What is Neural Network ? From AI to Generative AI

Lets first understand field of Artificial Intelligence (AI) and how it is related with machine learning, deep learning and generative AI.

The Evolution of Artificial Intelligence

The overarching goal of artificial intelligence is to teach machines to perform tasks normally done by humans. The evolutionary progression toward this goal is divided into four major stages.

1. Traditional Artificial Intelligence

Traditional Artificial Intelligence relies on explicit “IF-Then Rules” to tell a computer exactly what to do. This approach proved difficult because it is impossible to manually cover all potential situations.

2. Machine Learning

To solve this, Machine Learning emerged, allowing computers to learn relationships between inputs and outputs by processing large amounts of examples using statistical techniques like Linear Regression, Random Forests, SVM, and Neural Networks. While these machine learning models worked well with structured data, they struggled to process unstructured data like images, video, text, and audio.

3. Deep Learning

This limitation required a shift toward Deep Learning, which successfully created “smart representations” of complex unstructured data but still lacked the ability to create new content.

4. Generative AI

Generative AI overcomes these previous boundaries, giving machines the ability to successfully generate entirely new outputs.

Goal: Teach machines to perform tasks normally done by humans
Artificial
Intelligence
→ Machine
Learning
→ Deep
Learning
→ Generative
AI
(Traditional)
↓
{ IF – Then Rules }
↓
( But Difficult )
{ Impossible to cover all situations }
[ Explicitly telling computer what to do ]
→ Give lot of examples to the machine (computer)
→ Learn/establish relationship between inputs and outputs
( Statistical Techniques )
e.g. Linear Regression, Random Forests, SVM, Neural networks
⇓
( Machine Learning model )
( Worked with structured data )
What about unstructured data such as images, videos, text, audio?
{ Better representations of data }
Smart representations
{ Couldn’t generate } →

What is a Neural Network?

In the evolution of artificial intelligence, Neural Networks serve as the critical bridge that transitions standard Machine Learning into Deep Learning and Generative AI.

The Biological Inspiration

Inspired by the biological neurons in the human brain, a neural network is a computational model designed to recognize complex patterns. Rather than relying on traditional “IF-Then” rules, it processes data through interconnected nodes called artificial neurons.

The Core Mechanism

The network stacks these simple neurons in layers: an input layer for raw data, hidden processing layers, and an output layer for the final prediction. As examples pass through the network, it establishes relationships by adjusting internal mathematical dials known as weights and biases, to accurately map inputs to outputs.

Scaling into Deep Learning

While early machine learning handled structured data well, it struggled with complex unstructured data like video and audio. By stacking many hidden layers together, a standard neural network becomes a “Deep Neural Network.” This depth allows Deep Learning to extract the necessary “smart representations” of unstructured data, which ultimately powers the creative capabilities of modern Generative AI.

References:

  • R. Ramakrishnan, “Introduction to Neural Networks and Deep Learning,” 2024.

‌

error: