AI and Large Language Models: A Story of Language

AI: A Story of
Language
Suresh Peiris
Co-Founder | Inforwaves &
Articom.io

Once Upon a
Time... The Dream
of Intelligent
Machines

Early Chapters
For decades, AI was a story of
slow, steady progress, with
breakthroughs in logic-based
systems and early machine
learning.
It was powerful but often confined
to specific, narrow tasks.
The Beginning
Our story begins not with silicon,
but with an idea: the dream of
creating machines that can think,
learn, and reason like humans.
This is the core quest of Artificial
Intelligence.
The Plot Twist
In the last decade, three key
elements converged to change
everything:
● Big Data: An ocean of information to
learn from.
● Powerful GPUs: The engine to
process this data at incredible
speeds.
● Advanced Algorithms: A new way
to think about learning.
The Evolution…

Today's Chapter:
We are living through the
most exciting chapter yet—
the age of Generative AI and
autonomous Agents.

To build intelligent machines, we first
needed to teach them our most powerful
tool: language.
A New Language for Machines -
The Rise of LLMs
Enter Large Language Models (LLMs).
Think of them as AI apprentices that have
read a library the size of the entire
internet.
They are built on complex neural networks,
learning not just words, but the relationships,
context, and nuance between them.
This is the foundational technology behind the
chatbots, writing assistants, and code helpers
we see today, making human-computer
interaction more natural than ever before.

The Spark of Genius -
"Attention Is All You
Need"
The Old Way: Word-by-word processing; struggled with
long-range dependencies, forgetting early parts of
sentences.
The Breakthrough Idea (Transformer Architecture):
Introduced "Attention Mechanism"; allowed
simultaneous processing of all words and identification
of important contextual elements.
Why It Was Revolutionary: Enabled training of massive
and powerful language models; made AI more
parallelizable, efficient, and scalable; foundation for
modern AI like GPT and Claude.

What is
an LLM?
LLMs Explained
[...] [...] [...]
[...]
0.02
0.03
0.9 0.01 0.0 …
Dogs Rain Drops Fish Wind …
and
cats
raining
It’s

Roses are red,
Violets are blue,
Sugar is sweet,
LLMs Explained

for(var i = 0;; i < 10; i++) {

Modern LLMs are
large
LLMs Explained

Classic Natural
Language Problems
LLMs Explained
Entity extraction Classification Summarization
Sentiment Analysis Translation …

LLMs let us
prototype fast
LLMs Explained

AI Studio
Endpoints
Prompting
https://ai.google.dev/docs/prompt_best_practices

Anthropic's Claude
3.5 Sonnet:
The meticulous scholar. It sets
new benchmarks in graduate-level
reasoning, complex code
generation, and interpreting
charts and graphs, all while being
incredibly fast and cost-effective.
OpenAI's GPT-4o
A master storyteller that isn't
just limited to text. It seamlessly
understands and discusses audio,
images, and text together,
making interactions fluid and
intuitive.
The Open-Source
Heroes:
Models like DeepSeek-V2 and
Llama 3 are champions of
accessibility. They offer powerful
capabilities to everyone, fostering
a global community of builders.
An Explosion of Creativity

This new AI
intelligence isn't just
for understanding;
it's for creating.
Image & Video
Generation:
Tools like Midjourney, Stable
Diffusion, and OpenAI's Sora,
Veo3 are turning simple text
prompts into breathtaking visuals
and video clips. They are our new
paintbrushes for the digital age.
This is a story of
democratizing
creativity, allowing
anyone to visualize
their imagination.
From Words to Worlds -
The Art of Generation

Demo Time!
1. Google’s Veo 3
2. Gemini’s Multi Language Capabilities
3. Via Vertex AI - Code Level

Enter AI Agents:
These are not just chatbots.
They are digital "employees" that
can understand a complex goal
(e.g., "plan a business trip to
Tokyo"), break it down into steps,
and use tools (like booking
websites, calendars, and email) to
get it done.
The Rise of the Agents -
AI Gets a Job 🤖
How They Talk to Each
Other:
For agents to collaborate, they
need a common language.
Protocols like A2A (Agent-to-
Agent) and the Agent
Communication Protocol (ACP)
are being developed to create a
standard way for them to
communicate and work together.
The Big Picture:
This is about automating entire
workflows, creating a truly smart
and interconnected digital
ecosystem.

The Power of Open Source: The future of AI is being built in the open.
● Popular Repos to Watch: Check out repositories for tools like vLLM (for running
models efficiently) and LangChain or LlamaIndex (for building applications on
top of LLMs). These projects are crucial for innovation.
The Challenge of Data Privacy: With great power comes great responsibility. As
we use these models, we must address critical questions:
● How do we ensure our personal data isn't memorized?
● How do we build fair and unbiased systems?
The Path Forward: Techniques like differential privacy and a strong focus on AI
ethics are not optional—they are essential for building a future with AI that we can all
trust.
Building the Future Together -
Open Source & Ethics

Customer Support:
AI agents are providing 24/7, instant, and
helpful support. Eg: Articom.io
The Impact -
A New World of Opportunity
Hiring:
AI is helping find the best talent faster and
more fairly.
Finance:
AI is detecting fraud and making financial
advice more accessible.
Education:
AI is creating personalized learning paths that help
every student succeed. Eg: NotebookLM

The Prologue (Beginner):
1. Learn Python - the language of AI.
2. Take a foundational course, like Andrew Ng's "Machine Learning" on Coursera.
3. Practice on Kaggle by working on real-world data challenges.
The Adventure (Intermediate):
4. Master a deep learning framework like PyTorch or TensorFlow.
5. Build a project! Use a repo like LangChain to create your first LLM-powered
application.
6. Use my sample repo on Google ADK to learn more about AI.
Writing Your Own Chapter - Your
Path into AI
️ 🗺️

● The Architect (ML Engineer): The one who builds and deploys the AI
models.
● The Explorer (Data Scientist): The one who finds valuable insights in
data.
● The Pioneer (Research Scientist): The one who writes the next chapter by
pushing the boundaries of what's possible.
● The Storyteller (AI Product Manager): The one who shapes the vision
for how AI can solve real-world problems.
● The Guardian (AI Ethicist): The one who ensures the story unfolds
responsibly and fairly for everyone.
Your Career in the AI Story -
The Roles Awaiting You 💼

The AI story is being written
live, every single day. The most
exciting chapters are still
ahead.
What part will you play?

- Suresh Peiris
@sureshmichael
Q&A

AI and Large Language Models: A Story of Language

More Related Content

Similar to AI and Large Language Models: A Story of Language

More from Suresh Peiris

Recently uploaded

AI and Large Language Models: A Story of Language