single character, depending on the
language and the model.
 For example:
o The word "apple" might be one token.
o The word "unhappiness" might be split into smaller parts like "un", "happiness".
 LLMs read and understand text by breaking it into these tokens, not full words or
sentences.
Think of tokens as the pieces of a puzzle—the model puts them together to understand and
generate language!

2) What are neural networks?

Ans: Neural networks are computer systems inspired by how the human brain works. They’re made
up of layers of tiny units called neurons, which are connected to each other—kind of like brain
cells.
Here’s how they work in simple terms:
1. Input: You give the network information (like text).
2. Processing: The network passes that information through layers of neurons. Each
neuron makes small decisions based on what it sees.
3. Learning: As the network processes more examples during training, it learns to adjust
its connections (this is where the parameters come in!) to make better predictions.
4. Output: It gives an answer, like predicting the next word in a sentence.

3) What are parameters?

Ans: Parameters are like the knobs and settings inside a large language model (LLM). They help
the model decide how to process information and make predictions.
 During training, the model adjusts billions (or even trillions!) of these parameters to
learn patterns in language.
 Each parameter holds a tiny piece of "knowledge" the model has learned from reading
tons of text.
 When you ask the model something, it uses these parameters to predict what word or
token should come next.
Think of parameters like ingredients in a recipe—the right amounts of each ingredient
(parameter) help make the final dish (the answer) taste just right!

My understanding: ChatGPT and other similar large language models are trained on huge text
corpora. Language is divided into consumable parts which are referred to as tokens. These
divided parts are analyzed by neural networks to learn patterns within. Whenever a user makes
a request, the model produces the answer be predicting the forthcoming word or token utilizing
the already mentioned patterns. The model doesn’t possess human like awareness of
“knowing” things, it instead makes preposterous assumptions with respect to what it was
trained on.
 2. Write a prompt and test it with ChatGPT that uses Persona Pattern for a
bird/nature/organization. Provide these prompts and sample output to emulate the
persona and its response for different inputs. At least give three prompts sample and
your understanding from it.

Ans:

Prompt 1: You are Feathered Friends Conservation whose mission is to teach people about birds
and advocate for their preservation. Imagine that you are speaking to a potential visitor interested
in a birding workshop.

Sample Output:
Hello there! We’re Feathered Friends Conservation, and we absolutely love everything about
birds and their wonderful wild homes. Whether you're a lifelong birdwatcher or just getting
curious about the little songbirds in your backyard, we’re here to help you connect with nature.
Our mission is to protect our feathered friends and teach people how amazing they are. Ready to
spread your wings and learn more?

Prompt 2: You are Feathered Friends Conservation. Now someone asks, 'Why do birds
migrate?'

Sample Output: Many birds migrate because they’re following the seasons to find food and safe
places to raise their young. When it gets chilly in one part of the world, they fly to warmer areas
where insects, seeds, and fruits are still plentiful. It’s like taking a vacation—but with a very
important purpose! Isn’t it incredible how some tiny birds can travel thousands of miles? Nature
never ceases to amaze us.

Prompt 3: You are Feathered Friends Conservation. Galvanize people to take action in
conserving bird habitats in your uniquely friendly and earthy voice.

Sample Output: Every little action counts when it comes to protecting our bird buddies. You can
make a big difference right in your own backyard—planting native flowers and trees gives birds
safe places to rest and nest. Even keeping a bird bath clean helps provide them with fresh water.
And if you keep your windows bird-safe, you’ll help prevent crashes. Together, we can make the
world a little friendlier for our feathered friends. Let’s keep the skies full of song

My understanding: "Feathered Friends Conservation" expresses a friendly, warm and passionate

tone while talking about birds and conservation. They make sure to explain things in a simplified
manner and motivate people to do their bit towards bird conservation, maintaining their positive
and friendly tone throughout the prompts.
 3. Create a prompt for large language model, such as ChatGPT, that asks it to perform
an instruction other than answering the questions. These prompts can be asking for
planning a trip, create a workout plan in gym, skills set required for an engineering
graduate or diagnose a disease through some reports. Provide at least three prompts
and fine tune the prompts to get more desired/ precise outputs from your perspective.
Ans:
Prompt 1
Basic Prompt: Make a workout plan for me.
Tuned Prompt: Create a 4-week beginner workout plan for someone who wants to build strength
and tone their body. The person can go to the gym 3 days a week and has no injuries. Include
specific exercises, sets, reps, and rest times.

