We’ve all had small talk. Whether you love it or hate it.

“Hey, how’s it going?”

“How about that weather?”

“Do your gradients even descend, bro?”

Sure, that last one’s questionable, but you get what I mean. Boring words just to pass the time.

But we’ve also had much deeper conversations. Can you think of someone who just captures your attention when you talk to them? Someone who makes time seem non-existent?

Well, believe it or not, machine learning is the small-talker. And deep learning is the meaningful conversationalist.

And in the world of AI…both small-talk and deep conversations hold weight.

The relationship between AI, machine learning, and deep learning looks like this:

Artificial Intelligence is the big, all-encompassing idea of teaching machines to be “smart” like humans. It’s the more generalized field of making computers behave more intelligently, whether through hard-coded rules or specialized reasoning.

Machine Learning is a more focused subset of AI. It’s all about getting computers to learn on their own by identifying patterns in data.

Deep Learning is an even smaller, focused subset within ML. It uses something called Neural Networks (yet another subset in itself) to tackle some really complex problems. It’s like ML, but with extra IQ points, and it’s the not-so-secret sauce behind things like image recognition and speech understanding,

So in short, AI is the big idea, ML is the practical approach, and DL is the one with extra brain cells and a caffeine habit. ☕🧠

Let’s talk about things strictly related to machine learning, to differentiate it from deep learning. To really “get” machine learning, there are some essentials you can’t skip.

Data Preprocessing: This is cleaning and organizing raw data so your machine learning model can learn it instead of getting confused. Data in ML is kind of like a high-maintenance partner. 🤷‍♂️

Feature Engineering: This is choosing and creating the right inputs (features) for your machine learning model to help it make accurate predictions. It’s kind of like dressing your data in its nicest outfit so the model will actually pay attention.

Model Selection: This is the process of choosing the most appropriate algorithm or model architecture for a given problem/dataset. Don’t use a sledgehammer when a screwdriver will do just fine, right? 😉

Evaluation Metrics: These are metrics to measure how well a model performs on a given task. Oftentimes, these metrics use values such as accuracy, precision, recall, or F1 Score. Think of this as your model’s report card.

As you can see, machine learning is the “small-talker” in the AI family…but that makes it powerful. Sometimes the clear, straightforward chatter gets the job done best.

Sometimes small talk is appropriate! If you literally just met someone, you probably wouldn’t immediately jump into debating them on philosophy.

Or maybe you would. In that case, more power to you, Cornel West.

If ML is the small-talker who wants to know how your day is going…then Deep Learning is the one who wants to unpack your life story…pixel by pixel, sentence by sentence, soundwave by soundwave. It’s the critical thinker of the AI family.

Think of deep learning as machine learning on steroids: instead of humans defining features, DL automatically discovers them through large neural networks (digital brains in DL).

Neural Networks:

The main structure: nodes pass information through layers, and activation functions decide what gets passed along. Read more about them here.

Architecture Types:

Different neural networks have different “shapes” for different jobs. For instance, CNN (Convolutional Neural Network), which is used for spotting patterns in images, RNN (Recurrent Neural Network), for sequential data, and Transformers, which process sequence data all at once.

Backpropagation:

This is how the neural network learns. You make a guess, see how far off you were, then adjust the wires (called weights in DL), so you get smarter over time.

In short, DL is useful for tackling problems that are too intricate for traditional ML.

As you can see, ML and DL are part of the same family — DL being a specialized field within ML.

Use ML when you have smaller, structured datasets, need more interpretability, and want quicker and lighter models. Use DL when you’re dealing with massive amounts of complex data (like images, text, or audio).

In short, ML is the versatile generalist, DL is the heavy-duty specialist.

And with that… may your training always converge, and your predictions be ever in your favor. 🎯

Tech moves fast, but you're still playing catch-up?

That's exactly why 100K+ engineers working at Google, Meta, and Apple read The Code twice a week.

Here's what you get:

All delivered twice a week in just 2 short emails.

Join 100K+ engineers