From Q-Learning to Deep Learning: Understanding the Neural Foundation

If you've been following our AI journey so far, we've explored Q-learning and temporal difference methods in reinforcement learning. But before we can jump into the exciting world of Deep Q-Learning, we need to take a step back and understand the fundamental building blocks: Artificial Neural Networks (ANNs).

Why this detour? Because Deep Q-Learning combines the power of Q-learning with the pattern recognition capabilities of neural networks. Over the next few articles, we'll explore ANNs in detail, and today we're starting with the big question: Why are neural networks suddenly everywhere, and why now?

Let me take you on a journey to understand this fascinating technology that's reshaping our world.

A Brief Walk Down Memory Lane

Imagine this: In 1994, most people didn't even know what the internet was. Fast forward to today, and we can't imagine life without it. Similarly, neural networks and deep learning have been conceptually available for decades, gaining significant attention in the 1980s before mysteriously fading from the spotlight.

But why did this promising technology disappear only to resurface with such tremendous force in recent years?

The answer isn't that neural networks weren't good enough. Rather, the world wasn't ready for them yet. Deep learning needs two critical ingredients to flourish:

1. Massive amounts of data

2. Significant processing power

And until recently, we simply didn't have enough of either.

The Incredible Evolution of Data Storage

Picture this: In 1956, a 5-megabyte hard drive was the size of a small room and required a forklift to transport. Companies would pay $2,500 per month (in 1956 dollars!) just to rent it.

By 1980, things had improved slightly—a 10-megabyte hard drive cost $3,500 to purchase. That's roughly the storage needed for a single high-resolution photo today.

Now in 2025, we carry terabytes of data in our pockets, and cloud storage is essentially free. For perspective:

• 1956 to 1980: Storage capacity doubled (over 24 years)

• 1980 to 2025: Storage capacity increased by millions of times (over 45 years)

This isn't just progress—it's exponential growth. And we're not stopping there! Scientists are now exploring DNA as a storage medium, where potentially all the world's data could fit in just one kilogram of DNA storage.

To put this in everyday terms: Imagine if your house, which could once only store a single book, can now hold every book ever written—with room to spare.

Processing Power: From Rat to Superhuman

Meanwhile, computing power has followed a similar exponential trajectory (often called Moore's Law).

In 2025, an average consumer computer can process information at roughly the speed of a human brain—a milestone that was unimaginable in the 1980s when neural networks first gained attention. Back then, computers had the processing power of perhaps a simple insect's nervous system.

Think about it like this: If the computing power of the 1980s was equivalent to a bicycle, today's computing power is like a supersonic jet.

So What Exactly IS Deep Learning?

Now that we understand why deep learning is taking off now, let's demystify what it actually is.

At its core, deep learning attempts to mimic how the human brain works. Our brains contain approximately 100 billion neurons, each connected to up to 1,000 neighbours. This creates an unimaginably complex network that processes information, learns patterns, and makes decisions.

Deep learning creates an artificial version of this structure called a neural network. Here's how it works in simple terms:

1. Input Layer: This is where we feed in our known information. Think of it like your senses— eyes, ears, nose — taking in data from the world.

2. Hidden Layers: This is where the "deep" in deep learning comes from. While "shallow" learning might have just one hidden layer, deep learning uses many interconnected layers—similar to how your brain has billions of neurons processing information before reaching a conclusion.

3. Output Layer: This produces the prediction or decision, like "this email is spam" or "this photo contains a cat."

A Real-World Example

Imagine you're trying to teach a computer to recognize dogs in photos. Here's how deep learning approaches this problem:

• Input Layer: The pixels from the image (thousands of individual data points)

• Hidden Layers: Multiple layers that progressively identify: Basic edges and shapes, then combinations of shapes (like circles for eyes, triangles for ears), then higher-level features (like furry textures, wet noses), then complete concepts (like different dog breeds).

• Output Layer: The final determination: "Yes, this is a dog" or "No, this is not a dog"

Unlike traditional programming where humans would have to specify exact rules for identifying a dog, deep learning allows the system to discover these patterns itself—much like how a human child learns to recognize animals.

Why This Matters For You

Even if you're not a tech professional, deep learning is already touching your life in countless ways:

• The smartphone voice assistant that understands your questions

• The streaming service that knows exactly what show you might enjoy next

• The photo app that automatically organizes pictures of your family

• The translation service that helps you communicate in foreign countries

• The medical diagnostic tool that's helping your doctor detect diseases earlier

As we continue down this path, we're approaching systems that can think and learn at speeds comparable to humans, opening up possibilities that were once confined to science fiction.

The Future Is Already Here

The neural networks that seemed like an interesting but impractical academic pursuit in the 1980s have finally found their moment. With our vast data reserves and powerful computers, we're witnessing the fulfilment of that early promise.

As someone living through this extraordinary technological transition, you have a front-row seat to one of the most significant shifts in human history—comparable perhaps only to the invention of the printing press, electricity, or the internet itself.

What neural networks will help us discover or create next is limited only by our imagination. And in a world of exponential growth, today's cutting-edge will seem quaint tomorrow.

So the next time you hear about a breakthrough in AI or deep learning, remember: This isn't just the latest tech fad. It's the continuation of a journey that began decades ago, finally enabled by the remarkable progress in data storage and computing power that makes deep learning not just possible, but practical.