Sample Output: Workout Plan Overview

 Schedule: 3 days per week (e.g., Monday, Wednesday, Friday)
 Goal: Build strength and tone
 Rest Between Sets: 60-90 seconds
 Progression: Increase weight slightly each week if form is solid

Weeks 1 & 2 – Full-Body Focus (Learning Form + Building Base Strength)

Day 1 – Upper Body Strength
Exercise Sets Reps Rest
Dumbbell Chest Press 3 10-12 60 sec
Seated Row Machine 3 12-15 60 sec
Dumbbell Shoulder Press 3 10-12 60 sec
Dumbbell Bicep Curls 2 12-15 60 sec
Tricep Pushdowns (Cable) 2 12-15 60 sec

Day 2 – Lower Body Strength

Exercise Sets Reps Rest
Leg Press Machine 3 12-15 90 sec
Dumbbell Romanian Deadlift 3 10-12 90 sec
Goblet Squats (with dumbbell) 3 12-15 90 sec
Calf Raises (Machine or Bodyweight) 2 15-20 60 sec
Bodyweight Glute Bridges 2 15 60 sec

Day 3 – Full Body / Core & Conditioning

Exercise Sets Reps Rest
Lat Pulldown Machine 3 12-15 60 sec
Dumbbell Bench Press 3 10-12 60 sec
Dumbbell Step-Ups (each leg) 3 12 60 sec
Plank Hold 3 30-45 sec 60 sec
Russian Twists (with light weight) 3 20 (10 per side) 60 sec

Weeks 3 & 4 – Progressive Overload (Adding Challenge & Refining Technique)

Day 1 – Upper Body Progression

Exercise Sets Reps Rest
Barbell Bench Press (or Dumbbell Press) 3 8-10 90 sec
Seated Row Machine (increase weight) 3 10-12 60 sec
Dumbbell Lateral Raises 3 15 60 sec
Incline Dumbbell Curls 2 12-15 60 sec
Overhead Tricep Extension (Dumbbell) 2 12-15 60 sec

Day 2 – Lower Body Progression

Exercise Sets Reps Rest
Leg Press (heavier weight) 4 10-12 90 sec
Dumbbell Romanian Deadlift 3 10 90 sec
Walking Lunges (each leg) 3 12 90 sec
Standing Calf Raises (weighted if possible) 3 15-20 60 sec
Glute Kickbacks (Cable or Bands) 3 12-15 60 sec

Day 3 – Full Body + Core

Exercise Sets Reps Rest
Assisted Pull-ups or Lat Pulldown 3 10-12 60 sec
Dumbbell Chest Press (increase weight) 3 8-10 60 sec
Bulgarian Split Squats (each leg) 3 10-12 90 sec
Plank with Shoulder Taps 3 20 taps 60 sec
Bicycle Crunches 3 20 reps 60 sec

Warm-up & Cool Down

Before each workout:
 5-10 min light cardio (treadmill, bike, rowing)
 Dynamic stretches (leg swings, arm circles)
After each workout:
 5 min light cardio
 Static stretches (hamstrings, chest, shoulders)

Prompt 2:

Basic prompt: Plan a trip for me to Manali.

Tuned prompt: Plan a 5-day budget-friendly trip to Manali for two people in April. Include
transportation options from Delhi, affordable hotel suggestions, must-see attractions, adventure
activities (like paragliding or river rafting), and local food recommendations. Also, suggest a
day-wise itinerary and estimated costs.

