As I embark on my journey into the realm of Large Language Models (LLMs), I'm discovering fascinating applications that redefine how we work. From leveraging AI like GitHub Co-pilot for coding to harnessing ChatGPT for email composition, the possibilities seem endless. However, I'm also intrigued by the limitations posed by these solutions being Software-as-a-Service (SaaS) products, lacking full control. In this blog, I delve into a topic often overlooked: Swagger API documentation. Join me as I explore the potential of local setups and document my journey. As a newcomer to the world of LLMs, I seek to uncover practical applications and share insights along the way."
Join us in a groundbreaking series as we delve into the world of Large Language Models (LLMs), examining both Software-as-a-Service (SaaS) solutions and local setups. Together, we'll compare their capabilities and uncover the potential of crafting our own local LLM for a unique purpose: Swagger documentation. This uncharted territory promises to revolutionize how we document APIs. Don't miss out on this pioneering exploration!.
My goal with this article is to provide a path to understand the LLM and potentially help software engineers to use LLM to full potential into their projects.
What is Machine Learning and why do we need to have Machine Learning in the first place?
Today as our dependency on the technology is growing and to surpass the intelligence of the humans, we started to use Machine Learning to predict / classify things to automate the process and eliminating the human intervention. As we know all of the nature and all the processes can be represented as a function which takes some input and produce some output.
"Functions Describes the world"
Quote from the introduction to Thomas Garrity's "Mathematical Maturity"
And the fact that the computers are good at crunching numbers we can use machine learning to basically approximate any function with appropriate data to train on.
Let me define Machine Learning in Leman terms
Machine learning is like teaching a computer to learn from examples rather than programming it with specific instructions. Just like how we learn from experiences, machine learning algorithms analyze data to recognize patterns and make predictions or decisions. Imagine you're teaching a child to differentiate between animals. You show them various animals like cat, dogs, and cows, explaining their unique features like appearance, sound they make, diet, etc. Over time, the child learns to identify each animal correctly without explicit instructions. Similarly, in machine learning, algorithms learn from data to perform tasks such as recognizing spam emails, recommending movies, or even driving cars. It's about enabling computers to learn and improve from data, making them more intelligent and adaptable.
Different areas of machine learning:
- Statistical Machine Learning
- Deep learning (Neural Networks)
- Reinforcement Learning
AI is the sub-domain of machine learning which covers the fields like computer vision and NLP. For me, I started learning AI in my Freshmen year in my undergrad degree and the first project I built was Clone of AlexNet.
As said by Jensen Huang, CEO of Nvidia in GTX 2024, in 2012 we gave the input of image 32X32 pixels and use to get one word as a answer, the potential was clear. Today, we give that one word/vector to the AI model and it generates millions of the pixel back, thats the age of Generative AI we are heading towards.
Hope you will enjoy the series and share with your friends who wants to build amazing understanding of LLM.
Tentative Articles:(Titles might be different)
- Exploring the realm of LLM - what and why?
- LLM SAAS offerings - OpenAI, Cloud and HuggingFace models
- LLM Local - deploying the LLM in local environment.
Top comments (0)