Intelligence, Mimicry & Dawn of AI

Poetic transcription on AI learning frameworks, and Human learning

Sujatha Ratnala
6 min readSep 4, 2024

“I think.. Therefore I am..”

We humans are blessed with amazing thinking and intelligence. While we put our intelligence to day to day use, the ancient philosophers, Newtons and Teslas, put their intelligence in eternal quest for knowledge, thereby discovering the patterns and mysteries that wrap this world.

“I seem to have been only like a boy playing on the sea-shore, and diverting myself finding a smoother pebble or a prettier shell, while the great ocean of truth lay all undiscovered before me.” — Isaac Newton

“God is a great mathematician, and he made the world with such accuracy.”

“If you want to find the secrets of the universe, think in terms of energy, frequency and vibration” — Nikola Tesla

This collective knowledge, favoured by time and technological advancements, has led to the dawn of AI.. Patterns and mimicry, with the hood of mathematics at the core. For mathematics is the language of mimicry!

The hard-to-understand AI Algorithms! Each of them is a facet of Intelligence and Learning. From the moment we are born, we are listening and learning all the time.

I do not know much on AI, and yet I’m tickled to write on the intersection of AI and Human learning. The convolution of experiences, the reinforcement of wisdom and the diffusion of knowledge. A poetic and philosophical exploration!

The Learning Elements:
Dataset, Model, Bias
OverFitting UnderFitting

Types of Learning:
Supervised, UnSupervised, Transfer

PART 1 — Decision making
Gradient Descent Algorithm, K Means, Convolution Network, Reinforced Learning.

PART 2 — Creativity
GAN, Diffussion

The Learning Game

Learning, Dataset & Model

We stand on the shoulders of giants, basking in the river of knowledge.

Our emotional side is about experiencing uncompressed life in the ups and downs. While the Logical side based on our attention mechanisms seems to be making inferences, takeaways, adding to the wisdom of things, just like AI.

Let’s take a case study of student’s life! We’ve been there! We have also seen friends, family tread the student life, and have learnt so much from the vast dataset of experiences. This accumulated wisdom forms a model, offering a balcony view of life, guiding us, as we navigate the world.

Similarly, AI learns from a vast dataset of inputs. It applies this learning to make predictions, such as in Real Estate or Cancer cell detection.

The Pitfalls: Bias, Over Fitting, Under Fitting

Despite this wisdom, we fall prey to pitfalls. We may be biased, judgmental not allowing us to understand that special student who does not fit the standard curve, making this an overfitting scenario.

On the other hand, we may underfit, neglecting the symptoms of a sick child and missing crucial cues.

Depending on the training and finetuning, AI algorithm too could fall prey to overfitting underfitting.

Types of Learning

And how did we learn? We were taught by others. Sometimes we picked up the knowledge ourselves. And sometimes we apply learning from one domain to solve the problem in another field.

Supervised Learning

In school, we learn alphabets, numbers and other topics. There is close supervision. We kind of know if we are right or wrong. And over time, we get a hang of it.

Unsupervised Learning

Learning in class vs Learning from Youtube lessons

As infants, listening to the sounds around, imitating and picking up sounds and accents.

Many soft skills of life

Transfer Learning

We live in a connected world and apply learnings from one job to another. In big or small ways, we keep doing this Transfer Learning and navigating the world all the time.

Apply math knowledge of school to real world problems.

Apply knowledge of Indian cooking style to decode Italian cooking.

Knowledge of one language enabling connection with another.

A Scientist gazing at the sky, inspired making a discovery.

PART 1- Decision Making Algorithms

In pursuit of excellence: Gradient Descent Algorithm

In the woods of the mind, in accordance to one’s ability and a known map, the windingtour, the uphill ascends and sometimes a detour to reach that pinnacle goal.

The algorithm travels around the hill till it finds that optimal prediction curve.

Making it Simple: K Means Algorithm

Often, for little creatures that we are, we cannot comprehend much.
It is easier to understand the world by means of comparison, than by leap of absolutes.

I want to be an engineer like my brother
I want to have a big house like Bill Gates
I want to be like my manager..

Role models, golden benchmarks, the standard metric system, all of them aid in simple comprehension.

This world is infinitely vast and hard to comprehend. We can only sample it. In the chapter on Vibhuti Yoga, the unending opulance of the world is conveyed in the language of superlatives. Much like the K Role Models.

In intelligence, game of dice
In animals, Im the lion and elephant
In birds, the hawk
In seasons, the spring
In mountains, the Himalayas..
In rivers, the Ganges..

Our friends, our books the K models and life shining through them.

In the AI world, instead of representing tremendous GPS-like coordinates for each data, they are placed relative to the K role-model coordinates

And how does AI use this K means algorithm? Compute 5 trending shopping insights at Amazon. Viola! — Gadgets, Fashion, Home Utility, School Supply, Groceries

Scaling Patterns in a Convoluted Journey

As Nikola Tesla once said, “If you want to find the secrets of the universe, think in terms of frequency and vibration.”

Like a maze, life is rich and convoluted with hidden patterns and paths that twist and turn.

Similarly, CNN imaging algorithm navigate complexity by scaling through layers, uncovering hidden patterns, and detecting objects within the labyrinth of data.

Reinforced Learning RL

When we receive good feedback -such as scoring well, getting compliments, or reaching milestones, it is like a reward signal telling — ‘You are doing good’ and this feeling further reinforcing, motivating to do better.

Much like AI agent adapting its strategy and path based on the positive and negative rewards.

PART2 — Beyond Classification: Unleashing Creativity: GenAI

Creative expression is the very soul of Human experience.

After hearing an intense lecture, an impulse to blog
After eating a fantastic dish at a restaurant, an impulse to recreate

Behind the hood of ChatGPT, DELL-E and other GenAI works.

In close harmony: GAN

After 30 years of friendship, the generator and processor of data are in sync and harmony.. Friend X narrates an event and Friend Y vividly imagines it.

Keep reading an author, keep watching a series, and soon you pick up cues

The creative prowess: Diffusion Algorithm

Learning and Creativity is a journey of exploitation and exploration. Exploitation of past learnings and attention mechanisms. And Exploration of newer ideas.

Like a master sculptor, crafting stunning images with ease. He leverages his past knowledge. Chipping the rough edges through layers of chiselling, using eye of judgement.. Sometimes, taking a detour and doing again.. In case of AI, the judgement of contour.

Similarly, our dreams can be seen as creative works of Diffusion, combining elements of prior works.

This line comes to my mind. ‘Existence precedes Essence.’ From a block of unpolished existence, layers of order and discretion kick in, thereby, churning the essence of creativity.

Human Experience and AI..

Human intuition and emotions are unparalleled. But the quest for learning and our logical sides made me think of parallels with AI.

In a poetic sense, we are perhaps like self-driving cars, but eager to explore the algorithms that make it work.

We are perhaps like ChatGPTs.. As we explore the mysteries, decoding the language of the world. Pretrained by experiences, we develop a ‘large language model’ in our mind and expressing presenting ourselves.

--

--

Sujatha Ratnala

I write.. I weave.. I walk.. कवयामि.. वयामि.. यामि.. Musings on Patterns, Science, Linguistics, Sanskrit et al..