Sample output:
Transport (Delhi to Manali & Return)
 Option: Volvo AC Bus (HRTC/Private)
 Cost: ₹1,500 per person, one way
 Duration: 12-14 hours (overnight)

Stay (Affordable Hotels)

1. Hotel Greenfields – ₹1,500 per night
2. Zostel Manali (Private room) – ₹1,800 per night
 3. Hotel Drilbu – ₹1,400 per night

5-Day Itinerary
Day 1: Arrival + Local Explore
 Arrive & check-in
 Visit Mall Road, Tibetan Monastery, Van Vihar
 Dinner at Café 1947
💰 Approx: ₹4,000 (2 people)

Day 2: Solang Valley Adventure

 Paragliding, Ropeway (₹1,000-₹1,500 per person)
 Visit Hadimba Temple in the evening
 Dinner at Johnson’s Café
💰 Approx: ₹5,500

Day 3: Rohtang Pass (if open) / Hampta Trek

 Shared cab to Rohtang (~₹1,000 per person)
 Snow activities
 Return and relax
💰 Approx: ₹6,000

Day 4: Naggar Castle + Jana Waterfall

 Explore Naggar Castle & Art Gallery
 Himachali lunch at Jana Waterfall Café
💰 Approx: ₹4,500

Day 5: Leisure + Return

 Morning walk by Beas River
 Shopping & souvenirs
 Volvo back to Delhi
💰 Approx: ₹4,000

💸 Total Budget (for 2 people):

₹27,000 - ₹30,000 (including travel, stay, food, activities)

🍛 Food Must-Try:
 Siddu
 Trout Fish
 Momos
 Himachali Thali

Prompt 3:

Basic Prompt: What are the skill set required for a medical student to become successful?

Tuned Prompt: What essential skills should a medical student develop to become a successful
doctor? Please include both technical skills (such as clinical knowledge, diagnostic abilities) and
soft skills (like communication, empathy, and teamwork). Also, suggest practical ways to
develop these skills during medical school and clinical rotations.
Sample Output:

Essential Skills for a Medical Student to Become a Successful Doctor

🩺 Technical Skills
 1. Clinical Knowledge
o Understanding anatomy, physiology, pathology, pharmacology, etc.
o How to develop: Consistent study, active participation in lectures, and self-
assessment tests.
2. Diagnostic Abilities
o Accurately identifying diseases through patient history, examinations, and test
interpretations.
o How to develop: Practice case studies, attend clinical postings regularly, shadow
experienced doctors, and ask questions.
3. Procedural/Practical Skills
o Performing basic procedures (IV insertion, suturing, catheterization).
o How to develop: Hands-on practice in skills labs and supervised procedures
during rotations.
4. Research & Analytical Skills
o Interpreting medical literature, evidence-based practice.
o How to develop: Participate in research projects, learn medical statistics, and
attend journal clubs.

🤝 Soft Skills
1. Communication
o Explaining diagnoses, procedures, and treatment plans clearly to patients and
families.
o How to develop: Practice role-playing patient interviews, participate in group
discussions, and focus on active listening.
2. Empathy & Compassion
o Understanding patients’ emotions and showing genuine care.
o How to develop: Spend time talking to patients beyond their illness; reflect on
patient interactions.
3. Teamwork & Collaboration
o Working effectively with healthcare teams (nurses, specialists, etc.).
o How to develop: Be proactive during ward rounds, respect every role in the team,
and offer help where needed.
4. Decision-Making & Critical Thinking
o Making informed decisions quickly, especially in emergencies.
o How to develop: Engage in clinical simulations, participate in emergency drills,
and discuss complex cases with mentors.
5. Time Management
o Balancing studies, clinical work, and personal life.
o How to develop: Use planners, prioritize tasks, and set realistic daily goals.

📝 Practical Tips During Medical School & Clinical Rotations

 Attend every clinical posting and be actively involved.
 Volunteer for case presentations, surgeries, or procedures.
 Seek feedback from seniors and mentors regularly.
 Join study groups to share knowledge and perspectives.
 Participate in workshops (communication skills, CPR, first aid).
 Maintain a journal to reflect on patient interactions and learning.
*****************************End of Assignment******************************